[Vendor] blevesearch v0.8.1 -> v1.0.7 (#11360)

* Update blevesearch v0.8.1 -> v1.0.7

* make vendor

Co-authored-by: zeripath <art27@cantab.net>
mj
6543 4 years ago committed by GitHub
parent a44854c287
commit fdf750e4d4
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23

@ -19,15 +19,11 @@ require (
gitea.com/macaron/toolbox v0.0.0-20190822013122-05ff0fc766b7
github.com/BurntSushi/toml v0.3.1
github.com/PuerkitoBio/goquery v1.5.0
github.com/RoaringBitmap/roaring v0.4.21 // indirect
github.com/RoaringBitmap/roaring v0.4.23 // indirect
github.com/bgentry/speakeasy v0.1.0 // indirect
github.com/blevesearch/bleve v0.8.1
github.com/blevesearch/blevex v0.0.0-20180227211930-4b158bb555a3 // indirect
github.com/blevesearch/go-porterstemmer v1.0.2 // indirect
github.com/blevesearch/segment v0.0.0-20160915185041-762005e7a34f // indirect
github.com/blevesearch/bleve v1.0.7
github.com/boombuler/barcode v0.0.0-20161226211916-fe0f26ff6d26 // indirect
github.com/couchbase/gomemcached v0.0.0-20191004160342-7b5da2ec40b2 // indirect
github.com/couchbase/vellum v0.0.0-20190829182332-ef2e028c01fd // indirect
github.com/cznic/b v0.0.0-20181122101859-a26611c4d92d // indirect
github.com/cznic/mathutil v0.0.0-20181122101859-297441e03548 // indirect
github.com/cznic/strutil v0.0.0-20181122101858-275e90344537 // indirect
@ -36,7 +32,6 @@ require (
github.com/dustin/go-humanize v1.0.0
github.com/editorconfig/editorconfig-core-go/v2 v2.1.1
github.com/emirpasic/gods v1.12.0
github.com/etcd-io/bbolt v1.3.3 // indirect
github.com/ethantkoenig/rupture v0.0.0-20180203182544-0a76f03a811a
github.com/facebookgo/ensure v0.0.0-20160127193407-b4ab57deab51 // indirect
github.com/facebookgo/stack v0.0.0-20160209184415-751773369052 // indirect
@ -53,7 +48,7 @@ require (
github.com/gobwas/glob v0.2.3
github.com/gogs/chardet v0.0.0-20191104214054-4b6791f73a28
github.com/gogs/cron v0.0.0-20171120032916-9f6c956d3e14
github.com/golang/protobuf v1.4.0 // indirect
github.com/golang/protobuf v1.4.1 // indirect
github.com/google/go-github/v24 v24.0.1
github.com/gorilla/context v1.1.1
github.com/hashicorp/go-retryablehttp v0.6.6 // indirect
@ -95,10 +90,9 @@ require (
github.com/sergi/go-diff v1.1.0
github.com/shurcooL/httpfs v0.0.0-20190527155220-6a4d4a70508b // indirect
github.com/shurcooL/vfsgen v0.0.0-20181202132449-6a9ea43bcacd
github.com/steveyen/gtreap v0.0.0-20150807155958-0abe01ef9be2 // indirect
github.com/stretchr/testify v1.4.0
github.com/tecbot/gorocksdb v0.0.0-20181010114359-8752a9433481 // indirect
github.com/tinylib/msgp v1.1.1 // indirect
github.com/tinylib/msgp v1.1.2 // indirect
github.com/tstranex/u2f v1.0.0
github.com/unknwon/cae v1.0.0
github.com/unknwon/com v1.0.1
@ -109,11 +103,10 @@ require (
github.com/yohcop/openid-go v0.0.0-20160914080427-2c050d2dae53
github.com/yuin/goldmark v1.1.25
github.com/yuin/goldmark-meta v0.0.0-20191126180153-f0638e958b60
go.etcd.io/bbolt v1.3.3 // indirect
golang.org/x/crypto v0.0.0-20200302210943-78000ba7a073
golang.org/x/net v0.0.0-20200324143707-d3edc9973b7e
golang.org/x/oauth2 v0.0.0-20200107190931-bf48bf16ab8d
golang.org/x/sys v0.0.0-20200323222414-85ca7c5b95cd
golang.org/x/sys v0.0.0-20200509044756-6aff5f38e54f
golang.org/x/text v0.3.2
golang.org/x/time v0.0.0-20200416051211-89c76fbcd5d1 // indirect
golang.org/x/tools v0.0.0-20200325010219-a49f79bcc224

@ -56,6 +56,8 @@ github.com/PuerkitoBio/urlesc v0.0.0-20170810143723-de5bf2ad4578 h1:d+Bc7a5rLufV
github.com/PuerkitoBio/urlesc v0.0.0-20170810143723-de5bf2ad4578/go.mod h1:uGdkoq3SwY9Y+13GIhn11/XLaGBb4BfwItxLd5jeuXE=
github.com/RoaringBitmap/roaring v0.4.21 h1:WJ/zIlNX4wQZ9x8Ey33O1UaD9TCTakYsdLFSBcTwH+8=
github.com/RoaringBitmap/roaring v0.4.21/go.mod h1:D0gp8kJQgE1A4LQ5wFLggQEyvDi06Mq5mKs52e1TwOo=
github.com/RoaringBitmap/roaring v0.4.23 h1:gpyfd12QohbqhFO4NVDUdoPOCXsyahYRQhINmlHxKeo=
github.com/RoaringBitmap/roaring v0.4.23/go.mod h1:D0gp8kJQgE1A4LQ5wFLggQEyvDi06Mq5mKs52e1TwOo=
github.com/Shopify/sarama v1.19.0/go.mod h1:FVkBWblsNy7DGZRfXLU0O9RCGt5g3g3yEuWXgklEdEo=
github.com/Shopify/toxiproxy v2.1.4+incompatible/go.mod h1:OXgGpZ6Cli1/URJOF1DMxUHB2q5Ap20/P/eIdh4G0pI=
github.com/Unknwon/com v0.0.0-20190321035513-0fed4efef755/go.mod h1:voKvFVpXBJxdIPeqjoJuLK+UVcRlo/JLjeToGxPYu68=
@ -83,14 +85,22 @@ github.com/beorn7/perks v1.0.1 h1:VlbKKnNfV8bJzeqoa4cOKqO6bYr3WgKZxO8Z16+hsOM=
github.com/beorn7/perks v1.0.1/go.mod h1:G2ZrVWU2WbWT9wwq4/hrbKbnv/1ERSJQ0ibhJ6rlkpw=
github.com/bgentry/speakeasy v0.1.0 h1:ByYyxL9InA1OWqxJqqp2A5pYHUrCiAL6K3J+LKSsQkY=
github.com/bgentry/speakeasy v0.1.0/go.mod h1:+zsyZBPWlz7T6j88CTgSN5bM796AkVf0kBD4zp0CCIs=
github.com/blevesearch/bleve v0.8.1 h1:20zBREtGe8dvBxCC+717SaxKcUVQOWk3/Fm75vabKpU=
github.com/blevesearch/bleve v0.8.1/go.mod h1:Y2lmIkzV6mcNfAnAdOd+ZxHkHchhBfU/xroGIp61wfw=
github.com/blevesearch/blevex v0.0.0-20180227211930-4b158bb555a3 h1:U6vnxZrTfItfiUiYx0lf/LgHjRSfaKK5QHSom3lEbnA=
github.com/blevesearch/blevex v0.0.0-20180227211930-4b158bb555a3/go.mod h1:WH+MU2F4T0VmSdaPX+Wu5GYoZBrYWdOZWSjzvYcDmqQ=
github.com/blevesearch/go-porterstemmer v1.0.2 h1:qe7n69gBd1OLY5sHKnxQHIbzn0LNJA4hpAf+5XDxV2I=
github.com/blevesearch/go-porterstemmer v1.0.2/go.mod h1:haWQqFT3RdOGz7PJuM3or/pWNJS1pKkoZJWCkWu0DVA=
github.com/blevesearch/segment v0.0.0-20160915185041-762005e7a34f h1:kqbi9lqXLLs+zfWlgo1PIiRQ86n33K1JKotjj4rSYOg=
github.com/blevesearch/segment v0.0.0-20160915185041-762005e7a34f/go.mod h1:IInt5XRvpiGE09KOk9mmCMLjHhydIhNPKPPFLFBB7L8=
github.com/blevesearch/bleve v1.0.7 h1:4PspZE7XABMSKcVpzAKp0E05Yer1PIYmTWk+1ngNr/c=
github.com/blevesearch/bleve v1.0.7/go.mod h1:3xvmBtaw12Y4C9iA1RTzwWCof5j5HjydjCTiDE2TeE0=
github.com/blevesearch/blevex v0.0.0-20190916190636-152f0fe5c040 h1:SjYVcfJVZoCfBlg+fkaq2eoZHTf5HaJfaTeTkOtyfHQ=
github.com/blevesearch/blevex v0.0.0-20190916190636-152f0fe5c040/go.mod h1:WH+MU2F4T0VmSdaPX+Wu5GYoZBrYWdOZWSjzvYcDmqQ=
github.com/blevesearch/go-porterstemmer v1.0.3 h1:GtmsqID0aZdCSNiY8SkuPJ12pD4jI+DdXTAn4YRcHCo=
github.com/blevesearch/go-porterstemmer v1.0.3/go.mod h1:angGc5Ht+k2xhJdZi511LtmxuEf0OVpvUUNrwmM1P7M=
github.com/blevesearch/mmap-go v1.0.2 h1:JtMHb+FgQCTTYIhtMvimw15dJwu1Y5lrZDMOFXVWPk0=
github.com/blevesearch/mmap-go v1.0.2/go.mod h1:ol2qBqYaOUsGdm7aRMRrYGgPvnwLe6Y+7LMvAB5IbSA=
github.com/blevesearch/segment v0.9.0 h1:5lG7yBCx98or7gK2cHMKPukPZ/31Kag7nONpoBt22Ac=
github.com/blevesearch/segment v0.9.0/go.mod h1:9PfHYUdQCgHktBgvtUOF4x+pc4/l8rdH0u5spnW85UQ=
github.com/blevesearch/snowballstem v0.9.0 h1:lMQ189YspGP6sXvZQ4WZ+MLawfV8wOmPoD/iWeNXm8s=
github.com/blevesearch/snowballstem v0.9.0/go.mod h1:PivSj3JMc8WuaFkTSRDW2SlrulNWPl4ABg1tC/hlgLs=
github.com/blevesearch/zap/v11 v11.0.7 h1:nnmAOP6eXBkqEa1Srq1eqA5Wmn4w+BZjLdjynNxvd+M=
github.com/blevesearch/zap/v11 v11.0.7/go.mod h1:bJoY56fdU2m/IP4LLz/1h4jY2thBoREvoqbuJ8zhm9k=
github.com/blevesearch/zap/v12 v12.0.7 h1:y8FWSAYkdc4p1dn4YLxNNr1dxXlSUsakJh2Fc/r6cj4=
github.com/blevesearch/zap/v12 v12.0.7/go.mod h1:70DNK4ZN4tb42LubeDbfpp6xnm8g3ROYVvvZ6pEoXD8=
github.com/boombuler/barcode v0.0.0-20161226211916-fe0f26ff6d26 h1:NGpwhs9FOwddM6TptNrq2ycby4s24TcppSe5uG4DA/Q=
github.com/boombuler/barcode v0.0.0-20161226211916-fe0f26ff6d26/go.mod h1:paBWMcWSl3LHKBqUq+rly7CNSldXjb2rDl3JlRe0mD8=
github.com/bradfitz/gomemcache v0.0.0-20190329173943-551aad21a668 h1:U/lr3Dgy4WK+hNk4tyD+nuGjpVLPEHuJSFXMw11/HPA=
@ -108,6 +118,7 @@ github.com/coreos/go-oidc v2.1.0+incompatible/go.mod h1:CgnwVTmzoESiwO9qyAFEMiHo
github.com/coreos/go-semver v0.2.0/go.mod h1:nnelYz7RCh+5ahJtPPxZlU+153eP4D4r3EedlOD2RNk=
github.com/coreos/go-systemd v0.0.0-20190321100706-95778dfbb74e/go.mod h1:F5haX7vjVVG0kc13fIWeqUViNPyEJxv/OmvnBo0Yme4=
github.com/coreos/pkg v0.0.0-20180928190104-399ea9e2e55f/go.mod h1:E3G3o1h8I7cfcXa63jLwjI0eiQQMgzzUDFVpN/nH/eA=
github.com/couchbase/ghistogram v0.1.0/go.mod h1:s1Jhy76zqfEecpNWJfWUiKZookAFaiGOEoyzgHt9i7k=
github.com/couchbase/gomemcached v0.0.0-20190515232915-c4b4ca0eb21d h1:XMf4E1U+b9E3ElF0mjvfXZdflBRZz4gLp16nQ/QSHQM=
github.com/couchbase/gomemcached v0.0.0-20190515232915-c4b4ca0eb21d/go.mod h1:srVSlQLB8iXBVXHgnqemxUXqN6FCvClgCMPCsjBDR7c=
github.com/couchbase/gomemcached v0.0.0-20191004160342-7b5da2ec40b2 h1:vZryARwW4PSFXd9arwegEywvMTvPuXL3/oa+4L5NTe8=
@ -116,8 +127,9 @@ github.com/couchbase/goutils v0.0.0-20190315194238-f9d42b11473b h1:bZ9rKU2/V8sY+
github.com/couchbase/goutils v0.0.0-20190315194238-f9d42b11473b/go.mod h1:BQwMFlJzDjFDG3DJUdU0KORxn88UlsOULuxLExMh3Hs=
github.com/couchbase/goutils v0.0.0-20191018232750-b49639060d85 h1:0WMIDtuXCKEm4wtAJgAAXa/qtM5O9MariLwgHaRlYmk=
github.com/couchbase/goutils v0.0.0-20191018232750-b49639060d85/go.mod h1:BQwMFlJzDjFDG3DJUdU0KORxn88UlsOULuxLExMh3Hs=
github.com/couchbase/vellum v0.0.0-20190829182332-ef2e028c01fd h1:zeuJhcG3f8eePshH3KxkNE+Xtl53pVln9MOUPMyr/1w=
github.com/couchbase/vellum v0.0.0-20190829182332-ef2e028c01fd/go.mod h1:xbc8Ff/oG7h2ejd7AlwOpfd+6QZntc92ygpAOfGwcKY=
github.com/couchbase/moss v0.1.0/go.mod h1:9MaHIaRuy9pvLPUJxB8sh8OrLfyDczECVL37grCIubs=
github.com/couchbase/vellum v1.0.1 h1:qrj9ohvZedvc51S5KzPfJ6P6z0Vqzv7Lx7k3mVc2WOk=
github.com/couchbase/vellum v1.0.1/go.mod h1:FcwrEivFpNi24R3jLOs3n+fs5RnuQnQqCLBJ1uAg1W4=
github.com/couchbaselabs/go-couchbase v0.0.0-20190708161019-23e7ca2ce2b7 h1:1XjEY/gnjQ+AfXef2U6dxCquhiRzkEpxZuWqs+QxTL8=
github.com/couchbaselabs/go-couchbase v0.0.0-20190708161019-23e7ca2ce2b7/go.mod h1:mby/05p8HE5yHEAKiIH/555NoblMs7PtW6NrYshDruc=
github.com/cpuguy83/go-md2man v1.0.10/go.mod h1:SmD6nW6nTyfqj6ABTjUi3V3JVMnlJmwcJI5acqYI6dE=
@ -153,8 +165,6 @@ github.com/edsrzf/mmap-go v1.0.0 h1:CEBF7HpRnUCSJgGUb5h1Gm7e3VkmVDrR8lvWVLtrOFw=
github.com/edsrzf/mmap-go v1.0.0/go.mod h1:YO35OhQPt3KJa3ryjFM5Bs14WD66h8eGKpfaBNrHW5M=
github.com/emirpasic/gods v1.12.0 h1:QAUIPSaCu4G+POclxeqb3F+WPpdKqFGlw36+yOzGlrg=
github.com/emirpasic/gods v1.12.0/go.mod h1:YfzfFFoVP/catgzJb4IKIqXjX78Ha8FMSDh3ymbK86o=
github.com/etcd-io/bbolt v1.3.3 h1:gSJmxrs37LgTqR/oyJBWok6k6SvXEUerFTbltIhXkBM=
github.com/etcd-io/bbolt v1.3.3/go.mod h1:ZF2nL25h33cCyBtcyWeZ2/I3HQOfTP+0PIEvHjkjCrw=
github.com/ethantkoenig/rupture v0.0.0-20180203182544-0a76f03a811a h1:M1bRpaZAn4GSsqu3hdK2R8H0AH9O6vqCTCbm2oAFGfE=
github.com/ethantkoenig/rupture v0.0.0-20180203182544-0a76f03a811a/go.mod h1:MkKY/CB98aVE4VxO63X5vTQKUgcn+3XP15LMASe3lYs=
github.com/facebookgo/ensure v0.0.0-20160127193407-b4ab57deab51 h1:0JZ+dUmQeA8IIVUMzysrX4/AKuQwWhV2dYQuPZdvdSQ=
@ -292,6 +302,8 @@ github.com/golang/protobuf v1.4.0-rc.2/go.mod h1:LlEzMj4AhA7rCAGe4KMBDvJI+AwstrU
github.com/golang/protobuf v1.4.0-rc.4.0.20200313231945-b860323f09d0/go.mod h1:WU3c8KckQ9AFe+yFwt9sWVRKCVIyN9cPHBJSNnbL67w=
github.com/golang/protobuf v1.4.0 h1:oOuy+ugB+P/kBdUnG5QaMXSIyJ1q38wWSojYCb3z5VQ=
github.com/golang/protobuf v1.4.0/go.mod h1:jodUvKwWbYaEsadDk5Fwe5c77LiNKVO9IDvqG2KuDX0=
github.com/golang/protobuf v1.4.1 h1:ZFgWrT+bLgsYPirOnRfKLYJLvssAegOj/hgyMFdJZe0=
github.com/golang/protobuf v1.4.1/go.mod h1:U8fpvMrcmy5pZrNK1lt4xCsGvpyWQ/VVv6QDs8UjoX8=
github.com/golang/snappy v0.0.0-20180518054509-2e65f85255db/go.mod h1:/XxbfmMg8lxefKM7IXC3fBNl/7bRcc72aCRzEWrmP2Q=
github.com/golang/snappy v0.0.1 h1:Qgr9rKW7uDUkrbSmQeiDsGa8SjGyCOGtuasMWwvp2P4=
github.com/golang/snappy v0.0.1/go.mod h1:/XxbfmMg8lxefKM7IXC3fBNl/7bRcc72aCRzEWrmP2Q=
@ -396,6 +408,7 @@ github.com/klauspost/compress v1.9.2 h1:LfVyl+ZlLlLDeQ/d2AqfGIIH4qEDu0Ed2S5GyhCW
github.com/klauspost/compress v1.9.2/go.mod h1:RyIbtBH6LamlWaDj8nUwkbUhJ87Yi3uG0guNDohfE1A=
github.com/klauspost/compress v1.10.2 h1:Znfn6hXZAHaLPNnlqUYRrBSReFHYybslgv4PTiyz6P0=
github.com/klauspost/compress v1.10.2/go.mod h1:aoV0uJVorq1K+umq18yTdKaF57EivdYsUV+/s2qKfXs=
github.com/kljensen/snowball v0.6.0/go.mod h1:27N7E8fVU5H68RlUmnWwZCfxgt4POBJfENGMvNRhldw=
github.com/konsorten/go-windows-terminal-sequences v1.0.1/go.mod h1:T0+1ngSBFLxvqU3pZ+m/2kptfBszLMUkC4ZK/EgS/cQ=
github.com/kr/logfmt v0.0.0-20140226030751-b84e30acd515/go.mod h1:+0opPa2QZZtGFBFZlji/RkVcI2GknAs/DXo4wKdlNEc=
github.com/kr/pretty v0.1.0 h1:L/CwN0zerZDmRFUapSPitk6f+Q3+0za1rQkzVuMiMFI=
@ -464,6 +477,8 @@ github.com/mrjones/oauth v0.0.0-20180629183705-f4e24b6d100c h1:3wkDRdxK92dF+c1ke
github.com/mrjones/oauth v0.0.0-20180629183705-f4e24b6d100c/go.mod h1:skjdDftzkFALcuGzYSklqYd8gvat6F1gZJ4YPVbkZpM=
github.com/mschoch/smat v0.0.0-20160514031455-90eadee771ae h1:VeRdUYdCw49yizlSbMEn2SZ+gT+3IUKx8BqxyQdz+BY=
github.com/mschoch/smat v0.0.0-20160514031455-90eadee771ae/go.mod h1:qAyveg+e4CE+eKJXWVjKXM4ck2QobLqTDytGJbLLhJg=
github.com/mschoch/smat v0.2.0 h1:8imxQsjDm8yFEAVBe7azKmKSgzSkZXDuKkSq9374khM=
github.com/mschoch/smat v0.2.0/go.mod h1:kc9mz7DoBKqDyiRL7VZN8KvXQMWeTaVnttLRXOlotKw=
github.com/msteinert/pam v0.0.0-20151204160544-02ccfbfaf0cc h1:z1PgdCCmYYVL0BoJTUgmAq1p7ca8fzYIPsNyfsN3xAU=
github.com/msteinert/pam v0.0.0-20151204160544-02ccfbfaf0cc/go.mod h1:np1wUFZ6tyoke22qDJZY40URn9Ae51gX7ljIWXN5TJs=
github.com/mwitkow/go-conntrack v0.0.0-20161129095857-cc309e4a2223/go.mod h1:qRWi+5nqEBWmkhHvq77mSJWrCKwh8bxhgT7d/eI7P4U=
@ -534,6 +549,7 @@ github.com/prometheus/tsdb v0.7.1/go.mod h1:qhTCs0VvXwvX/y3TZrWD7rabWM+ijKTux40T
github.com/quasoft/websspi v1.0.0 h1:5nDgdM5xSur9s+B5w2xQ5kxf5nUGqgFgU4W0aDLZ8Mw=
github.com/quasoft/websspi v1.0.0/go.mod h1:HmVdl939dQ0WIXZhyik+ARdI03M6bQzaSEKcgpFmewk=
github.com/rcrowley/go-metrics v0.0.0-20181016184325-3113b8401b8a/go.mod h1:bCqnVzQkZxMG4s8nGwiZ5l3QUCyqpo9Y+/ZMZ9VjZe4=
github.com/rcrowley/go-metrics v0.0.0-20190826022208-cac0b30c2563/go.mod h1:bCqnVzQkZxMG4s8nGwiZ5l3QUCyqpo9Y+/ZMZ9VjZe4=
github.com/remyoudompheng/bigfft v0.0.0-20190321074620-2f0d2b0e0001 h1:YDeskXpkNDhPdWN3REluVa46HQOVuVkjkd2sWnrABNQ=
github.com/remyoudompheng/bigfft v0.0.0-20190321074620-2f0d2b0e0001/go.mod h1:qqbHyh8v60DhA7CoWK5oRCqLrMHRGoxYCSS9EjAz6Eo=
github.com/rogpeppe/fastuuid v0.0.0-20150106093220-6724a57986af/go.mod h1:XWv6SoW27p1b0cqNHllgS5HIMJraePCO15w5zCzIWYg=
@ -578,8 +594,8 @@ github.com/spf13/pflag v1.0.3/go.mod h1:DYY7MBk1bdzusC3SYhjObp+wFpr4gzcvqqNjLnIn
github.com/spf13/viper v1.3.2/go.mod h1:ZiWeW+zYFKm7srdB9IoDzzZXaJaI5eL9QjNiN/DMA2s=
github.com/spf13/viper v1.4.0 h1:yXHLWeravcrgGyFSyCgdYpXQ9dR9c/WED3pg1RhxqEU=
github.com/spf13/viper v1.4.0/go.mod h1:PTJ7Z/lr49W6bUbkmS1V3by4uWynFiR9p7+dSq/yZzE=
github.com/steveyen/gtreap v0.0.0-20150807155958-0abe01ef9be2 h1:JNEGSiWg6D3lcBCMCBqN3ELniXujt+0QNHLhNnO0w3s=
github.com/steveyen/gtreap v0.0.0-20150807155958-0abe01ef9be2/go.mod h1:mjqs7N0Q6m5HpR7QfXVBZXZWSqTjQLeTujjA/xUp2uw=
github.com/steveyen/gtreap v0.1.0 h1:CjhzTa274PyJLJuMZwIzCO1PfC00oRa8d1Kc78bFXJM=
github.com/steveyen/gtreap v0.1.0/go.mod h1:kl/5J7XbrOmlIbYIXdRHDDE5QxHqpk0cmkT7Z4dM9/Y=
github.com/stretchr/objx v0.1.0/go.mod h1:HFkY916IF+rwdDfMAkV7OtwuqBVzrE8GR6GFx+wExME=
github.com/stretchr/objx v0.1.1/go.mod h1:HFkY916IF+rwdDfMAkV7OtwuqBVzrE8GR6GFx+wExME=
github.com/stretchr/objx v0.2.0 h1:Hbg2NidpLE8veEBkEZTL3CvlkUIVzuU9jDplZO54c48=
@ -596,8 +612,8 @@ github.com/tidwall/pretty v1.0.0 h1:HsD+QiTn7sK6flMKIvNmpqz1qrpP3Ps6jOKIKMooyg4=
github.com/tidwall/pretty v1.0.0/go.mod h1:XNkn88O1ChpSDQmQeStsy+sBenx6DDtFZJxhVysOjyk=
github.com/tinylib/msgp v1.1.0 h1:9fQd+ICuRIu/ue4vxJZu6/LzxN0HwMds2nq/0cFvxHU=
github.com/tinylib/msgp v1.1.0/go.mod h1:+d+yLhGm8mzTaHzB+wgMYrodPfmZrzkirds8fDWklFE=
github.com/tinylib/msgp v1.1.1 h1:TnCZ3FIuKeaIy+F45+Cnp+caqdXGy4z74HvwXN+570Y=
github.com/tinylib/msgp v1.1.1/go.mod h1:+d+yLhGm8mzTaHzB+wgMYrodPfmZrzkirds8fDWklFE=
github.com/tinylib/msgp v1.1.2 h1:gWmO7n0Ys2RBEb7GPYB9Ujq8Mk5p2U08lRnmMcGy6BQ=
github.com/tinylib/msgp v1.1.2/go.mod h1:+d+yLhGm8mzTaHzB+wgMYrodPfmZrzkirds8fDWklFE=
github.com/tmc/grpc-websocket-proxy v0.0.0-20190109142713-0ad062ec5ee5/go.mod h1:ncp9v5uamzpCO7NfCPTXjqaC+bZgJeR0sMTm6dMHP7U=
github.com/toqueteos/trie v1.0.0 h1:8i6pXxNUXNRAqP246iibb7w/pSFquNTQ+uNfriG7vlk=
github.com/toqueteos/trie v1.0.0/go.mod h1:Ywk48QhEqhU1+DwhMkJ2x7eeGxDHiGkAdc9+0DYcbsM=
@ -637,8 +653,8 @@ github.com/yuin/goldmark-meta v0.0.0-20191126180153-f0638e958b60/go.mod h1:i9Vhc
github.com/ziutek/mymysql v1.5.4 h1:GB0qdRGsTwQSBVYuVShFBKaXSnSnYYC2d9knnE1LHFs=
github.com/ziutek/mymysql v1.5.4/go.mod h1:LMSpPZ6DbqWFxNCHW77HeMg9I646SAhApZ/wKdgO/C0=
go.etcd.io/bbolt v1.3.2/go.mod h1:IbVyRI1SCnLcuJnV2u8VeU0CEYM7e686BmAb1XKL+uU=
go.etcd.io/bbolt v1.3.3 h1:MUGmc65QhB3pIlaQ5bB4LwqSj6GIonVJXpZiaKNyaKk=
go.etcd.io/bbolt v1.3.3/go.mod h1:IbVyRI1SCnLcuJnV2u8VeU0CEYM7e686BmAb1XKL+uU=
go.etcd.io/bbolt v1.3.4 h1:hi1bXHMVrlQh6WwxAy+qZCV/SYIlqo+Ushwdpa4tAKg=
go.etcd.io/bbolt v1.3.4/go.mod h1:G5EMThwa9y8QZGBClrRx5EY+Yw9kAhnjy3bSjsnlVTQ=
go.mongodb.org/mongo-driver v1.0.3/go.mod h1:u7ryQJ+DOzQmeO7zB6MHyr8jkEQvC8vH7qLUO4lqsUM=
go.mongodb.org/mongo-driver v1.1.1 h1:Sq1fR+0c58RME5EoqKdjkiQAmPjmfHlZOoRI6fTUOcs=
go.mongodb.org/mongo-driver v1.1.1/go.mod h1:u7ryQJ+DOzQmeO7zB6MHyr8jkEQvC8vH7qLUO4lqsUM=
@ -728,6 +744,7 @@ golang.org/x/sys v0.0.0-20181107165924-66b7b1311ac8/go.mod h1:STP8DvDyc/dI5b8T5h
golang.org/x/sys v0.0.0-20181116152217-5ac8a444bdc5/go.mod h1:STP8DvDyc/dI5b8T5hshtkjS+E42TnysNCUPdjciGhY=
golang.org/x/sys v0.0.0-20181122145206-62eef0e2fa9b/go.mod h1:STP8DvDyc/dI5b8T5hshtkjS+E42TnysNCUPdjciGhY=
golang.org/x/sys v0.0.0-20181205085412-a5c9d58dba9a/go.mod h1:STP8DvDyc/dI5b8T5hshtkjS+E42TnysNCUPdjciGhY=
golang.org/x/sys v0.0.0-20181221143128-b4a75ba826a6/go.mod h1:STP8DvDyc/dI5b8T5hshtkjS+E42TnysNCUPdjciGhY=
golang.org/x/sys v0.0.0-20190215142949-d0b11bdaac8a/go.mod h1:STP8DvDyc/dI5b8T5hshtkjS+E42TnysNCUPdjciGhY=
golang.org/x/sys v0.0.0-20190221075227-b4e8571b14e0/go.mod h1:STP8DvDyc/dI5b8T5hshtkjS+E42TnysNCUPdjciGhY=
golang.org/x/sys v0.0.0-20190222072716-a9d3bda3a223/go.mod h1:STP8DvDyc/dI5b8T5hshtkjS+E42TnysNCUPdjciGhY=
@ -746,10 +763,13 @@ golang.org/x/sys v0.0.0-20190907184412-d223b2b6db03/go.mod h1:h1NjWce9XRLGQEsW7w
golang.org/x/sys v0.0.0-20191010194322-b09406accb47 h1:/XfQ9z7ib8eEJX2hdgFTZJ/ntt0swNk5oYBziWeTCvY=
golang.org/x/sys v0.0.0-20191010194322-b09406accb47/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
golang.org/x/sys v0.0.0-20191026070338-33540a1f6037/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
golang.org/x/sys v0.0.0-20200202164722-d101bd2416d5/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
golang.org/x/sys v0.0.0-20200302150141-5c8b2ff67527 h1:uYVVQ9WP/Ds2ROhcaGPeIdVq0RIXVLwsHlnvJ+cT1So=
golang.org/x/sys v0.0.0-20200302150141-5c8b2ff67527/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
golang.org/x/sys v0.0.0-20200323222414-85ca7c5b95cd h1:xhmwyvizuTgC2qz7ZlMluP20uW+C3Rm0FD/WLDX8884=
golang.org/x/sys v0.0.0-20200323222414-85ca7c5b95cd/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
golang.org/x/sys v0.0.0-20200509044756-6aff5f38e54f h1:mOhmO9WsBaJCNmaZHPtHs9wOcdqdKCjF6OPJlmDM3KI=
golang.org/x/sys v0.0.0-20200509044756-6aff5f38e54f/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
golang.org/x/text v0.3.0/go.mod h1:NqM8EUOU14njkJ3fqMW+pc6Ldnwhi/IjpwHt7yyuwOQ=
golang.org/x/text v0.3.1-0.20180807135948-17ff2d5776d2/go.mod h1:NqM8EUOU14njkJ3fqMW+pc6Ldnwhi/IjpwHt7yyuwOQ=
golang.org/x/text v0.3.2 h1:tW2bmiBqwgJj/UpqtC8EpXEZVYOwU0yG4iWbprSVAcs=
@ -822,6 +842,8 @@ google.golang.org/protobuf v0.0.0-20200228230310-ab0ca4ff8a60/go.mod h1:cfTl7dwQ
google.golang.org/protobuf v1.20.1-0.20200309200217-e05f789c0967/go.mod h1:A+miEFZTKqfCUM6K7xSMQL9OKL/b6hQv+e19PK+JZNE=
google.golang.org/protobuf v1.21.0 h1:qdOKuR/EIArgaWNjetjgTzgVTAZ+S/WXVrq9HW9zimw=
google.golang.org/protobuf v1.21.0/go.mod h1:47Nbq4nVaFHyn7ilMalzfO3qCViNmqZ2kzikPIcrTAo=
google.golang.org/protobuf v1.22.0 h1:cJv5/xdbk1NnMPR1VP9+HU6gupuG9MLBoH1r6RHZ2MY=
google.golang.org/protobuf v1.22.0/go.mod h1:EGpADcykh3NcUnDUJcl1+ZksZNG86OlYog2l/sGQquU=
gopkg.in/alecthomas/kingpin.v2 v2.2.6/go.mod h1:FMv+mEhP44yOT+4EoQTLFTRgOQ1FBLkstjWtayDeSgw=
gopkg.in/alexcesaro/quotedprintable.v3 v3.0.0-20150716171945-2caba252f4dc h1:2gGKlE2+asNV9m7xrywl36YYNnBG5ZQ0r/BOOxqPpmk=
gopkg.in/alexcesaro/quotedprintable.v3 v3.0.0-20150716171945-2caba252f4dc/go.mod h1:m7x9LTH6d71AHyAX77c9yqWCCa3UKHcVEj9y7hAtKDk=

@ -14,6 +14,7 @@ go:
- "1.10.x"
- "1.11.x"
- "1.12.x"
- "1.13.x"
- tip
# whitelist

@ -29,11 +29,17 @@ Roaring bitmaps are found to work well in many important applications:
The ``roaring`` Go library is used by
* [Cloud Torrent](https://github.com/jpillora/cloud-torrent): a self-hosted remote torrent client
* [runv](https://github.com/hyperhq/runv): an Hypervisor-based runtime for the Open Containers Initiative
* [Cloud Torrent](https://github.com/jpillora/cloud-torrent)
* [runv](https://github.com/hyperhq/runv)
* [InfluxDB](https://www.influxdata.com)
* [Pilosa](https://www.pilosa.com/)
* [Bleve](http://www.blevesearch.com)
* [lindb](https://github.com/lindb/lindb)
* [Elasticell](https://github.com/deepfabric/elasticell)
* [SourceGraph](https://github.com/sourcegraph/sourcegraph)
* [M3](https://github.com/m3db/m3)
* [trident](https://github.com/NetApp/trident)
This library is used in production in several systems, it is part of the [Awesome Go collection](https://awesome-go.com).

@ -24,6 +24,18 @@ func (ac *arrayContainer) fillLeastSignificant16bits(x []uint32, i int, mask uin
}
}
func (ac *arrayContainer) iterate(cb func(x uint16) bool) bool {
iterator := shortIterator{ac.content, 0}
for iterator.hasNext() {
if !cb(iterator.next()) {
return false
}
}
return true
}
func (ac *arrayContainer) getShortIterator() shortPeekable {
return &shortIterator{ac.content, 0}
}

@ -96,6 +96,18 @@ func (bc *bitmapContainer) maximum() uint16 {
return uint16(0)
}
func (bc *bitmapContainer) iterate(cb func(x uint16) bool) bool {
iterator := bitmapContainerShortIterator{bc, bc.NextSetBit(0)}
for iterator.hasNext() {
if !cb(iterator.next()) {
return false
}
}
return true
}
type bitmapContainerShortIterator struct {
ptr *bitmapContainer
i int

@ -416,6 +416,38 @@ func (rb *Bitmap) String() string {
return buffer.String()
}
// Iterate iterates over the bitmap, calling the given callback with each value in the bitmap. If the callback returns
// false, the iteration is halted.
// The iteration results are undefined if the bitmap is modified (e.g., with Add or Remove).
// There is no guarantee as to what order the values will be iterated
func (rb *Bitmap) Iterate(cb func(x uint32) bool) {
for i := 0; i < rb.highlowcontainer.size(); i++ {
hs := uint32(rb.highlowcontainer.getKeyAtIndex(i)) << 16
c := rb.highlowcontainer.getContainerAtIndex(i)
var shouldContinue bool
// This is hacky but it avoids allocations from invoking an interface method with a closure
switch t := c.(type) {
case *arrayContainer:
shouldContinue = t.iterate(func(x uint16) bool {
return cb(uint32(x) | hs)
})
case *runContainer16:
shouldContinue = t.iterate(func(x uint16) bool {
return cb(uint32(x) | hs)
})
case *bitmapContainer:
shouldContinue = t.iterate(func(x uint16) bool {
return cb(uint32(x) | hs)
})
}
if !shouldContinue {
break
}
}
}
// Iterator creates a new IntPeekable to iterate over the integers contained in the bitmap, in sorted order;
// the iterator becomes invalid if the bitmap is modified (e.g., with Add or Remove).
func (rb *Bitmap) Iterator() IntPeekable {
@ -475,41 +507,72 @@ func (rb *Bitmap) Equals(o interface{}) bool {
// AddOffset adds the value 'offset' to each and every value in a bitmap, generating a new bitmap in the process
func AddOffset(x *Bitmap, offset uint32) (answer *Bitmap) {
containerOffset := highbits(offset)
inOffset := lowbits(offset)
return AddOffset64(x, int64(offset))
}
// AddOffset64 adds the value 'offset' to each and every value in a bitmap, generating a new bitmap in the process
// If offset + element is outside of the range [0,2^32), that the element will be dropped
func AddOffset64(x *Bitmap, offset int64) (answer *Bitmap) {
// we need "offset" to be a long because we want to support values
// between -0xFFFFFFFF up to +-0xFFFFFFFF
var containerOffset64 int64
if offset < 0 {
containerOffset64 = (offset - (1 << 16) + 1) / (1 << 16)
} else {
containerOffset64 = offset >> 16
}
if containerOffset64 >= (1<<16) || containerOffset64 <= -(1<<16) {
return New()
}
containerOffset := int32(containerOffset64)
inOffset := (uint16)(offset - containerOffset64*(1<<16))
if inOffset == 0 {
answer = x.Clone()
for pos := 0; pos < answer.highlowcontainer.size(); pos++ {
key := answer.highlowcontainer.getKeyAtIndex(pos)
key := int32(answer.highlowcontainer.getKeyAtIndex(pos))
key += containerOffset
answer.highlowcontainer.keys[pos] = key
if key >= 0 && key <= MaxUint16 {
answer.highlowcontainer.keys[pos] = uint16(key)
}
}
} else {
answer = New()
for pos := 0; pos < x.highlowcontainer.size(); pos++ {
key := x.highlowcontainer.getKeyAtIndex(pos)
key := int32(x.highlowcontainer.getKeyAtIndex(pos))
key += containerOffset
c := x.highlowcontainer.getContainerAtIndex(pos)
offsetted := c.addOffset(inOffset)
if offsetted[0].getCardinality() > 0 {
if offsetted[0].getCardinality() > 0 && (key >= 0 && key <= MaxUint16) {
curSize := answer.highlowcontainer.size()
lastkey := uint16(0)
lastkey := int32(0)
if curSize > 0 {
lastkey = answer.highlowcontainer.getKeyAtIndex(curSize - 1)
lastkey = int32(answer.highlowcontainer.getKeyAtIndex(curSize - 1))
}
if curSize > 0 && lastkey == key {
prev := answer.highlowcontainer.getContainerAtIndex(curSize - 1)
orrseult := prev.ior(offsetted[0])
answer.highlowcontainer.setContainerAtIndex(curSize-1, orrseult)
} else {
answer.highlowcontainer.appendContainer(key, offsetted[0], false)
answer.highlowcontainer.appendContainer(uint16(key), offsetted[0], false)
}
}
if offsetted[1].getCardinality() > 0 {
answer.highlowcontainer.appendContainer(key+1, offsetted[1], false)
if offsetted[1].getCardinality() > 0 && ((key+1) >= 0 && (key+1) <= MaxUint16) {
answer.highlowcontainer.appendContainer(uint16(key+1), offsetted[1], false)
}
}
}
return answer
}

@ -4,9 +4,10 @@ import (
"bytes"
"encoding/binary"
"fmt"
"io"
snappy "github.com/glycerine/go-unsnap-stream"
"github.com/tinylib/msgp/msgp"
"io"
)
//go:generate msgp -unexported
@ -38,6 +39,7 @@ type container interface {
inot(firstOfRange, endx int) container // i stands for inplace, range is [firstOfRange,endx)
xor(r container) container
getShortIterator() shortPeekable
iterate(cb func(x uint16) bool) bool
getReverseIterator() shortIterable
getManyIterator() manyIterable
contains(i uint16) bool
@ -488,20 +490,15 @@ func (ra *roaringArray) writeTo(w io.Writer) (n int64, err error) {
nw += 2
binary.LittleEndian.PutUint16(buf[2:], uint16(len(ra.keys)-1))
nw += 2
// compute isRun bitmap
var ir []byte
isRun := newBitmapContainer()
// compute isRun bitmap without temporary allocation
var runbitmapslice = buf[nw:nw+isRunSizeInBytes]
for i, c := range ra.containers {
switch c.(type) {
case *runContainer16:
isRun.iadd(uint16(i))
runbitmapslice[i / 8] |= 1<<(uint(i)%8)
}
}
// convert to little endian
ir = isRun.asLittleEndianByteSlice()[:isRunSizeInBytes]
nw += copy(buf[nw:], ir)
nw += isRunSizeInBytes
} else {
binary.LittleEndian.PutUint32(buf[0:], uint32(serialCookieNoRunContainer))
nw += 4

@ -1162,6 +1162,18 @@ func (rc *runContainer16) newRunIterator16() *runIterator16 {
return &runIterator16{rc: rc, curIndex: 0, curPosInIndex: 0}
}
func (rc *runContainer16) iterate(cb func(x uint16) bool) bool {
iterator := runIterator16{rc, 0, 0}
for iterator.hasNext() {
if !cb(iterator.next()) {
return false
}
}
return true
}
// hasNext returns false if calling next will panic. It
// returns true when there is at least one more value
// available in the iteration sequence.

@ -14,6 +14,7 @@ func equal(a, b []uint16) bool {
func difference(set1 []uint16, set2 []uint16, buffer []uint16) int {
if 0 == len(set2) {
buffer = buffer[:len(set1)]
for k := 0; k < len(set1); k++ {
buffer[k] = set1[k]
}

@ -112,7 +112,7 @@ func highbits(x uint32) uint16 {
return uint16(x >> 16)
}
func lowbits(x uint32) uint16 {
return uint16(x & 0xFFFF)
return uint16(x & maxLowBit)
}
const maxLowBit = 0xFFFF

@ -3,9 +3,9 @@ sudo: false
language: go
go:
- "1.10.x"
- "1.11.x"
- "1.12.x"
- "1.13.x"
- "1.14.x"
script:
- go get golang.org/x/tools/cmd/cover
@ -16,11 +16,7 @@ script:
- go test -race -v $(go list ./... | grep -v vendor/)
- go vet $(go list ./... | grep -v vendor/)
- go test ./test -v -indexType scorch
- if [[ ${TRAVIS_GO_VERSION} =~ ^1\.10 ]]; then
echo "errcheck skipped for go version" $TRAVIS_GO_VERSION;
else
errcheck -ignorepkg fmt $(go list ./... | grep -v vendor/);
fi
- errcheck -ignorepkg fmt $(go list ./... | grep -v vendor/);
- docs/project-code-coverage.sh
- docs/build_children.sh

@ -0,0 +1,49 @@
// Copyright (c) 2020 Couchbase, Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
package en
import (
"github.com/blevesearch/bleve/analysis"
"github.com/blevesearch/bleve/registry"
"github.com/blevesearch/snowballstem"
"github.com/blevesearch/snowballstem/english"
)
const SnowballStemmerName = "stemmer_en_snowball"
type EnglishStemmerFilter struct {
}
func NewEnglishStemmerFilter() *EnglishStemmerFilter {
return &EnglishStemmerFilter{}
}
func (s *EnglishStemmerFilter) Filter(input analysis.TokenStream) analysis.TokenStream {
for _, token := range input {
env := snowballstem.NewEnv(string(token.Term))
english.Stem(env)
token.Term = []byte(env.Current())
}
return input
}
func EnglishStemmerFilterConstructor(config map[string]interface{}, cache *registry.Cache) (analysis.TokenFilter, error) {
return NewEnglishStemmerFilter(), nil
}
func init() {
registry.RegisterTokenFilter(SnowballStemmerName, EnglishStemmerFilterConstructor)
}

@ -33,14 +33,16 @@ var minLonRad = minLon * degreesToRadian
var minLatRad = minLat * degreesToRadian
var maxLonRad = maxLon * degreesToRadian
var maxLatRad = maxLat * degreesToRadian
var geoTolerance = 1E-6
var geoTolerance = 1e-6
var lonScale = float64((uint64(0x1)<<GeoBits)-1) / 360.0
var latScale = float64((uint64(0x1)<<GeoBits)-1) / 180.0
var geoHashMaxLength = 12
// Point represents a geo point.
type Point struct {
Lon float64
Lat float64
Lon float64 `json:"lon"`
Lat float64 `json:"lat"`
}
// MortonHash computes the morton hash value for the provided geo point

@ -85,11 +85,13 @@ func ExtractGeoPoint(thing interface{}) (lon, lat float64, success bool) {
}
} else {
// geohash
if len(geoStr) <= geoHashMaxLength {
lat, lon = DecodeGeoHash(geoStr)
foundLat = true
foundLon = true
}
}
}
// is it a map
if l, ok := thing.(map[string]interface{}); ok {

@ -0,0 +1,25 @@
module github.com/blevesearch/bleve
go 1.13
require (
github.com/RoaringBitmap/roaring v0.4.21
github.com/blevesearch/blevex v0.0.0-20190916190636-152f0fe5c040
github.com/blevesearch/go-porterstemmer v1.0.3
github.com/blevesearch/segment v0.9.0
github.com/blevesearch/snowballstem v0.9.0
github.com/blevesearch/zap/v11 v11.0.7
github.com/blevesearch/zap/v12 v12.0.7
github.com/couchbase/ghistogram v0.1.0 // indirect
github.com/couchbase/moss v0.1.0
github.com/couchbase/vellum v1.0.1
github.com/golang/protobuf v1.3.2
github.com/kljensen/snowball v0.6.0
github.com/rcrowley/go-metrics v0.0.0-20190826022208-cac0b30c2563
github.com/spf13/cobra v0.0.5
github.com/steveyen/gtreap v0.1.0
github.com/syndtr/goleveldb v1.0.0
github.com/willf/bitset v1.1.10
go.etcd.io/bbolt v1.3.4
golang.org/x/text v0.3.0
)

@ -21,7 +21,6 @@ import (
"github.com/RoaringBitmap/roaring"
"github.com/blevesearch/bleve/index"
"github.com/blevesearch/bleve/index/scorch/segment"
"github.com/blevesearch/bleve/index/scorch/segment/zap"
)
type segmentIntroduction struct {
@ -77,11 +76,6 @@ OUTER:
case persist := <-s.persists:
s.introducePersist(persist)
case revertTo := <-s.revertToSnapshots:
err := s.revertToSnapshot(revertTo)
if err != nil {
continue OUTER
}
}
var epochCurr uint64
@ -312,6 +306,8 @@ func (s *Scorch) introducePersist(persist *persistIntroduction) {
close(persist.applied)
}
// The introducer should definitely handle the segmentMerge.notify
// channel before exiting the introduceMerge.
func (s *Scorch) introduceMerge(nextMerge *segmentMerge) {
atomic.AddUint64(&s.stats.TotIntroduceMergeBeg, 1)
defer atomic.AddUint64(&s.stats.TotIntroduceMergeEnd, 1)
@ -409,11 +405,11 @@ func (s *Scorch) introduceMerge(nextMerge *segmentMerge) {
atomic.AddUint64(&s.stats.TotIntroducedSegmentsMerge, 1)
switch nextMerge.new.(type) {
case *zap.SegmentBase:
case segment.PersistedSegment:
fileSegments++
default:
docsToPersistCount += nextMerge.new.Count() - newSegmentDeleted.GetCardinality()
memSegments++
case *zap.Segment:
fileSegments++
}
}
@ -443,86 +439,11 @@ func (s *Scorch) introduceMerge(nextMerge *segmentMerge) {
close(nextMerge.notify)
}
func (s *Scorch) revertToSnapshot(revertTo *snapshotReversion) error {
atomic.AddUint64(&s.stats.TotIntroduceRevertBeg, 1)
defer atomic.AddUint64(&s.stats.TotIntroduceRevertEnd, 1)
if revertTo.snapshot == nil {
err := fmt.Errorf("Cannot revert to a nil snapshot")
revertTo.applied <- err
return err
}
// acquire lock
s.rootLock.Lock()
// prepare a new index snapshot, based on next snapshot
newSnapshot := &IndexSnapshot{
parent: s,
segment: make([]*SegmentSnapshot, len(revertTo.snapshot.segment)),
offsets: revertTo.snapshot.offsets,
internal: revertTo.snapshot.internal,
epoch: s.nextSnapshotEpoch,
refs: 1,
creator: "revertToSnapshot",
}
s.nextSnapshotEpoch++
var docsToPersistCount, memSegments, fileSegments uint64
// iterate through segments
for i, segmentSnapshot := range revertTo.snapshot.segment {
newSnapshot.segment[i] = &SegmentSnapshot{
id: segmentSnapshot.id,
segment: segmentSnapshot.segment,
deleted: segmentSnapshot.deleted,
cachedDocs: segmentSnapshot.cachedDocs,
creator: segmentSnapshot.creator,
}
newSnapshot.segment[i].segment.AddRef()
// remove segment from ineligibleForRemoval map
filename := zapFileName(segmentSnapshot.id)
delete(s.ineligibleForRemoval, filename)
if isMemorySegment(segmentSnapshot) {
docsToPersistCount += segmentSnapshot.Count()
memSegments++
} else {
fileSegments++
}
}
atomic.StoreUint64(&s.stats.TotItemsToPersist, docsToPersistCount)
atomic.StoreUint64(&s.stats.TotMemorySegmentsAtRoot, memSegments)
atomic.StoreUint64(&s.stats.TotFileSegmentsAtRoot, fileSegments)
if revertTo.persisted != nil {
s.rootPersisted = append(s.rootPersisted, revertTo.persisted)
}
newSnapshot.updateSize()
// swap in new snapshot
rootPrev := s.root
s.root = newSnapshot
atomic.StoreUint64(&s.stats.CurRootEpoch, s.root.epoch)
// release lock
s.rootLock.Unlock()
if rootPrev != nil {
_ = rootPrev.DecRef()
}
close(revertTo.applied)
return nil
}
func isMemorySegment(s *SegmentSnapshot) bool {
switch s.segment.(type) {
case *zap.SegmentBase:
return true
default:
case segment.PersistedSegment:
return false
default:
return true
}
}

@ -25,7 +25,6 @@ import (
"github.com/RoaringBitmap/roaring"
"github.com/blevesearch/bleve/index/scorch/mergeplan"
"github.com/blevesearch/bleve/index/scorch/segment"
"github.com/blevesearch/bleve/index/scorch/segment/zap"
)
func (s *Scorch) mergerLoop() {
@ -131,18 +130,18 @@ func (s *Scorch) parseMergePlannerOptions() (*mergeplan.MergePlanOptions,
func (s *Scorch) planMergeAtSnapshot(ourSnapshot *IndexSnapshot,
options *mergeplan.MergePlanOptions) error {
// build list of zap segments in this snapshot
var onlyZapSnapshots []mergeplan.Segment
// build list of persisted segments in this snapshot
var onlyPersistedSnapshots []mergeplan.Segment
for _, segmentSnapshot := range ourSnapshot.segment {
if _, ok := segmentSnapshot.segment.(*zap.Segment); ok {
onlyZapSnapshots = append(onlyZapSnapshots, segmentSnapshot)
if _, ok := segmentSnapshot.segment.(segment.PersistedSegment); ok {
onlyPersistedSnapshots = append(onlyPersistedSnapshots, segmentSnapshot)
}
}
atomic.AddUint64(&s.stats.TotFileMergePlan, 1)
// give this list to the planner
resultMergePlan, err := mergeplan.Plan(onlyZapSnapshots, options)
resultMergePlan, err := mergeplan.Plan(onlyPersistedSnapshots, options)
if err != nil {
atomic.AddUint64(&s.stats.TotFileMergePlanErr, 1)
return fmt.Errorf("merge planning err: %v", err)
@ -157,8 +156,8 @@ func (s *Scorch) planMergeAtSnapshot(ourSnapshot *IndexSnapshot,
atomic.AddUint64(&s.stats.TotFileMergePlanTasks, uint64(len(resultMergePlan.Tasks)))
// process tasks in serial for now
var notifications []chan *IndexSnapshot
var filenames []string
for _, task := range resultMergePlan.Tasks {
if len(task.Segments) == 0 {
atomic.AddUint64(&s.stats.TotFileMergePlanTasksSegmentsEmpty, 1)
@ -169,24 +168,24 @@ func (s *Scorch) planMergeAtSnapshot(ourSnapshot *IndexSnapshot,
oldMap := make(map[uint64]*SegmentSnapshot)
newSegmentID := atomic.AddUint64(&s.nextSegmentID, 1)
segmentsToMerge := make([]*zap.Segment, 0, len(task.Segments))
segmentsToMerge := make([]segment.Segment, 0, len(task.Segments))
docsToDrop := make([]*roaring.Bitmap, 0, len(task.Segments))
for _, planSegment := range task.Segments {
if segSnapshot, ok := planSegment.(*SegmentSnapshot); ok {
oldMap[segSnapshot.id] = segSnapshot
if zapSeg, ok := segSnapshot.segment.(*zap.Segment); ok {
if persistedSeg, ok := segSnapshot.segment.(segment.PersistedSegment); ok {
if segSnapshot.LiveSize() == 0 {
atomic.AddUint64(&s.stats.TotFileMergeSegmentsEmpty, 1)
oldMap[segSnapshot.id] = nil
} else {
segmentsToMerge = append(segmentsToMerge, zapSeg)
segmentsToMerge = append(segmentsToMerge, segSnapshot.segment)
docsToDrop = append(docsToDrop, segSnapshot.deleted)
}
// track the files getting merged for unsetting the
// removal ineligibility. This helps to unflip files
// even with fast merger, slow persister work flows.
path := zapSeg.Path()
path := persistedSeg.Path()
filenames = append(filenames,
strings.TrimPrefix(path, s.path+string(os.PathSeparator)))
}
@ -203,8 +202,8 @@ func (s *Scorch) planMergeAtSnapshot(ourSnapshot *IndexSnapshot,
fileMergeZapStartTime := time.Now()
atomic.AddUint64(&s.stats.TotFileMergeZapBeg, 1)
newDocNums, _, err := zap.Merge(segmentsToMerge, docsToDrop, path,
DefaultChunkFactor, s.closeCh, s)
newDocNums, _, err := s.segPlugin.Merge(segmentsToMerge, docsToDrop, path,
s.closeCh, s)
atomic.AddUint64(&s.stats.TotFileMergeZapEnd, 1)
fileMergeZapTime := uint64(time.Since(fileMergeZapStartTime))
@ -222,17 +221,12 @@ func (s *Scorch) planMergeAtSnapshot(ourSnapshot *IndexSnapshot,
return fmt.Errorf("merging failed: %v", err)
}
seg, err = zap.Open(path)
seg, err = s.segPlugin.Open(path)
if err != nil {
s.unmarkIneligibleForRemoval(filename)
atomic.AddUint64(&s.stats.TotFileMergePlanTasksErr, 1)
return err
}
err = zap.ValidateMerge(segmentsToMerge, nil, docsToDrop, seg.(*zap.Segment))
if err != nil {
s.unmarkIneligibleForRemoval(filename)
return fmt.Errorf("merge validation failed: %v", err)
}
oldNewDocNums = make(map[uint64][]uint64)
for i, segNewDocNums := range newDocNums {
oldNewDocNums[task.Segments[i].Id()] = segNewDocNums
@ -246,9 +240,8 @@ func (s *Scorch) planMergeAtSnapshot(ourSnapshot *IndexSnapshot,
old: oldMap,
oldNewDocNums: oldNewDocNums,
new: seg,
notify: make(chan *IndexSnapshot, 1),
notify: make(chan *IndexSnapshot),
}
notifications = append(notifications, sm.notify)
// give it to the introducer
select {
@ -259,20 +252,21 @@ func (s *Scorch) planMergeAtSnapshot(ourSnapshot *IndexSnapshot,
atomic.AddUint64(&s.stats.TotFileMergeIntroductions, 1)
}
atomic.AddUint64(&s.stats.TotFileMergePlanTasksDone, 1)
introStartTime := time.Now()
// it is safe to blockingly wait for the merge introduction
// here as the introducer is bound to handle the notify channel.
newSnapshot := <-sm.notify
introTime := uint64(time.Since(introStartTime))
atomic.AddUint64(&s.stats.TotFileMergeZapIntroductionTime, introTime)
if atomic.LoadUint64(&s.stats.MaxFileMergeZapIntroductionTime) < introTime {
atomic.StoreUint64(&s.stats.MaxFileMergeZapIntroductionTime, introTime)
}
for _, notification := range notifications {
select {
case <-s.closeCh:
atomic.AddUint64(&s.stats.TotFileMergeIntroductionsSkipped, 1)
return segment.ErrClosed
case newSnapshot := <-notification:
atomic.AddUint64(&s.stats.TotFileMergeIntroductionsDone, 1)
if newSnapshot != nil {
_ = newSnapshot.DecRef()
}
}
atomic.AddUint64(&s.stats.TotFileMergePlanTasksDone, 1)
}
// once all the newly merged segment introductions are done,
@ -297,8 +291,8 @@ type segmentMerge struct {
// persisted segment, and synchronously introduce that new segment
// into the root
func (s *Scorch) mergeSegmentBases(snapshot *IndexSnapshot,
sbs []*zap.SegmentBase, sbsDrops []*roaring.Bitmap, sbsIndexes []int,
chunkFactor uint32) (*IndexSnapshot, uint64, error) {
sbs []segment.Segment, sbsDrops []*roaring.Bitmap,
sbsIndexes []int) (*IndexSnapshot, uint64, error) {
atomic.AddUint64(&s.stats.TotMemMergeBeg, 1)
memMergeZapStartTime := time.Now()
@ -310,7 +304,7 @@ func (s *Scorch) mergeSegmentBases(snapshot *IndexSnapshot,
path := s.path + string(os.PathSeparator) + filename
newDocNums, _, err :=
zap.MergeSegmentBases(sbs, sbsDrops, path, chunkFactor, s.closeCh, s)
s.segPlugin.Merge(sbs, sbsDrops, path, s.closeCh, s)
atomic.AddUint64(&s.stats.TotMemMergeZapEnd, 1)
@ -325,15 +319,11 @@ func (s *Scorch) mergeSegmentBases(snapshot *IndexSnapshot,
return nil, 0, err
}
seg, err := zap.Open(path)
seg, err := s.segPlugin.Open(path)
if err != nil {
atomic.AddUint64(&s.stats.TotMemMergeErr, 1)
return nil, 0, err
}
err = zap.ValidateMerge(nil, sbs, sbsDrops, seg.(*zap.Segment))
if err != nil {
return nil, 0, fmt.Errorf("in-memory merge validation failed: %v", err)
}
// update persisted stats
atomic.AddUint64(&s.stats.TotPersistedItems, seg.Count())
@ -344,7 +334,7 @@ func (s *Scorch) mergeSegmentBases(snapshot *IndexSnapshot,
old: make(map[uint64]*SegmentSnapshot),
oldNewDocNums: make(map[uint64][]uint64),
new: seg,
notify: make(chan *IndexSnapshot, 1),
notify: make(chan *IndexSnapshot),
}
for i, idx := range sbsIndexes {
@ -360,14 +350,13 @@ func (s *Scorch) mergeSegmentBases(snapshot *IndexSnapshot,
case s.merges <- sm:
}
select { // wait for introduction to complete
case <-s.closeCh:
return nil, 0, segment.ErrClosed
case newSnapshot := <-sm.notify:
// blockingly wait for the introduction to complete
newSnapshot := <-sm.notify
if newSnapshot != nil {
atomic.AddUint64(&s.stats.TotMemMergeSegments, uint64(len(sbs)))
atomic.AddUint64(&s.stats.TotMemMergeDone, 1)
return newSnapshot, newSegmentID, nil
}
return newSnapshot, newSegmentID, nil
}
func (s *Scorch) ReportBytesWritten(bytesWritten uint64) {

@ -18,10 +18,8 @@ import (
"fmt"
"github.com/RoaringBitmap/roaring"
"github.com/blevesearch/bleve/index"
"github.com/blevesearch/bleve/index/scorch/segment"
"github.com/blevesearch/bleve/index/scorch/segment/zap"
)
var OptimizeConjunction = true
@ -81,25 +79,25 @@ func (o *OptimizeTFRConjunction) Finish() (index.Optimized, error) {
}
for i := range o.snapshot.segment {
itr0, ok := o.tfrs[0].iterators[i].(*zap.PostingsIterator)
if !ok || itr0.ActualBM == nil {
itr0, ok := o.tfrs[0].iterators[i].(segment.OptimizablePostingsIterator)
if !ok || itr0.ActualBitmap() == nil {
continue
}
itr1, ok := o.tfrs[1].iterators[i].(*zap.PostingsIterator)
if !ok || itr1.ActualBM == nil {
itr1, ok := o.tfrs[1].iterators[i].(segment.OptimizablePostingsIterator)
if !ok || itr1.ActualBitmap() == nil {
continue
}
bm := roaring.And(itr0.ActualBM, itr1.ActualBM)
bm := roaring.And(itr0.ActualBitmap(), itr1.ActualBitmap())
for _, tfr := range o.tfrs[2:] {
itr, ok := tfr.iterators[i].(*zap.PostingsIterator)
if !ok || itr.ActualBM == nil {
itr, ok := tfr.iterators[i].(segment.OptimizablePostingsIterator)
if !ok || itr.ActualBitmap() == nil {
continue
}
bm.And(itr.ActualBM)
bm.And(itr.ActualBitmap())
}
// in this conjunction optimization, the postings iterators
@ -107,10 +105,9 @@ func (o *OptimizeTFRConjunction) Finish() (index.Optimized, error) {
// regular conjunction searcher machinery will still be used,
// but the underlying bitmap will be smaller.
for _, tfr := range o.tfrs {
itr, ok := tfr.iterators[i].(*zap.PostingsIterator)
if ok && itr.ActualBM != nil {
itr.ActualBM = bm
itr.Actual = bm.Iterator()
itr, ok := tfr.iterators[i].(segment.OptimizablePostingsIterator)
if ok && itr.ActualBitmap() != nil {
itr.ReplaceActual(bm)
}
}
}
@ -191,9 +188,9 @@ OUTER:
continue OUTER
}
itr, ok := tfr.iterators[i].(*zap.PostingsIterator)
itr, ok := tfr.iterators[i].(segment.OptimizablePostingsIterator)
if !ok {
// We optimize zap postings iterators only.
// We only optimize postings iterators that support this operation.
return nil, nil
}
@ -201,12 +198,6 @@ OUTER:
// can perform several optimizations up-front here.
docNum1Hit, ok := itr.DocNum1Hit()
if ok {
if docNum1Hit == zap.DocNum1HitFinished {
// An empty docNum here means the entire AND is empty.
oTFR.iterators[i] = segment.AnEmptyPostingsIterator
continue OUTER
}
if docNum1HitLastOk && docNum1HitLast != docNum1Hit {
// The docNum1Hit doesn't match the previous
// docNum1HitLast, so the entire AND is empty.
@ -220,14 +211,14 @@ OUTER:
continue
}
if itr.ActualBM == nil {
if itr.ActualBitmap() == nil {
// An empty actual bitmap means the entire AND is empty.
oTFR.iterators[i] = segment.AnEmptyPostingsIterator
continue OUTER
}
// Collect the actual bitmap for more processing later.
actualBMs = append(actualBMs, itr.ActualBM)
actualBMs = append(actualBMs, itr.ActualBitmap())
}
if docNum1HitLastOk {
@ -245,11 +236,7 @@ OUTER:
// The actual bitmaps and docNum1Hits all contain or have
// the same 1-hit docNum, so that's our AND'ed result.
oTFR.iterators[i], err = zap.PostingsIteratorFrom1Hit(
docNum1HitLast, zap.NormBits1Hit, false, false)
if err != nil {
return nil, nil
}
oTFR.iterators[i] = segment.NewUnadornedPostingsIteratorFrom1Hit(docNum1HitLast)
continue OUTER
}
@ -263,11 +250,7 @@ OUTER:
if len(actualBMs) == 1 {
// If we've only 1 actual bitmap, then that's our result.
oTFR.iterators[i], err = zap.PostingsIteratorFromBitmap(
actualBMs[0], false, false)
if err != nil {
return nil, nil
}
oTFR.iterators[i] = segment.NewUnadornedPostingsIteratorFromBitmap(actualBMs[0])
continue OUTER
}
@ -279,11 +262,7 @@ OUTER:
bm.And(actualBM)
}
oTFR.iterators[i], err = zap.PostingsIteratorFromBitmap(
bm, false, false)
if err != nil {
return nil, nil
}
oTFR.iterators[i] = segment.NewUnadornedPostingsIteratorFromBitmap(bm)
}
return oTFR, nil
@ -337,13 +316,13 @@ func (o *OptimizeTFRDisjunctionUnadorned) Finish() (rv index.Optimized, err erro
var cMax uint64
for _, tfr := range o.tfrs {
itr, ok := tfr.iterators[i].(*zap.PostingsIterator)
itr, ok := tfr.iterators[i].(segment.OptimizablePostingsIterator)
if !ok {
return nil, nil
}
if itr.ActualBM != nil {
c := itr.ActualBM.GetCardinality()
if itr.ActualBitmap() != nil {
c := itr.ActualBitmap().GetCardinality()
if cMax < c {
cMax = c
}
@ -379,7 +358,7 @@ func (o *OptimizeTFRDisjunctionUnadorned) Finish() (rv index.Optimized, err erro
actualBMs = actualBMs[:0]
for _, tfr := range o.tfrs {
itr, ok := tfr.iterators[i].(*zap.PostingsIterator)
itr, ok := tfr.iterators[i].(segment.OptimizablePostingsIterator)
if !ok {
return nil, nil
}
@ -390,8 +369,8 @@ func (o *OptimizeTFRDisjunctionUnadorned) Finish() (rv index.Optimized, err erro
continue
}
if itr.ActualBM != nil {
actualBMs = append(actualBMs, itr.ActualBM)
if itr.ActualBitmap() != nil {
actualBMs = append(actualBMs, itr.ActualBitmap())
}
}
@ -410,10 +389,7 @@ func (o *OptimizeTFRDisjunctionUnadorned) Finish() (rv index.Optimized, err erro
bm.AddMany(docNums)
oTFR.iterators[i], err = zap.PostingsIteratorFromBitmap(bm, false, false)
if err != nil {
return nil, nil
}
oTFR.iterators[i] = segment.NewUnadornedPostingsIteratorFromBitmap(bm)
}
return oTFR, nil

@ -32,12 +32,9 @@ import (
"github.com/RoaringBitmap/roaring"
"github.com/blevesearch/bleve/index"
"github.com/blevesearch/bleve/index/scorch/segment"
"github.com/blevesearch/bleve/index/scorch/segment/zap"
bolt "github.com/etcd-io/bbolt"
bolt "go.etcd.io/bbolt"
)
var DefaultChunkFactor uint32 = 1024
// DefaultPersisterNapTimeMSec is kept to zero as this helps in direct
// persistence of segments with the default safe batch option.
// If the default safe batch option results in high number of
@ -253,7 +250,7 @@ func (s *Scorch) pausePersisterForMergerCatchUp(lastPersistedEpoch uint64,
persistWatchers = notifyMergeWatchers(lastPersistedEpoch, persistWatchers)
// Check the merger lag by counting the segment files on disk,
numFilesOnDisk, _ := s.diskFileStats()
numFilesOnDisk, _, _ := s.diskFileStats(nil)
// On finding fewer files on disk, persister takes a short pause
// for sufficient in-memory segments to pile up for the next
@ -280,7 +277,7 @@ func (s *Scorch) pausePersisterForMergerCatchUp(lastPersistedEpoch uint64,
// 2. The merger could be lagging behind on merging the disk files.
if numFilesOnDisk > uint64(po.PersisterNapUnderNumFiles) {
s.removeOldData()
numFilesOnDisk, _ = s.diskFileStats()
numFilesOnDisk, _, _ = s.diskFileStats(nil)
}
// Persister pause until the merger catches up to reduce the segment
@ -305,7 +302,7 @@ OUTER:
// let the watchers proceed if they lag behind
persistWatchers = notifyMergeWatchers(lastPersistedEpoch, persistWatchers)
numFilesOnDisk, _ = s.diskFileStats()
numFilesOnDisk, _, _ = s.diskFileStats(nil)
}
return lastMergedEpoch, persistWatchers
@ -360,13 +357,13 @@ var DefaultMinSegmentsForInMemoryMerge = 2
func (s *Scorch) persistSnapshotMaybeMerge(snapshot *IndexSnapshot) (
bool, error) {
// collect the in-memory zap segments (SegmentBase instances)
var sbs []*zap.SegmentBase
var sbs []segment.Segment
var sbsDrops []*roaring.Bitmap
var sbsIndexes []int
for i, segmentSnapshot := range snapshot.segment {
if sb, ok := segmentSnapshot.segment.(*zap.SegmentBase); ok {
sbs = append(sbs, sb)
if _, ok := segmentSnapshot.segment.(segment.PersistedSegment); !ok {
sbs = append(sbs, segmentSnapshot.segment)
sbsDrops = append(sbsDrops, segmentSnapshot.deleted)
sbsIndexes = append(sbsIndexes, i)
}
@ -377,7 +374,7 @@ func (s *Scorch) persistSnapshotMaybeMerge(snapshot *IndexSnapshot) (
}
newSnapshot, newSegmentID, err := s.mergeSegmentBases(
snapshot, sbs, sbsDrops, sbsIndexes, DefaultChunkFactor)
snapshot, sbs, sbsDrops, sbsIndexes)
if err != nil {
return false, err
}
@ -459,13 +456,13 @@ func (s *Scorch) persistSnapshotDirect(snapshot *IndexSnapshot) (err error) {
if err != nil {
return err
}
err = metaBucket.Put([]byte("type"), []byte(zap.Type))
err = metaBucket.Put(boltMetaDataSegmentTypeKey, []byte(s.segPlugin.Type()))
if err != nil {
return err
}
buf := make([]byte, binary.MaxVarintLen32)
binary.BigEndian.PutUint32(buf, zap.Version)
err = metaBucket.Put([]byte("version"), buf)
binary.BigEndian.PutUint32(buf, s.segPlugin.Version())
err = metaBucket.Put(boltMetaDataSegmentVersionKey, buf)
if err != nil {
return err
}
@ -494,11 +491,19 @@ func (s *Scorch) persistSnapshotDirect(snapshot *IndexSnapshot) (err error) {
return err
}
switch seg := segmentSnapshot.segment.(type) {
case *zap.SegmentBase:
case segment.PersistedSegment:
path := seg.Path()
filename := strings.TrimPrefix(path, s.path+string(os.PathSeparator))
err = snapshotSegmentBucket.Put(boltPathKey, []byte(filename))
if err != nil {
return err
}
filenames = append(filenames, filename)
case segment.UnpersistedSegment:
// need to persist this to disk
filename := zapFileName(segmentSnapshot.id)
path := s.path + string(os.PathSeparator) + filename
err = zap.PersistSegmentBase(seg, path)
err = seg.Persist(path)
if err != nil {
return fmt.Errorf("error persisting segment: %v", err)
}
@ -508,14 +513,7 @@ func (s *Scorch) persistSnapshotDirect(snapshot *IndexSnapshot) (err error) {
return err
}
filenames = append(filenames, filename)
case *zap.Segment:
path := seg.Path()
filename := strings.TrimPrefix(path, s.path+string(os.PathSeparator))
err = snapshotSegmentBucket.Put(boltPathKey, []byte(filename))
if err != nil {
return err
}
filenames = append(filenames, filename)
default:
return fmt.Errorf("unknown segment type: %T", seg)
}
@ -553,7 +551,7 @@ func (s *Scorch) persistSnapshotDirect(snapshot *IndexSnapshot) (err error) {
}
}()
for segmentID, path := range newSegmentPaths {
newSegments[segmentID], err = zap.Open(path)
newSegments[segmentID], err = s.segPlugin.Open(path)
if err != nil {
return fmt.Errorf("error opening new segment at %s, %v", path, err)
}
@ -609,6 +607,8 @@ var boltPathKey = []byte{'p'}
var boltDeletedKey = []byte{'d'}
var boltInternalKey = []byte{'i'}
var boltMetaDataKey = []byte{'m'}
var boltMetaDataSegmentTypeKey = []byte("type")
var boltMetaDataSegmentVersionKey = []byte("version")
func (s *Scorch) loadFromBolt() error {
return s.rootBolt.View(func(tx *bolt.Tx) error {
@ -693,6 +693,23 @@ func (s *Scorch) loadSnapshot(snapshot *bolt.Bucket) (*IndexSnapshot, error) {
refs: 1,
creator: "loadSnapshot",
}
// first we look for the meta-data bucket, this will tell us
// which segment type/version was used for this snapshot
// all operations for this scorch will use this type/version
metaBucket := snapshot.Bucket(boltMetaDataKey)
if metaBucket == nil {
_ = rv.DecRef()
return nil, fmt.Errorf("meta-data bucket missing")
}
segmentType := string(metaBucket.Get(boltMetaDataSegmentTypeKey))
segmentVersion := binary.BigEndian.Uint32(
metaBucket.Get(boltMetaDataSegmentVersionKey))
err := s.loadSegmentPlugin(segmentType, segmentVersion)
if err != nil {
_ = rv.DecRef()
return nil, fmt.Errorf(
"unable to load correct segment wrapper: %v", err)
}
var running uint64
c := snapshot.Cursor()
for k, _ := c.First(); k != nil; k, _ = c.Next() {
@ -737,7 +754,7 @@ func (s *Scorch) loadSegment(segmentBucket *bolt.Bucket) (*SegmentSnapshot, erro
return nil, fmt.Errorf("segment path missing")
}
segmentPath := s.path + string(os.PathSeparator) + string(pathBytes)
segment, err := zap.Open(segmentPath)
segment, err := s.segPlugin.Open(segmentPath)
if err != nil {
return nil, fmt.Errorf("error opening bolt segment: %v", err)
}

@ -28,10 +28,9 @@ import (
"github.com/blevesearch/bleve/document"
"github.com/blevesearch/bleve/index"
"github.com/blevesearch/bleve/index/scorch/segment"
"github.com/blevesearch/bleve/index/scorch/segment/zap"
"github.com/blevesearch/bleve/index/store"
"github.com/blevesearch/bleve/registry"
bolt "github.com/etcd-io/bbolt"
bolt "go.etcd.io/bbolt"
)
const Name = "scorch"
@ -67,7 +66,6 @@ type Scorch struct {
persists chan *persistIntroduction
merges chan *segmentMerge
introducerNotifier chan *epochWatcher
revertToSnapshots chan *snapshotReversion
persisterNotifier chan *epochWatcher
rootBolt *bolt.DB
asyncTasks sync.WaitGroup
@ -78,6 +76,8 @@ type Scorch struct {
pauseLock sync.RWMutex
pauseCount uint64
segPlugin segment.Plugin
}
type internalStats struct {
@ -101,7 +101,25 @@ func NewScorch(storeName string,
nextSnapshotEpoch: 1,
closeCh: make(chan struct{}),
ineligibleForRemoval: map[string]bool{},
segPlugin: defaultSegmentPlugin,
}
// check if the caller has requested a specific segment type/version
forcedSegmentVersion, ok := config["forceSegmentVersion"].(int)
if ok {
forcedSegmentType, ok2 := config["forceSegmentType"].(string)
if !ok2 {
return nil, fmt.Errorf(
"forceSegmentVersion set to %d, must also specify forceSegmentType", forcedSegmentVersion)
}
err := rv.loadSegmentPlugin(forcedSegmentType,
uint32(forcedSegmentVersion))
if err != nil {
return nil, err
}
}
rv.root = &IndexSnapshot{parent: rv, refs: 1, creator: "NewScorch"}
ro, ok := config["read_only"].(bool)
if ok {
@ -221,8 +239,8 @@ func (s *Scorch) openBolt() error {
s.persists = make(chan *persistIntroduction)
s.merges = make(chan *segmentMerge)
s.introducerNotifier = make(chan *epochWatcher, 1)
s.revertToSnapshots = make(chan *snapshotReversion)
s.persisterNotifier = make(chan *epochWatcher, 1)
s.closeCh = make(chan struct{})
if !s.readOnly && s.path != "" {
err := s.removeOldZapFiles() // Before persister or merger create any new files.
@ -263,7 +281,10 @@ func (s *Scorch) Close() (err error) {
err = s.rootBolt.Close()
s.rootLock.Lock()
if s.root != nil {
_ = s.root.DecRef()
err2 := s.root.DecRef()
if err == nil {
err = err2
}
}
s.root = nil
s.rootLock.Unlock()
@ -349,7 +370,7 @@ func (s *Scorch) Batch(batch *index.Batch) (err error) {
var newSegment segment.Segment
var bufBytes uint64
if len(analysisResults) > 0 {
newSegment, bufBytes, err = zap.AnalysisResultsToSegmentBase(analysisResults, DefaultChunkFactor)
newSegment, bufBytes, err = s.segPlugin.New(analysisResults)
if err != nil {
return err
}
@ -466,8 +487,9 @@ func (s *Scorch) Stats() json.Marshaler {
return &s.stats
}
func (s *Scorch) diskFileStats() (uint64, uint64) {
var numFilesOnDisk, numBytesUsedDisk uint64
func (s *Scorch) diskFileStats(rootSegmentPaths map[string]struct{}) (uint64,
uint64, uint64) {
var numFilesOnDisk, numBytesUsedDisk, numBytesOnDiskByRoot uint64
if s.path != "" {
finfos, err := ioutil.ReadDir(s.path)
if err == nil {
@ -475,24 +497,47 @@ func (s *Scorch) diskFileStats() (uint64, uint64) {
if !finfo.IsDir() {
numBytesUsedDisk += uint64(finfo.Size())
numFilesOnDisk++
if rootSegmentPaths != nil {
fname := s.path + string(os.PathSeparator) + finfo.Name()
if _, fileAtRoot := rootSegmentPaths[fname]; fileAtRoot {
numBytesOnDiskByRoot += uint64(finfo.Size())
}
}
}
}
}
}
// if no root files path given, then consider all disk files.
if rootSegmentPaths == nil {
return numFilesOnDisk, numBytesUsedDisk, numBytesUsedDisk
}
return numFilesOnDisk, numBytesUsedDisk
return numFilesOnDisk, numBytesUsedDisk, numBytesOnDiskByRoot
}
func (s *Scorch) rootDiskSegmentsPaths() map[string]struct{} {
rv := make(map[string]struct{}, len(s.root.segment))
for _, segmentSnapshot := range s.root.segment {
if seg, ok := segmentSnapshot.segment.(segment.PersistedSegment); ok {
rv[seg.Path()] = struct{}{}
}
}
return rv
}
func (s *Scorch) StatsMap() map[string]interface{} {
m := s.stats.ToMap()
numFilesOnDisk, numBytesUsedDisk := s.diskFileStats()
s.rootLock.RLock()
rootSegPaths := s.rootDiskSegmentsPaths()
m["CurFilesIneligibleForRemoval"] = uint64(len(s.ineligibleForRemoval))
s.rootLock.RUnlock()
numFilesOnDisk, numBytesUsedDisk, numBytesOnDiskByRoot := s.diskFileStats(rootSegPaths)
m["CurOnDiskBytes"] = numBytesUsedDisk
m["CurOnDiskFiles"] = numFilesOnDisk
s.rootLock.RLock()
m["CurFilesIneligibleForRemoval"] = uint64(len(s.ineligibleForRemoval))
s.rootLock.RUnlock()
// TODO: consider one day removing these backwards compatible
// names for apps using the old names
m["updates"] = m["TotUpdates"]
@ -507,8 +552,11 @@ func (s *Scorch) StatsMap() map[string]interface{} {
m["num_items_introduced"] = m["TotIntroducedItems"]
m["num_items_persisted"] = m["TotPersistedItems"]
m["num_recs_to_persist"] = m["TotItemsToPersist"]
m["num_bytes_used_disk"] = m["CurOnDiskBytes"]
m["num_files_on_disk"] = m["CurOnDiskFiles"]
// total disk bytes found in index directory inclusive of older snapshots
m["num_bytes_used_disk"] = numBytesUsedDisk
// total disk bytes by the latest root index, exclusive of older snapshots
m["num_bytes_used_disk_by_root"] = numBytesOnDiskByRoot
m["num_files_on_disk"] = numFilesOnDisk
m["num_root_memorysegments"] = m["TotMemorySegmentsAtRoot"]
m["num_root_filesegments"] = m["TotFileSegmentsAtRoot"]
m["num_persister_nap_pause_completed"] = m["TotPersisterNapPauseCompleted"]

@ -105,10 +105,6 @@ func (e *EmptyDictionaryIterator) Contains(key []byte) (bool, error) {
return false, nil
}
func (e *EmptyPostingsIterator) Advance(uint64) (Posting, error) {
return nil, nil
}
type EmptyPostingsList struct{}
func (e *EmptyPostingsList) Iterator(includeFreq, includeNorm, includeLocations bool,
@ -130,6 +126,10 @@ func (e *EmptyPostingsIterator) Next() (Posting, error) {
return nil, nil
}
func (e *EmptyPostingsIterator) Advance(uint64) (Posting, error) {
return nil, nil
}
func (e *EmptyPostingsIterator) Size() int {
return 0
}

@ -0,0 +1,58 @@
// Copyright (c) 2020 Couchbase, Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
package segment
import (
"github.com/RoaringBitmap/roaring"
"github.com/blevesearch/bleve/index"
)
// Plugin represents the essential functions required by a package to plug in
// it's segment implementation
type Plugin interface {
// Type is the name for this segment plugin
Type() string
// Version is a numeric value identifying a specific version of this type.
// When incompatible changes are made to a particular type of plugin, the
// version must be incremented.
Version() uint32
// New takes a set of AnalysisResults and turns them into a new Segment
New(results []*index.AnalysisResult) (Segment, uint64, error)
// Open attempts to open the file at the specified path and
// return the corresponding Segment
Open(path string) (Segment, error)
// Merge takes a set of Segments, and creates a new segment on disk at
// the specified path.
// Drops is a set of bitmaps (one for each segment) indicating which
// documents can be dropped from the segments during the merge.
// If the closeCh channel is closed, Merge will cease doing work at
// the next opportunity, and return an error (closed).
// StatsReporter can optionally be provided, in which case progress
// made during the merge is reported while operation continues.
// Returns:
// A slice of new document numbers (one for each input segment),
// this allows the caller to know a particular document's new
// document number in the newly merged segment.
// The number of bytes written to the new segment file.
// An error, if any occurred.
Merge(segments []Segment, drops []*roaring.Bitmap, path string,
closeCh chan struct{}, s StatsReporter) (
[][]uint64, uint64, error)
}

@ -50,6 +50,16 @@ type Segment interface {
DecRef() error
}
type UnpersistedSegment interface {
Segment
Persist(path string) error
}
type PersistedSegment interface {
Segment
Path() string
}
type TermDictionary interface {
PostingsList(term []byte, except *roaring.Bitmap, prealloc PostingsList) (PostingsList, error)
@ -96,6 +106,12 @@ type PostingsIterator interface {
Size() int
}
type OptimizablePostingsIterator interface {
ActualBitmap() *roaring.Bitmap
DocNum1Hit() (uint64, bool)
ReplaceActual(*roaring.Bitmap)
}
type Posting interface {
Number() uint64

@ -0,0 +1,148 @@
// Copyright (c) 2020 Couchbase, Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
package segment
import (
"github.com/RoaringBitmap/roaring"
"math"
"reflect"
)
var reflectStaticSizeUnadornedPostingsIteratorBitmap int
var reflectStaticSizeUnadornedPostingsIterator1Hit int
var reflectStaticSizeUnadornedPosting int
func init() {
var pib UnadornedPostingsIteratorBitmap
reflectStaticSizeUnadornedPostingsIteratorBitmap = int(reflect.TypeOf(pib).Size())
var pi1h UnadornedPostingsIterator1Hit
reflectStaticSizeUnadornedPostingsIterator1Hit = int(reflect.TypeOf(pi1h).Size())
var up UnadornedPosting
reflectStaticSizeUnadornedPosting = int(reflect.TypeOf(up).Size())
}
type UnadornedPostingsIteratorBitmap struct{
actual roaring.IntPeekable
actualBM *roaring.Bitmap
}
func (i *UnadornedPostingsIteratorBitmap) Next() (Posting, error) {
return i.nextAtOrAfter(0)
}
func (i *UnadornedPostingsIteratorBitmap) Advance(docNum uint64) (Posting, error) {
return i.nextAtOrAfter(docNum)
}
func (i *UnadornedPostingsIteratorBitmap) nextAtOrAfter(atOrAfter uint64) (Posting, error) {
docNum, exists := i.nextDocNumAtOrAfter(atOrAfter)
if !exists {
return nil, nil
}
return UnadornedPosting(docNum), nil
}
func (i *UnadornedPostingsIteratorBitmap) nextDocNumAtOrAfter(atOrAfter uint64) (uint64, bool) {
if i.actual == nil || !i.actual.HasNext() {
return 0, false
}
i.actual.AdvanceIfNeeded(uint32(atOrAfter))
if !i.actual.HasNext() {
return 0, false // couldn't find anything
}
return uint64(i.actual.Next()), true
}
func (i *UnadornedPostingsIteratorBitmap) Size() int {
return reflectStaticSizeUnadornedPostingsIteratorBitmap
}
func NewUnadornedPostingsIteratorFromBitmap(bm *roaring.Bitmap) PostingsIterator {
return &UnadornedPostingsIteratorBitmap{
actualBM: bm,
actual: bm.Iterator(),
}
}
const docNum1HitFinished = math.MaxUint64
type UnadornedPostingsIterator1Hit struct{
docNum uint64
}
func (i *UnadornedPostingsIterator1Hit) Next() (Posting, error) {
return i.nextAtOrAfter(0)
}
func (i *UnadornedPostingsIterator1Hit) Advance(docNum uint64) (Posting, error) {
return i.nextAtOrAfter(docNum)
}
func (i *UnadornedPostingsIterator1Hit) nextAtOrAfter(atOrAfter uint64) (Posting, error) {
docNum, exists := i.nextDocNumAtOrAfter(atOrAfter)
if !exists {
return nil, nil
}
return UnadornedPosting(docNum), nil
}
func (i *UnadornedPostingsIterator1Hit) nextDocNumAtOrAfter(atOrAfter uint64) (uint64, bool) {
if i.docNum == docNum1HitFinished {
return 0, false
}
if i.docNum < atOrAfter {
// advanced past our 1-hit
i.docNum = docNum1HitFinished // consume our 1-hit docNum
return 0, false
}
docNum := i.docNum
i.docNum = docNum1HitFinished // consume our 1-hit docNum
return docNum, true
}
func (i *UnadornedPostingsIterator1Hit) Size() int {
return reflectStaticSizeUnadornedPostingsIterator1Hit
}
func NewUnadornedPostingsIteratorFrom1Hit(docNum1Hit uint64) PostingsIterator {
return &UnadornedPostingsIterator1Hit{
docNum1Hit,
}
}
type UnadornedPosting uint64
func (p UnadornedPosting) Number() uint64 {
return uint64(p)
}
func (p UnadornedPosting) Frequency() uint64 {
return 0
}
func (p UnadornedPosting) Norm() float64 {
return 0
}
func (p UnadornedPosting) Locations() []Location {
return nil
}
func (p UnadornedPosting) Size() int {
return reflectStaticSizeUnadornedPosting
}

@ -0,0 +1,77 @@
// Copyright (c) 2019 Couchbase, Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
package scorch
import (
"fmt"
"github.com/blevesearch/bleve/index/scorch/segment"
zapv11 "github.com/blevesearch/zap/v11"
zapv12 "github.com/blevesearch/zap/v12"
)
var supportedSegmentPlugins map[string]map[uint32]segment.Plugin
var defaultSegmentPlugin segment.Plugin
func init() {
ResetPlugins()
RegisterPlugin(zapv12.Plugin(), false)
RegisterPlugin(zapv11.Plugin(), true)
}
func ResetPlugins() {
supportedSegmentPlugins = map[string]map[uint32]segment.Plugin{}
}
func RegisterPlugin(plugin segment.Plugin, makeDefault bool) {
if _, ok := supportedSegmentPlugins[plugin.Type()]; !ok {
supportedSegmentPlugins[plugin.Type()] = map[uint32]segment.Plugin{}
}
supportedSegmentPlugins[plugin.Type()][plugin.Version()] = plugin
if makeDefault {
defaultSegmentPlugin = plugin
}
}
func SupportedSegmentTypes() (rv []string) {
for k := range supportedSegmentPlugins {
rv = append(rv, k)
}
return
}
func SupportedSegmentTypeVersions(typ string) (rv []uint32) {
for k := range supportedSegmentPlugins[typ] {
rv = append(rv, k)
}
return rv
}
func (s *Scorch) loadSegmentPlugin(forcedSegmentType string,
forcedSegmentVersion uint32) error {
if versions, ok := supportedSegmentPlugins[forcedSegmentType]; ok {
if segPlugin, ok := versions[uint32(forcedSegmentVersion)]; ok {
s.segPlugin = segPlugin
return nil
}
return fmt.Errorf(
"unsupported version %d for segment type: %s, supported: %v",
forcedSegmentVersion, forcedSegmentType,
SupportedSegmentTypeVersions(forcedSegmentType))
}
return fmt.Errorf("unsupported segment type: %s, supported: %v",
forcedSegmentType, SupportedSegmentTypes())
}

@ -17,9 +17,10 @@ package scorch
import (
"fmt"
"log"
"os"
"github.com/blevesearch/bleve/index/scorch/segment"
bolt "github.com/etcd-io/bbolt"
bolt "go.etcd.io/bbolt"
)
type RollbackPoint struct {
@ -34,13 +35,22 @@ func (r *RollbackPoint) GetInternal(key []byte) []byte {
// RollbackPoints returns an array of rollback points available for
// the application to rollback to, with more recent rollback points
// (higher epochs) coming first.
func (s *Scorch) RollbackPoints() ([]*RollbackPoint, error) {
if s.rootBolt == nil {
return nil, fmt.Errorf("RollbackPoints: root is nil")
func RollbackPoints(path string) ([]*RollbackPoint, error) {
if len(path) == 0 {
return nil, fmt.Errorf("RollbackPoints: invalid path")
}
rootBoltPath := path + string(os.PathSeparator) + "root.bolt"
rootBoltOpt := &bolt.Options{
ReadOnly: true,
}
rootBolt, err := bolt.Open(rootBoltPath, 0600, rootBoltOpt)
if err != nil || rootBolt == nil {
return nil, err
}
// start a read-only bolt transaction
tx, err := s.rootBolt.Begin(false)
tx, err := rootBolt.Begin(false)
if err != nil {
return nil, fmt.Errorf("RollbackPoints: failed to start" +
" read-only transaction")
@ -49,6 +59,7 @@ func (s *Scorch) RollbackPoints() ([]*RollbackPoint, error) {
// read-only bolt transactions to be rolled back
defer func() {
_ = tx.Rollback()
_ = rootBolt.Close()
}()
snapshots := tx.Bucket(boltSnapshotsBucket)
@ -105,69 +116,98 @@ func (s *Scorch) RollbackPoints() ([]*RollbackPoint, error) {
return rollbackPoints, nil
}
// Rollback atomically and durably (if unsafeBatch is unset) brings
// the store back to the point in time as represented by the
// RollbackPoint. Rollback() should only be passed a RollbackPoint
// that came from the same store using the RollbackPoints() API.
func (s *Scorch) Rollback(to *RollbackPoint) error {
// Rollback atomically and durably brings the store back to the point
// in time as represented by the RollbackPoint.
// Rollback() should only be passed a RollbackPoint that came from the
// same store using the RollbackPoints() API along with the index path.
func Rollback(path string, to *RollbackPoint) error {
if to == nil {
return fmt.Errorf("Rollback: RollbackPoint is nil")
}
if s.rootBolt == nil {
return fmt.Errorf("Rollback: root is nil")
if len(path) == 0 {
return fmt.Errorf("Rollback: index path is empty")
}
revert := &snapshotReversion{}
s.rootLock.Lock()
rootBoltPath := path + string(os.PathSeparator) + "root.bolt"
rootBoltOpt := &bolt.Options{
ReadOnly: false,
}
rootBolt, err := bolt.Open(rootBoltPath, 0600, rootBoltOpt)
if err != nil || rootBolt == nil {
return err
}
defer func() {
err1 := rootBolt.Close()
if err1 != nil && err == nil {
err = err1
}
}()
err := s.rootBolt.View(func(tx *bolt.Tx) error {
// pick all the younger persisted epochs in bolt store
// including the target one.
var found bool
var eligibleEpochs []uint64
err = rootBolt.View(func(tx *bolt.Tx) error {
snapshots := tx.Bucket(boltSnapshotsBucket)
if snapshots == nil {
return fmt.Errorf("Rollback: no snapshots available")
}
pos := segment.EncodeUvarintAscending(nil, to.epoch)
snapshot := snapshots.Bucket(pos)
if snapshot == nil {
return fmt.Errorf("Rollback: snapshot not found")
return nil
}
indexSnapshot, err := s.loadSnapshot(snapshot)
sc := snapshots.Cursor()
for sk, _ := sc.Last(); sk != nil && !found; sk, _ = sc.Prev() {
_, snapshotEpoch, err := segment.DecodeUvarintAscending(sk)
if err != nil {
return fmt.Errorf("Rollback: unable to load snapshot: %v", err)
continue
}
// add segments referenced by loaded index snapshot to the
// ineligibleForRemoval map
for _, segSnap := range indexSnapshot.segment {
filename := zapFileName(segSnap.id)
s.ineligibleForRemoval[filename] = true
if snapshotEpoch == to.epoch {
found = true
}
eligibleEpochs = append(eligibleEpochs, snapshotEpoch)
}
revert.snapshot = indexSnapshot
revert.applied = make(chan error)
revert.persisted = make(chan error)
return nil
})
s.rootLock.Unlock()
if len(eligibleEpochs) == 0 {
return fmt.Errorf("Rollback: no persisted epochs found in bolt")
}
if !found {
return fmt.Errorf("Rollback: target epoch %d not found in bolt", to.epoch)
}
// start a write transaction
tx, err := rootBolt.Begin(true)
if err != nil {
return err
}
// introduce the reversion
s.revertToSnapshots <- revert
defer func() {
if err == nil {
err = tx.Commit()
} else {
_ = tx.Rollback()
}
if err == nil {
err = rootBolt.Sync()
}
}()
// block until this snapshot is applied
err = <-revert.applied
snapshots := tx.Bucket(boltSnapshotsBucket)
if snapshots == nil {
return nil
}
for _, epoch := range eligibleEpochs {
k := segment.EncodeUvarintAscending(nil, epoch)
if err != nil {
return fmt.Errorf("Rollback: failed with err: %v", err)
continue
}
if epoch == to.epoch {
// return here as it already processed until the given epoch
return nil
}
err = snapshots.DeleteBucket(k)
if err == bolt.ErrBucketNotFound {
err = nil
}
}
return <-revert.persisted
return err
}

@ -102,6 +102,8 @@ type Stats struct {
TotFileMergeZapEnd uint64
TotFileMergeZapTime uint64
MaxFileMergeZapTime uint64
TotFileMergeZapIntroductionTime uint64
MaxFileMergeZapIntroductionTime uint64
TotFileMergeIntroductions uint64
TotFileMergeIntroductionsDone uint64

@ -17,7 +17,7 @@ package boltdb
import (
"bytes"
bolt "github.com/etcd-io/bbolt"
bolt "go.etcd.io/bbolt"
)
type Iterator struct {

@ -16,7 +16,7 @@ package boltdb
import (
"github.com/blevesearch/bleve/index/store"
bolt "github.com/etcd-io/bbolt"
bolt "go.etcd.io/bbolt"
)
type Reader struct {

@ -30,7 +30,7 @@ import (
"github.com/blevesearch/bleve/index/store"
"github.com/blevesearch/bleve/registry"
bolt "github.com/etcd-io/bbolt"
bolt "go.etcd.io/bbolt"
)
const (

@ -820,7 +820,8 @@ func (udc *UpsideDownCouch) Batch(batch *index.Batch) (err error) {
if numUpdates > 0 {
go func() {
for _, doc := range batch.IndexOps {
for k := range batch.IndexOps {
doc := batch.IndexOps[k]
if doc != nil {
aw := index.NewAnalysisWork(udc, doc, resultChan)
// put the work on the queue

@ -17,6 +17,7 @@ package collector
import (
"context"
"reflect"
"strconv"
"time"
"github.com/blevesearch/bleve/index"
@ -90,6 +91,18 @@ func NewTopNCollectorAfter(size int, sort search.SortOrder, after []string) *Top
rv.searchAfter = &search.DocumentMatch{
Sort: after,
}
for pos, ss := range sort {
if ss.RequiresDocID() {
rv.searchAfter.ID = after[pos]
}
if ss.RequiresScoring() {
if score, err := strconv.ParseFloat(after[pos], 64); err == nil {
rv.searchAfter.Score = score
}
}
}
return rv
}

@ -58,6 +58,11 @@ OUTER:
// push back towards beginning
// without cross maxbegin
for start > 0 && used < s.fragmentSize {
if start > len(orig) {
// bail if out of bounds, possibly due to token replacement
// e.g with a regexp replacement
continue OUTER
}
r, size := utf8.DecodeLastRune(orig[0:start])
if r == utf8.RuneError {
continue OUTER // bail

@ -224,7 +224,8 @@ func ComputeGeoRange(term uint64, shift uint,
func buildRectFilter(dvReader index.DocValueReader, field string,
minLon, minLat, maxLon, maxLat float64) FilterFunc {
return func(d *search.DocumentMatch) bool {
var lon, lat float64
// check geo matches against all numeric type terms indexed
var lons, lats []float64
var found bool
err := dvReader.VisitDocValues(d.IndexInternalID, func(field string, term []byte) {
// only consider the values which are shifted 0
@ -234,15 +235,19 @@ func buildRectFilter(dvReader index.DocValueReader, field string,
var i64 int64
i64, err = prefixCoded.Int64()
if err == nil {
lon = geo.MortonUnhashLon(uint64(i64))
lat = geo.MortonUnhashLat(uint64(i64))
lons = append(lons, geo.MortonUnhashLon(uint64(i64)))
lats = append(lats, geo.MortonUnhashLat(uint64(i64)))
found = true
}
}
})
if err == nil && found {
return geo.BoundingBoxContains(lon, lat,
minLon, minLat, maxLon, maxLat)
for i := range lons {
if geo.BoundingBoxContains(lons[i], lats[i],
minLon, minLat, maxLon, maxLat) {
return true
}
}
}
return false
}

@ -83,7 +83,7 @@ func boxSearcher(indexReader index.IndexReader,
return boxSearcher, nil
}
// build geoboundinggox searcher for that bounding box
// build geoboundingbox searcher for that bounding box
boxSearcher, err := NewGeoBoundingBoxSearcher(indexReader,
topLeftLon, bottomRightLat, bottomRightLon, topLeftLat, field, boost,
options, checkBoundaries)
@ -96,7 +96,8 @@ func boxSearcher(indexReader index.IndexReader,
func buildDistFilter(dvReader index.DocValueReader, field string,
centerLon, centerLat, maxDist float64) FilterFunc {
return func(d *search.DocumentMatch) bool {
var lon, lat float64
// check geo matches against all numeric type terms indexed
var lons, lats []float64
var found bool
err := dvReader.VisitDocValues(d.IndexInternalID, func(field string, term []byte) {
@ -106,18 +107,20 @@ func buildDistFilter(dvReader index.DocValueReader, field string,
if err == nil && shift == 0 {
i64, err := prefixCoded.Int64()
if err == nil {
lon = geo.MortonUnhashLon(uint64(i64))
lat = geo.MortonUnhashLat(uint64(i64))
lons = append(lons, geo.MortonUnhashLon(uint64(i64)))
lats = append(lats, geo.MortonUnhashLat(uint64(i64)))
found = true
}
}
})
if err == nil && found {
dist := geo.Haversin(lon, lat, centerLon, centerLat)
for i := range lons {
dist := geo.Haversin(lons[i], lats[i], centerLon, centerLat)
if dist <= maxDist/1000 {
return true
}
}
}
return false
}
}

@ -15,6 +15,7 @@
package searcher
import (
"fmt"
"github.com/blevesearch/bleve/geo"
"github.com/blevesearch/bleve/index"
"github.com/blevesearch/bleve/numeric"
@ -26,6 +27,10 @@ func NewGeoBoundedPolygonSearcher(indexReader index.IndexReader,
polygon []geo.Point, field string, boost float64,
options search.SearcherOptions) (search.Searcher, error) {
if len(polygon) < 3 {
return nil, fmt.Errorf("Too few points specified for the polygon boundary")
}
// compute the bounding box enclosing the polygon
topLeftLon, topLeftLat, bottomRightLon, bottomRightLat, err :=
geo.BoundingRectangleForPolygon(polygon)
@ -63,7 +68,8 @@ func almostEqual(a, b float64) bool {
func buildPolygonFilter(dvReader index.DocValueReader, field string,
polygon []geo.Point) FilterFunc {
return func(d *search.DocumentMatch) bool {
var lon, lat float64
// check geo matches against all numeric type terms indexed
var lons, lats []float64
var found bool
err := dvReader.VisitDocValues(d.IndexInternalID, func(field string, term []byte) {
@ -73,8 +79,8 @@ func buildPolygonFilter(dvReader index.DocValueReader, field string,
if err == nil && shift == 0 {
i64, err := prefixCoded.Int64()
if err == nil {
lon = geo.MortonUnhashLon(uint64(i64))
lat = geo.MortonUnhashLat(uint64(i64))
lons = append(lons, geo.MortonUnhashLon(uint64(i64)))
lats = append(lats, geo.MortonUnhashLat(uint64(i64)))
found = true
}
}
@ -84,26 +90,36 @@ func buildPolygonFilter(dvReader index.DocValueReader, field string,
// the polygon. ie it might fail for certain points on the polygon boundaries.
if err == nil && found {
nVertices := len(polygon)
var inside bool
if len(polygon) < 3 {
return false
}
rayIntersectsSegment := func(point, a, b geo.Point) bool {
return (a.Lat > point.Lat) != (b.Lat > point.Lat) &&
point.Lon < (b.Lon-a.Lon)*(point.Lat-a.Lat)/(b.Lat-a.Lat)+a.Lon
}
for i := range lons {
pt := geo.Point{Lon: lons[i], Lat: lats[i]}
inside := rayIntersectsSegment(pt, polygon[len(polygon)-1], polygon[0])
// check for a direct vertex match
if almostEqual(polygon[0].Lat, lat) &&
almostEqual(polygon[0].Lon, lon) {
if almostEqual(polygon[0].Lat, lats[i]) &&
almostEqual(polygon[0].Lon, lons[i]) {
return true
}
for i := 1; i < nVertices; i++ {
if almostEqual(polygon[i].Lat, lat) &&
almostEqual(polygon[i].Lon, lon) {
for j := 1; j < nVertices; j++ {
if almostEqual(polygon[j].Lat, lats[i]) &&
almostEqual(polygon[j].Lon, lons[i]) {
return true
}
if (polygon[i].Lat > lat) != (polygon[i-1].Lat > lat) &&
lon < (polygon[i-1].Lon-polygon[i].Lon)*(lat-polygon[i].Lat)/
(polygon[i-1].Lat-polygon[i].Lat)+polygon[i].Lon {
if rayIntersectsSegment(pt, polygon[j-1], polygon[j]) {
inside = !inside
}
}
return inside
if inside {
return true
}
}
}
return false
}

@ -0,0 +1,3 @@
module github.com/blevesearch/go-porterstemmer
go 1.13

@ -0,0 +1,16 @@
language: go
os:
- linux
- osx
- windows
go:
- 1.11.4
env:
global:
- GO111MODULE=on
install:
- go mod download
- go get github.com/mattn/goveralls
script:
- go test -v -covermode=count -coverprofile=coverage.out -bench . -cpu 1,4
- '[ "${TRAVIS_PULL_REQUEST}" = "false" ] && $HOME/gopath/bin/goveralls -coverprofile=coverage.out -service=travis-ci -repotoken $COVERALLS_TOKEN || true'

@ -0,0 +1,3 @@
module github.com/blevesearch/mmap-go
require golang.org/x/sys v0.0.0-20181221143128-b4a75ba826a6

@ -0,0 +1,2 @@
golang.org/x/sys v0.0.0-20181221143128-b4a75ba826a6 h1:IcgEB62HYgAhX0Nd/QrVgZlxlcyxbGQHElLUhW2X4Fo=
golang.org/x/sys v0.0.0-20181221143128-b4a75ba826a6/go.mod h1:STP8DvDyc/dI5b8T5hshtkjS+E42TnysNCUPdjciGhY=

@ -24,6 +24,7 @@ import (
type addrinfo struct {
file windows.Handle
mapview windows.Handle
writable bool
}
var handleLock sync.Mutex
@ -32,13 +33,16 @@ var handleMap = map[uintptr]*addrinfo{}
func mmap(len int, prot, flags, hfile uintptr, off int64) ([]byte, error) {
flProtect := uint32(windows.PAGE_READONLY)
dwDesiredAccess := uint32(windows.FILE_MAP_READ)
writable := false
switch {
case prot&COPY != 0:
flProtect = windows.PAGE_WRITECOPY
dwDesiredAccess = windows.FILE_MAP_COPY
writable = true
case prot&RDWR != 0:
flProtect = windows.PAGE_READWRITE
dwDesiredAccess = windows.FILE_MAP_WRITE
writable = true
}
if prot&EXEC != 0 {
flProtect <<= 4
@ -69,6 +73,7 @@ func mmap(len int, prot, flags, hfile uintptr, off int64) ([]byte, error) {
handleMap[addr] = &addrinfo{
file: windows.Handle(hfile),
mapview: h,
writable: writable,
}
handleLock.Unlock()
@ -96,8 +101,13 @@ func (m MMap) flush() error {
return errors.New("unknown base address")
}
errno = windows.FlushFileBuffers(handle.file)
return os.NewSyscallError("FlushFileBuffers", errno)
if handle.writable {
if err := windows.FlushFileBuffers(handle.file); err != nil {
return os.NewSyscallError("FlushFileBuffers", err)
}
}
return nil
}
func (m MMap) lock() error {

@ -0,0 +1,3 @@
module github.com/blevesearch/segment
go 1.13

@ -0,0 +1,29 @@
Copyright (c) 2001, Dr Martin Porter
Copyright (c) 2004,2005, Richard Boulton
Copyright (c) 2013, Yoshiki Shibukawa
Copyright (c) 2006,2007,2009,2010,2011,2014-2019, Olly Betts
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions
are met:
1. Redistributions of source code must retain the above copyright notice,
this list of conditions and the following disclaimer.
2. Redistributions in binary form must reproduce the above copyright notice,
this list of conditions and the following disclaimer in the documentation
and/or other materials provided with the distribution.
3. Neither the name of the Snowball project nor the names of its contributors
may be used to endorse or promote products derived from this software
without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR
ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON
ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

@ -0,0 +1,66 @@
# snowballstem
This repository contains the Go stemmers generated by the [Snowball](https://github.com/snowballstem/snowball) project. They are maintained outside of the core bleve package so that they may be more easily be reused in other contexts.
## Usage
All these stemmers export a single `Stem()` method which operates on a snowball `Env` structure. The `Env` structure maintains all state for the stemmer. A new `Env` is created to point at an initial string. After stemming, the results of the `Stem()` operation can be retrieved using the `Current()` method. The `Env` structure can be reused for subsequent calls by using the `SetCurrent()` method.
## Example
```
package main
import (
"fmt"
"github.com/blevesearch/snowballstem"
"github.com/blevesearch/snowballstem/english"
)
func main() {
// words to stem
words := []string{
"running",
"jumping",
}
// build new environment
env := snowballstem.NewEnv("")
for _, word := range words {
// set up environment for word
env.SetCurrent(word)
// invoke stemmer
english.Stem(env)
// print results
fmt.Printf("%s stemmed to %s\n", word, env.Current())
}
}
```
Produces Output:
```
$ ./snowtest
running stemmed to run
jumping stemmed to jump
```
## Testing
The test harness for these stemmers is hosted in the main [Snowball](https://github.com/snowballstem/snowball) repository. There are functional tests built around the separate [snowballstem-data](https://github.com/snowballstem/snowball-data) repository, and there is support for fuzz-testing the stemmers there as well.
## Generating the Stemmers
```
$ export SNOWBALL=/path/to/github.com/snowballstem/snowball/after/snowball/built
$ go generate
```
## Updated the Go Generate Commands
A simple tool is provided to automate these from the snowball algorithms directory:
```
$ go run gengen.go /path/to/github.com/snowballstem/snowball/algorithms
```

@ -0,0 +1,16 @@
package snowballstem
import "fmt"
type AmongF func(env *Env, ctx interface{}) bool
type Among struct {
Str string
A int32
B int32
F AmongF
}
func (a *Among) String() string {
return fmt.Sprintf("str: `%s`, a: %d, b: %d, f: %p", a.Str, a.A, a.B, a.F)
}

File diff suppressed because it is too large Load Diff

@ -0,0 +1,389 @@
package snowballstem
import (
"log"
"strings"
"unicode/utf8"
)
// Env represents the Snowball execution environment
type Env struct {
current string
Cursor int
Limit int
LimitBackward int
Bra int
Ket int
}
// NewEnv creates a new Snowball execution environment on the provided string
func NewEnv(val string) *Env {
return &Env{
current: val,
Cursor: 0,
Limit: len(val),
LimitBackward: 0,
Bra: 0,
Ket: len(val),
}
}
func (env *Env) Current() string {
return env.current
}
func (env *Env) SetCurrent(s string) {
env.current = s
env.Cursor = 0
env.Limit = len(s)
env.LimitBackward = 0
env.Bra = 0
env.Ket = len(s)
}
func (env *Env) ReplaceS(bra, ket int, s string) int32 {
adjustment := int32(len(s)) - (int32(ket) - int32(bra))
result, _ := splitAt(env.current, bra)
rsplit := ket
if ket < bra {
rsplit = bra
}
_, rhs := splitAt(env.current, rsplit)
result += s
result += rhs
newLim := int32(env.Limit) + adjustment
env.Limit = int(newLim)
if env.Cursor >= ket {
newCur := int32(env.Cursor) + adjustment
env.Cursor = int(newCur)
} else if env.Cursor > bra {
env.Cursor = bra
}
env.current = result
return adjustment
}
func (env *Env) EqS(s string) bool {
if env.Cursor >= env.Limit {
return false
}
if strings.HasPrefix(env.current[env.Cursor:], s) {
env.Cursor += len(s)
for !onCharBoundary(env.current, env.Cursor) {
env.Cursor++
}
return true
}
return false
}
func (env *Env) EqSB(s string) bool {
if int32(env.Cursor)-int32(env.LimitBackward) < int32(len(s)) {
return false
} else if !onCharBoundary(env.current, env.Cursor-len(s)) ||
!strings.HasPrefix(env.current[env.Cursor-len(s):], s) {
return false
} else {
env.Cursor -= len(s)
return true
}
}
func (env *Env) SliceFrom(s string) bool {
bra, ket := env.Bra, env.Ket
env.ReplaceS(bra, ket, s)
return true
}
func (env *Env) NextChar() {
env.Cursor++
for !onCharBoundary(env.current, env.Cursor) {
env.Cursor++
}
}
func (env *Env) PrevChar() {
env.Cursor--
for !onCharBoundary(env.current, env.Cursor) {
env.Cursor--
}
}
func (env *Env) ByteIndexForHop(delta int32) int32 {
if delta > 0 {
res := env.Cursor
for delta > 0 {
res++
delta--
for res <= len(env.current) && !onCharBoundary(env.current, res) {
res++
}
}
return int32(res)
} else if delta < 0 {
res := env.Cursor
for delta < 0 {
res--
delta++
for res >= 0 && !onCharBoundary(env.current, res) {
res--
}
}
return int32(res)
} else {
return int32(env.Cursor)
}
}
func (env *Env) InGrouping(chars []byte, min, max int32) bool {
if env.Cursor >= env.Limit {
return false
}
r, _ := utf8.DecodeRuneInString(env.current[env.Cursor:])
if r != utf8.RuneError {
if r > max || r < min {
return false
}
r -= min
if (chars[uint(r>>3)] & (0x1 << uint(r&0x7))) == 0 {
return false
}
env.NextChar()
return true
}
return false
}
func (env *Env) InGroupingB(chars []byte, min, max int32) bool {
if env.Cursor <= env.LimitBackward {
return false
}
env.PrevChar()
r, _ := utf8.DecodeRuneInString(env.current[env.Cursor:])
if r != utf8.RuneError {
env.NextChar()
if r > max || r < min {
return false
}
r -= min
if (chars[uint(r>>3)] & (0x1 << uint(r&0x7))) == 0 {
return false
}
env.PrevChar()
return true
}
return false
}
func (env *Env) OutGrouping(chars []byte, min, max int32) bool {
if env.Cursor >= env.Limit {
return false
}
r, _ := utf8.DecodeRuneInString(env.current[env.Cursor:])
if r != utf8.RuneError {
if r > max || r < min {
env.NextChar()
return true
}
r -= min
if (chars[uint(r>>3)] & (0x1 << uint(r&0x7))) == 0 {
env.NextChar()
return true
}
}
return false
}
func (env *Env) OutGroupingB(chars []byte, min, max int32) bool {
if env.Cursor <= env.LimitBackward {
return false
}
env.PrevChar()
r, _ := utf8.DecodeRuneInString(env.current[env.Cursor:])
if r != utf8.RuneError {
env.NextChar()
if r > max || r < min {
env.PrevChar()
return true
}
r -= min
if (chars[uint(r>>3)] & (0x1 << uint(r&0x7))) == 0 {
env.PrevChar()
return true
}
}
return false
}
func (env *Env) SliceDel() bool {
return env.SliceFrom("")
}
func (env *Env) Insert(bra, ket int, s string) {
adjustment := env.ReplaceS(bra, ket, s)
if bra <= env.Bra {
env.Bra = int(int32(env.Bra) + adjustment)
}
if bra <= env.Ket {
env.Ket = int(int32(env.Ket) + adjustment)
}
}
func (env *Env) SliceTo() string {
return env.current[env.Bra:env.Ket]
}
func (env *Env) FindAmong(amongs []*Among, ctx interface{}) int32 {
var i int32
j := int32(len(amongs))
c := env.Cursor
l := env.Limit
var commonI, commonJ int
firstKeyInspected := false
for {
k := i + ((j - i) >> 1)
var diff int32
common := min(commonI, commonJ)
w := amongs[k]
for lvar := common; lvar < len(w.Str); lvar++ {
if c+common == l {
diff--
break
}
diff = int32(env.current[c+common]) - int32(w.Str[lvar])
if diff != 0 {
break
}
common++
}
if diff < 0 {
j = k
commonJ = common
} else {
i = k
commonI = common
}
if j-i <= 1 {
if i > 0 {
break
}
if j == i {
break
}
if firstKeyInspected {
break
}
firstKeyInspected = true
}
}
for {
w := amongs[i]
if commonI >= len(w.Str) {
env.Cursor = c + len(w.Str)
if w.F != nil {
res := w.F(env, ctx)
env.Cursor = c + len(w.Str)
if res {
return w.B
}
} else {
return w.B
}
}
i = w.A
if i < 0 {
return 0
}
}
}
func (env *Env) FindAmongB(amongs []*Among, ctx interface{}) int32 {
var i int32
j := int32(len(amongs))
c := env.Cursor
lb := env.LimitBackward
var commonI, commonJ int
firstKeyInspected := false
for {
k := i + ((j - i) >> 1)
diff := int32(0)
common := min(commonI, commonJ)
w := amongs[k]
for lvar := len(w.Str) - int(common) - 1; lvar >= 0; lvar-- {
if c-common == lb {
diff--
break
}
diff = int32(env.current[c-common-1]) - int32(w.Str[lvar])
if diff != 0 {
break
}
// Count up commons. But not one character but the byte width of that char
common++
}
if diff < 0 {
j = k
commonJ = common
} else {
i = k
commonI = common
}
if j-i <= 1 {
if i > 0 {
break
}
if j == i {
break
}
if firstKeyInspected {
break
}
firstKeyInspected = true
}
}
for {
w := amongs[i]
if commonI >= len(w.Str) {
env.Cursor = c - len(w.Str)
if w.F != nil {
res := w.F(env, ctx)
env.Cursor = c - len(w.Str)
if res {
return w.B
}
} else {
return w.B
}
}
i = w.A
if i < 0 {
return 0
}
}
}
func (env *Env) Debug(count, lineNumber int) {
log.Printf("snowball debug, count: %d, line: %d", count, lineNumber)
}
func (env *Env) Clone() *Env {
clone := *env
return &clone
}
func (env *Env) AssignTo() string {
return env.Current()
}

@ -0,0 +1,61 @@
package snowballstem
// to regenerate these commands, run
// go run gengen.go /path/to/snowball/algorithms/directory
//go:generate $SNOWBALL/snowball $SNOWBALL/algorithms/arabic/stem_Unicode.sbl -go -o arabic/arabic_stemmer -gop arabic -gor github.com/blevesearch/snowballstem
//go:generate gofmt -s -w arabic/arabic_stemmer.go
//go:generate $SNOWBALL/snowball $SNOWBALL/algorithms/danish/stem_ISO_8859_1.sbl -go -o danish/danish_stemmer -gop danish -gor github.com/blevesearch/snowballstem
//go:generate gofmt -s -w danish/danish_stemmer.go
//go:generate $SNOWBALL/snowball $SNOWBALL/algorithms/dutch/stem_ISO_8859_1.sbl -go -o dutch/dutch_stemmer -gop dutch -gor github.com/blevesearch/snowballstem
//go:generate gofmt -s -w dutch/dutch_stemmer.go
//go:generate $SNOWBALL/snowball $SNOWBALL/algorithms/english/stem_ISO_8859_1.sbl -go -o english/english_stemmer -gop english -gor github.com/blevesearch/snowballstem
//go:generate gofmt -s -w english/english_stemmer.go
//go:generate $SNOWBALL/snowball $SNOWBALL/algorithms/finnish/stem_ISO_8859_1.sbl -go -o finnish/finnish_stemmer -gop finnish -gor github.com/blevesearch/snowballstem
//go:generate gofmt -s -w finnish/finnish_stemmer.go
//go:generate $SNOWBALL/snowball $SNOWBALL/algorithms/french/stem_ISO_8859_1.sbl -go -o french/french_stemmer -gop french -gor github.com/blevesearch/snowballstem
//go:generate gofmt -s -w french/french_stemmer.go
//go:generate $SNOWBALL/snowball $SNOWBALL/algorithms/german/stem_ISO_8859_1.sbl -go -o german/german_stemmer -gop german -gor github.com/blevesearch/snowballstem
//go:generate gofmt -s -w german/german_stemmer.go
//go:generate $SNOWBALL/snowball $SNOWBALL/algorithms/hungarian/stem_Unicode.sbl -go -o hungarian/hungarian_stemmer -gop hungarian -gor github.com/blevesearch/snowballstem
//go:generate gofmt -s -w hungarian/hungarian_stemmer.go
//go:generate $SNOWBALL/snowball $SNOWBALL/algorithms/irish/stem_ISO_8859_1.sbl -go -o irish/irish_stemmer -gop irish -gor github.com/blevesearch/snowballstem
//go:generate gofmt -s -w irish/irish_stemmer.go
//go:generate $SNOWBALL/snowball $SNOWBALL/algorithms/italian/stem_ISO_8859_1.sbl -go -o italian/italian_stemmer -gop italian -gor github.com/blevesearch/snowballstem
//go:generate gofmt -s -w italian/italian_stemmer.go
//go:generate $SNOWBALL/snowball $SNOWBALL/algorithms/norwegian/stem_ISO_8859_1.sbl -go -o norwegian/norwegian_stemmer -gop norwegian -gor github.com/blevesearch/snowballstem
//go:generate gofmt -s -w norwegian/norwegian_stemmer.go
//go:generate $SNOWBALL/snowball $SNOWBALL/algorithms/porter/stem_ISO_8859_1.sbl -go -o porter/porter_stemmer -gop porter -gor github.com/blevesearch/snowballstem
//go:generate gofmt -s -w porter/porter_stemmer.go
//go:generate $SNOWBALL/snowball $SNOWBALL/algorithms/portuguese/stem_ISO_8859_1.sbl -go -o portuguese/portuguese_stemmer -gop portuguese -gor github.com/blevesearch/snowballstem
//go:generate gofmt -s -w portuguese/portuguese_stemmer.go
//go:generate $SNOWBALL/snowball $SNOWBALL/algorithms/romanian/stem_Unicode.sbl -go -o romanian/romanian_stemmer -gop romanian -gor github.com/blevesearch/snowballstem
//go:generate gofmt -s -w romanian/romanian_stemmer.go
//go:generate $SNOWBALL/snowball $SNOWBALL/algorithms/russian/stem_Unicode.sbl -go -o russian/russian_stemmer -gop russian -gor github.com/blevesearch/snowballstem
//go:generate gofmt -s -w russian/russian_stemmer.go
//go:generate $SNOWBALL/snowball $SNOWBALL/algorithms/spanish/stem_ISO_8859_1.sbl -go -o spanish/spanish_stemmer -gop spanish -gor github.com/blevesearch/snowballstem
//go:generate gofmt -s -w spanish/spanish_stemmer.go
//go:generate $SNOWBALL/snowball $SNOWBALL/algorithms/swedish/stem_ISO_8859_1.sbl -go -o swedish/swedish_stemmer -gop swedish -gor github.com/blevesearch/snowballstem
//go:generate gofmt -s -w swedish/swedish_stemmer.go
//go:generate $SNOWBALL/snowball $SNOWBALL/algorithms/tamil/stem_Unicode.sbl -go -o tamil/tamil_stemmer -gop tamil -gor github.com/blevesearch/snowballstem
//go:generate gofmt -s -w tamil/tamil_stemmer.go
//go:generate $SNOWBALL/snowball $SNOWBALL/algorithms/turkish/stem_Unicode.sbl -go -o turkish/turkish_stemmer -gop turkish -gor github.com/blevesearch/snowballstem
//go:generate gofmt -s -w turkish/turkish_stemmer.go

@ -0,0 +1,3 @@
module github.com/blevesearch/snowballstem
go 1.13

@ -0,0 +1,34 @@
package snowballstem
import (
"math"
"unicode/utf8"
)
const MaxInt = math.MaxInt32
const MinInt = math.MinInt32
func splitAt(str string, mid int) (string, string) {
return str[:mid], str[mid:]
}
func min(a, b int) int {
if a < b {
return a
}
return b
}
func onCharBoundary(s string, pos int) bool {
if pos <= 0 || pos >= len(s) {
return true
}
return utf8.RuneStart(s[pos])
}
// RuneCountInString is a wrapper around utf8.RuneCountInString
// this allows us to not have to conditionally include
// the utf8 package into some stemmers and not others
func RuneCountInString(str string) int {
return utf8.RuneCountInString(str)
}

@ -0,0 +1,12 @@
#*
*.sublime-*
*~
.#*
.project
.settings
**/.idea/
**/*.iml
.DS_Store
/cmd/zap/zap
*.test
tags

@ -0,0 +1,202 @@
Apache License
Version 2.0, January 2004
http://www.apache.org/licenses/
TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
1. Definitions.
"License" shall mean the terms and conditions for use, reproduction,
and distribution as defined by Sections 1 through 9 of this document.
"Licensor" shall mean the copyright owner or entity authorized by
the copyright owner that is granting the License.
"Legal Entity" shall mean the union of the acting entity and all
other entities that control, are controlled by, or are under common
control with that entity. For the purposes of this definition,
"control" means (i) the power, direct or indirect, to cause the
direction or management of such entity, whether by contract or
otherwise, or (ii) ownership of fifty percent (50%) or more of the
outstanding shares, or (iii) beneficial ownership of such entity.
"You" (or "Your") shall mean an individual or Legal Entity
exercising permissions granted by this License.
"Source" form shall mean the preferred form for making modifications,
including but not limited to software source code, documentation
source, and configuration files.
"Object" form shall mean any form resulting from mechanical
transformation or translation of a Source form, including but
not limited to compiled object code, generated documentation,
and conversions to other media types.
"Work" shall mean the work of authorship, whether in Source or
Object form, made available under the License, as indicated by a
copyright notice that is included in or attached to the work
(an example is provided in the Appendix below).
"Derivative Works" shall mean any work, whether in Source or Object
form, that is based on (or derived from) the Work and for which the
editorial revisions, annotations, elaborations, or other modifications
represent, as a whole, an original work of authorship. For the purposes
of this License, Derivative Works shall not include works that remain
separable from, or merely link (or bind by name) to the interfaces of,
the Work and Derivative Works thereof.
"Contribution" shall mean any work of authorship, including
the original version of the Work and any modifications or additions
to that Work or Derivative Works thereof, that is intentionally
submitted to Licensor for inclusion in the Work by the copyright owner
or by an individual or Legal Entity authorized to submit on behalf of
the copyright owner. For the purposes of this definition, "submitted"
means any form of electronic, verbal, or written communication sent
to the Licensor or its representatives, including but not limited to
communication on electronic mailing lists, source code control systems,
and issue tracking systems that are managed by, or on behalf of, the
Licensor for the purpose of discussing and improving the Work, but
excluding communication that is conspicuously marked or otherwise
designated in writing by the copyright owner as "Not a Contribution."
"Contributor" shall mean Licensor and any individual or Legal Entity
on behalf of whom a Contribution has been received by Licensor and
subsequently incorporated within the Work.
2. Grant of Copyright License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
copyright license to reproduce, prepare Derivative Works of,
publicly display, publicly perform, sublicense, and distribute the
Work and such Derivative Works in Source or Object form.
3. Grant of Patent License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
(except as stated in this section) patent license to make, have made,
use, offer to sell, sell, import, and otherwise transfer the Work,
where such license applies only to those patent claims licensable
by such Contributor that are necessarily infringed by their
Contribution(s) alone or by combination of their Contribution(s)
with the Work to which such Contribution(s) was submitted. If You
institute patent litigation against any entity (including a
cross-claim or counterclaim in a lawsuit) alleging that the Work
or a Contribution incorporated within the Work constitutes direct
or contributory patent infringement, then any patent licenses
granted to You under this License for that Work shall terminate
as of the date such litigation is filed.
4. Redistribution. You may reproduce and distribute copies of the
Work or Derivative Works thereof in any medium, with or without
modifications, and in Source or Object form, provided that You
meet the following conditions:
(a) You must give any other recipients of the Work or
Derivative Works a copy of this License; and
(b) You must cause any modified files to carry prominent notices
stating that You changed the files; and
(c) You must retain, in the Source form of any Derivative Works
that You distribute, all copyright, patent, trademark, and
attribution notices from the Source form of the Work,
excluding those notices that do not pertain to any part of
the Derivative Works; and
(d) If the Work includes a "NOTICE" text file as part of its
distribution, then any Derivative Works that You distribute must
include a readable copy of the attribution notices contained
within such NOTICE file, excluding those notices that do not
pertain to any part of the Derivative Works, in at least one
of the following places: within a NOTICE text file distributed
as part of the Derivative Works; within the Source form or
documentation, if provided along with the Derivative Works; or,
within a display generated by the Derivative Works, if and
wherever such third-party notices normally appear. The contents
of the NOTICE file are for informational purposes only and
do not modify the License. You may add Your own attribution
notices within Derivative Works that You distribute, alongside
or as an addendum to the NOTICE text from the Work, provided
that such additional attribution notices cannot be construed
as modifying the License.
You may add Your own copyright statement to Your modifications and
may provide additional or different license terms and conditions
for use, reproduction, or distribution of Your modifications, or
for any such Derivative Works as a whole, provided Your use,
reproduction, and distribution of the Work otherwise complies with
the conditions stated in this License.
5. Submission of Contributions. Unless You explicitly state otherwise,
any Contribution intentionally submitted for inclusion in the Work
by You to the Licensor shall be under the terms and conditions of
this License, without any additional terms or conditions.
Notwithstanding the above, nothing herein shall supersede or modify
the terms of any separate license agreement you may have executed
with Licensor regarding such Contributions.
6. Trademarks. This License does not grant permission to use the trade
names, trademarks, service marks, or product names of the Licensor,
except as required for reasonable and customary use in describing the
origin of the Work and reproducing the content of the NOTICE file.
7. Disclaimer of Warranty. Unless required by applicable law or
agreed to in writing, Licensor provides the Work (and each
Contributor provides its Contributions) on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
implied, including, without limitation, any warranties or conditions
of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
PARTICULAR PURPOSE. You are solely responsible for determining the
appropriateness of using or redistributing the Work and assume any
risks associated with Your exercise of permissions under this License.
8. Limitation of Liability. In no event and under no legal theory,
whether in tort (including negligence), contract, or otherwise,
unless required by applicable law (such as deliberate and grossly
negligent acts) or agreed to in writing, shall any Contributor be
liable to You for damages, including any direct, indirect, special,
incidental, or consequential damages of any character arising as a
result of this License or out of the use or inability to use the
Work (including but not limited to damages for loss of goodwill,
work stoppage, computer failure or malfunction, or any and all
other commercial damages or losses), even if such Contributor
has been advised of the possibility of such damages.
9. Accepting Warranty or Additional Liability. While redistributing
the Work or Derivative Works thereof, You may choose to offer,
and charge a fee for, acceptance of support, warranty, indemnity,
or other liability obligations and/or rights consistent with this
License. However, in accepting such obligations, You may act only
on Your own behalf and on Your sole responsibility, not on behalf
of any other Contributor, and only if You agree to indemnify,
defend, and hold each Contributor harmless for any liability
incurred by, or claims asserted against, such Contributor by reason
of your accepting any such warranty or additional liability.
END OF TERMS AND CONDITIONS
APPENDIX: How to apply the Apache License to your work.
To apply the Apache License to your work, attach the following
boilerplate notice, with the fields enclosed by brackets "[]"
replaced with your own identifying information. (Don't include
the brackets!) The text should be enclosed in the appropriate
comment syntax for the file format. We also recommend that a
file or class name and description of purpose be included on the
same "printed page" as the copyright notice for easier
identification within third-party archives.
Copyright [yyyy] [name of copyright owner]
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.

@ -16,9 +16,10 @@ package zap
import (
"bufio"
"github.com/couchbase/vellum"
"math"
"os"
"github.com/couchbase/vellum"
)
const Version uint32 = 11
@ -27,6 +28,10 @@ const Type string = "zap"
const fieldNotUninverted = math.MaxUint64
func (sb *SegmentBase) Persist(path string) error {
return PersistSegmentBase(sb, path)
}
// PersistSegmentBase persists SegmentBase in the zap file format.
func PersistSegmentBase(sb *SegmentBase, path string) error {
flag := os.O_RDWR | os.O_CREATE

@ -76,10 +76,6 @@ func (di *docValueReader) cloneInto(rv *docValueReader) *docValueReader {
return rv
}
func (di *docValueReader) fieldName() string {
return di.field
}
func (di *docValueReader) curChunkNumber() uint64 {
return di.curChunkNum
}

@ -0,0 +1,12 @@
module github.com/blevesearch/zap/v11
go 1.12
require (
github.com/RoaringBitmap/roaring v0.4.21
github.com/blevesearch/bleve v1.0.7
github.com/blevesearch/mmap-go v1.0.2
github.com/couchbase/vellum v1.0.1
github.com/golang/snappy v0.0.1
github.com/spf13/cobra v0.0.5
)

@ -31,32 +31,30 @@ import (
var DefaultFileMergerBufferSize = 1024 * 1024
// ValidateMerge can be set by applications to perform additional checks
// on a new segment produced by a merge, by default this does nothing.
// Caller should provide EITHER segments or memSegments, but not both.
// This API is experimental and may be removed at any time.
var ValidateMerge = func(segments []*Segment, memSegments []*SegmentBase, drops []*roaring.Bitmap, newSegment *Segment) error {
return nil
}
const docDropped = math.MaxUint64 // sentinel docNum to represent a deleted doc
// Merge takes a slice of zap segments and bit masks describing which
// Merge takes a slice of segments and bit masks describing which
// documents may be dropped, and creates a new segment containing the
// remaining data. This new segment is built at the specified path,
// with the provided chunkFactor.
func Merge(segments []*Segment, drops []*roaring.Bitmap, path string,
chunkFactor uint32, closeCh chan struct{}, s seg.StatsReporter) (
// remaining data. This new segment is built at the specified path.
func (*ZapPlugin) Merge(segments []seg.Segment, drops []*roaring.Bitmap, path string,
closeCh chan struct{}, s seg.StatsReporter) (
[][]uint64, uint64, error) {
segmentBases := make([]*SegmentBase, len(segments))
for segmenti, segment := range segments {
segmentBases[segmenti] = &segment.SegmentBase
switch segmentx := segment.(type) {
case *Segment:
segmentBases[segmenti] = &segmentx.SegmentBase
case *SegmentBase:
segmentBases[segmenti] = segmentx
default:
panic(fmt.Sprintf("oops, unexpected segment type: %T", segment))
}
return MergeSegmentBases(segmentBases, drops, path, chunkFactor, closeCh, s)
}
return mergeSegmentBases(segmentBases, drops, path, defaultChunkFactor, closeCh, s)
}
func MergeSegmentBases(segmentBases []*SegmentBase, drops []*roaring.Bitmap, path string,
func mergeSegmentBases(segmentBases []*SegmentBase, drops []*roaring.Bitmap, path string,
chunkFactor uint32, closeCh chan struct{}, s seg.StatsReporter) (
[][]uint64, uint64, error) {
flag := os.O_RDWR | os.O_CREATE

@ -25,6 +25,7 @@ import (
"github.com/blevesearch/bleve/analysis"
"github.com/blevesearch/bleve/document"
"github.com/blevesearch/bleve/index"
"github.com/blevesearch/bleve/index/scorch/segment"
"github.com/couchbase/vellum"
"github.com/golang/snappy"
)
@ -41,10 +42,17 @@ var ValidateDocFields = func(field document.Field) error {
return nil
}
var defaultChunkFactor uint32 = 1024
// AnalysisResultsToSegmentBase produces an in-memory zap-encoded
// SegmentBase from analysis results
func AnalysisResultsToSegmentBase(results []*index.AnalysisResult,
chunkFactor uint32) (*SegmentBase, uint64, error) {
func (z *ZapPlugin) New(results []*index.AnalysisResult) (
segment.Segment, uint64, error) {
return z.newWithChunkFactor(results, defaultChunkFactor)
}
func (*ZapPlugin) newWithChunkFactor(results []*index.AnalysisResult,
chunkFactor uint32) (segment.Segment, uint64, error) {
s := interimPool.Get().(*interim)
var br bytes.Buffer

@ -0,0 +1,37 @@
// Copyright (c) 2020 Couchbase, Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
package zap
import (
"github.com/blevesearch/bleve/index/scorch/segment"
)
// ZapPlugin implements the Plugin interface of
// the blevesearch/bleve/index/scorch/segment pkg
type ZapPlugin struct{}
func (*ZapPlugin) Type() string {
return Type
}
func (*ZapPlugin) Version() uint32 {
return Version
}
// Plugin returns an instance segment.Plugin for use
// by the Scorch indexing scheme
func Plugin() segment.Plugin {
return &ZapPlugin{}
}

@ -793,10 +793,23 @@ func (p *PostingsIterator) DocNum1Hit() (uint64, bool) {
return 0, false
}
// ActualBitmap returns the underlying actual bitmap
// which can be used up the stack for optimizations
func (p *PostingsIterator) ActualBitmap() *roaring.Bitmap {
return p.ActualBM
}
// ReplaceActual replaces the ActualBM with the provided
// bitmap
func (p *PostingsIterator) ReplaceActual(abm *roaring.Bitmap) {
p.ActualBM = abm
p.Actual = abm.Iterator()
}
// PostingsIteratorFromBitmap constructs a PostingsIterator given an
// "actual" bitmap.
func PostingsIteratorFromBitmap(bm *roaring.Bitmap,
includeFreqNorm, includeLocs bool) (*PostingsIterator, error) {
includeFreqNorm, includeLocs bool) (segment.PostingsIterator, error) {
return &PostingsIterator{
ActualBM: bm,
Actual: bm.Iterator(),
@ -807,11 +820,11 @@ func PostingsIteratorFromBitmap(bm *roaring.Bitmap,
// PostingsIteratorFrom1Hit constructs a PostingsIterator given a
// 1-hit docNum.
func PostingsIteratorFrom1Hit(docNum1Hit, normBits1Hit uint64,
includeFreqNorm, includeLocs bool) (*PostingsIterator, error) {
func PostingsIteratorFrom1Hit(docNum1Hit uint64,
includeFreqNorm, includeLocs bool) (segment.PostingsIterator, error) {
return &PostingsIterator{
docNum1Hit: docNum1Hit,
normBits1Hit: normBits1Hit,
normBits1Hit: NormBits1Hit,
includeFreqNorm: includeFreqNorm,
includeLocs: includeLocs,
}, nil

@ -27,7 +27,7 @@ import (
"github.com/blevesearch/bleve/index/scorch/segment"
"github.com/blevesearch/bleve/size"
"github.com/couchbase/vellum"
mmap "github.com/edsrzf/mmap-go"
mmap "github.com/blevesearch/mmap-go"
"github.com/golang/snappy"
)
@ -39,7 +39,7 @@ func init() {
}
// Open returns a zap impl of a segment
func Open(path string) (segment.Segment, error) {
func (*ZapPlugin) Open(path string) (segment.Segment, error) {
f, err := os.Open(path)
if err != nil {
return nil, err
@ -116,7 +116,7 @@ func (sb *SegmentBase) updateSize() {
cap(sb.mem)
// fieldsMap
for k, _ := range sb.fieldsMap {
for k := range sb.fieldsMap {
sizeInBytes += (len(k) + size.SizeOfString) + size.SizeOfUint16
}

@ -0,0 +1,12 @@
#*
*.sublime-*
*~
.#*
.project
.settings
**/.idea/
**/*.iml
.DS_Store
/cmd/zap/zap
*.test
tags

@ -0,0 +1,202 @@
Apache License
Version 2.0, January 2004
http://www.apache.org/licenses/
TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
1. Definitions.
"License" shall mean the terms and conditions for use, reproduction,
and distribution as defined by Sections 1 through 9 of this document.
"Licensor" shall mean the copyright owner or entity authorized by
the copyright owner that is granting the License.
"Legal Entity" shall mean the union of the acting entity and all
other entities that control, are controlled by, or are under common
control with that entity. For the purposes of this definition,
"control" means (i) the power, direct or indirect, to cause the
direction or management of such entity, whether by contract or
otherwise, or (ii) ownership of fifty percent (50%) or more of the
outstanding shares, or (iii) beneficial ownership of such entity.
"You" (or "Your") shall mean an individual or Legal Entity
exercising permissions granted by this License.
"Source" form shall mean the preferred form for making modifications,
including but not limited to software source code, documentation
source, and configuration files.
"Object" form shall mean any form resulting from mechanical
transformation or translation of a Source form, including but
not limited to compiled object code, generated documentation,
and conversions to other media types.
"Work" shall mean the work of authorship, whether in Source or
Object form, made available under the License, as indicated by a
copyright notice that is included in or attached to the work
(an example is provided in the Appendix below).
"Derivative Works" shall mean any work, whether in Source or Object
form, that is based on (or derived from) the Work and for which the
editorial revisions, annotations, elaborations, or other modifications
represent, as a whole, an original work of authorship. For the purposes
of this License, Derivative Works shall not include works that remain
separable from, or merely link (or bind by name) to the interfaces of,
the Work and Derivative Works thereof.
"Contribution" shall mean any work of authorship, including
the original version of the Work and any modifications or additions
to that Work or Derivative Works thereof, that is intentionally
submitted to Licensor for inclusion in the Work by the copyright owner
or by an individual or Legal Entity authorized to submit on behalf of
the copyright owner. For the purposes of this definition, "submitted"
means any form of electronic, verbal, or written communication sent
to the Licensor or its representatives, including but not limited to
communication on electronic mailing lists, source code control systems,
and issue tracking systems that are managed by, or on behalf of, the
Licensor for the purpose of discussing and improving the Work, but
excluding communication that is conspicuously marked or otherwise
designated in writing by the copyright owner as "Not a Contribution."
"Contributor" shall mean Licensor and any individual or Legal Entity
on behalf of whom a Contribution has been received by Licensor and
subsequently incorporated within the Work.
2. Grant of Copyright License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
copyright license to reproduce, prepare Derivative Works of,
publicly display, publicly perform, sublicense, and distribute the
Work and such Derivative Works in Source or Object form.
3. Grant of Patent License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
(except as stated in this section) patent license to make, have made,
use, offer to sell, sell, import, and otherwise transfer the Work,
where such license applies only to those patent claims licensable
by such Contributor that are necessarily infringed by their
Contribution(s) alone or by combination of their Contribution(s)
with the Work to which such Contribution(s) was submitted. If You
institute patent litigation against any entity (including a
cross-claim or counterclaim in a lawsuit) alleging that the Work
or a Contribution incorporated within the Work constitutes direct
or contributory patent infringement, then any patent licenses
granted to You under this License for that Work shall terminate
as of the date such litigation is filed.
4. Redistribution. You may reproduce and distribute copies of the
Work or Derivative Works thereof in any medium, with or without
modifications, and in Source or Object form, provided that You
meet the following conditions:
(a) You must give any other recipients of the Work or
Derivative Works a copy of this License; and
(b) You must cause any modified files to carry prominent notices
stating that You changed the files; and
(c) You must retain, in the Source form of any Derivative Works
that You distribute, all copyright, patent, trademark, and
attribution notices from the Source form of the Work,
excluding those notices that do not pertain to any part of
the Derivative Works; and
(d) If the Work includes a "NOTICE" text file as part of its
distribution, then any Derivative Works that You distribute must
include a readable copy of the attribution notices contained
within such NOTICE file, excluding those notices that do not
pertain to any part of the Derivative Works, in at least one
of the following places: within a NOTICE text file distributed
as part of the Derivative Works; within the Source form or
documentation, if provided along with the Derivative Works; or,
within a display generated by the Derivative Works, if and
wherever such third-party notices normally appear. The contents
of the NOTICE file are for informational purposes only and
do not modify the License. You may add Your own attribution
notices within Derivative Works that You distribute, alongside
or as an addendum to the NOTICE text from the Work, provided
that such additional attribution notices cannot be construed
as modifying the License.
You may add Your own copyright statement to Your modifications and
may provide additional or different license terms and conditions
for use, reproduction, or distribution of Your modifications, or
for any such Derivative Works as a whole, provided Your use,
reproduction, and distribution of the Work otherwise complies with
the conditions stated in this License.
5. Submission of Contributions. Unless You explicitly state otherwise,
any Contribution intentionally submitted for inclusion in the Work
by You to the Licensor shall be under the terms and conditions of
this License, without any additional terms or conditions.
Notwithstanding the above, nothing herein shall supersede or modify
the terms of any separate license agreement you may have executed
with Licensor regarding such Contributions.
6. Trademarks. This License does not grant permission to use the trade
names, trademarks, service marks, or product names of the Licensor,
except as required for reasonable and customary use in describing the
origin of the Work and reproducing the content of the NOTICE file.
7. Disclaimer of Warranty. Unless required by applicable law or
agreed to in writing, Licensor provides the Work (and each
Contributor provides its Contributions) on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
implied, including, without limitation, any warranties or conditions
of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
PARTICULAR PURPOSE. You are solely responsible for determining the
appropriateness of using or redistributing the Work and assume any
risks associated with Your exercise of permissions under this License.
8. Limitation of Liability. In no event and under no legal theory,
whether in tort (including negligence), contract, or otherwise,
unless required by applicable law (such as deliberate and grossly
negligent acts) or agreed to in writing, shall any Contributor be
liable to You for damages, including any direct, indirect, special,
incidental, or consequential damages of any character arising as a
result of this License or out of the use or inability to use the
Work (including but not limited to damages for loss of goodwill,
work stoppage, computer failure or malfunction, or any and all
other commercial damages or losses), even if such Contributor
has been advised of the possibility of such damages.
9. Accepting Warranty or Additional Liability. While redistributing
the Work or Derivative Works thereof, You may choose to offer,
and charge a fee for, acceptance of support, warranty, indemnity,
or other liability obligations and/or rights consistent with this
License. However, in accepting such obligations, You may act only
on Your own behalf and on Your sole responsibility, not on behalf
of any other Contributor, and only if You agree to indemnify,
defend, and hold each Contributor harmless for any liability
incurred by, or claims asserted against, such Contributor by reason
of your accepting any such warranty or additional liability.
END OF TERMS AND CONDITIONS
APPENDIX: How to apply the Apache License to your work.
To apply the Apache License to your work, attach the following
boilerplate notice, with the fields enclosed by brackets "[]"
replaced with your own identifying information. (Don't include
the brackets!) The text should be enclosed in the appropriate
comment syntax for the file format. We also recommend that a
file or class name and description of purpose be included on the
same "printed page" as the copyright notice for easier
identification within third-party archives.
Copyright [yyyy] [name of copyright owner]
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.

@ -0,0 +1,158 @@
# zap file format
Advanced ZAP File Format Documentation is [here](zap.md).
The file is written in the reverse order that we typically access data. This helps us write in one pass since later sections of the file require file offsets of things we've already written.
Current usage:
- mmap the entire file
- crc-32 bytes and version are in fixed position at end of the file
- reading remainder of footer could be version specific
- remainder of footer gives us:
- 3 important offsets (docValue , fields index and stored data index)
- 2 important values (number of docs and chunk factor)
- field data is processed once and memoized onto the heap so that we never have to go back to disk for it
- access to stored data by doc number means first navigating to the stored data index, then accessing a fixed position offset into that slice, which gives us the actual address of the data. the first bytes of that section tell us the size of data so that we know where it ends.
- access to all other indexed data follows the following pattern:
- first know the field name -> convert to id
- next navigate to term dictionary for that field
- some operations stop here and do dictionary ops
- next use dictionary to navigate to posting list for a specific term
- walk posting list
- if necessary, walk posting details as we go
- if location info is desired, consult location bitmap to see if it is there
## stored fields section
- for each document
- preparation phase:
- produce a slice of metadata bytes and data bytes
- produce these slices in field id order
- field value is appended to the data slice
- metadata slice is varint encoded with the following values for each field value
- field id (uint16)
- field type (byte)
- field value start offset in uncompressed data slice (uint64)
- field value length (uint64)
- field number of array positions (uint64)
- one additional value for each array position (uint64)
- compress the data slice using snappy
- file writing phase:
- remember the start offset for this document
- write out meta data length (varint uint64)
- write out compressed data length (varint uint64)
- write out the metadata bytes
- write out the compressed data bytes
## stored fields idx
- for each document
- write start offset (remembered from previous section) of stored data (big endian uint64)
With this index and a known document number, we have direct access to all the stored field data.
## posting details (freq/norm) section
- for each posting list
- produce a slice containing multiple consecutive chunks (each chunk is varint stream)
- produce a slice remembering offsets of where each chunk starts
- preparation phase:
- for each hit in the posting list
- if this hit is in next chunk close out encoding of last chunk and record offset start of next
- encode term frequency (uint64)
- encode norm factor (float32)
- file writing phase:
- remember start position for this posting list details
- write out number of chunks that follow (varint uint64)
- write out length of each chunk (each a varint uint64)
- write out the byte slice containing all the chunk data
If you know the doc number you're interested in, this format lets you jump to the correct chunk (docNum/chunkFactor) directly and then seek within that chunk until you find it.
## posting details (location) section
- for each posting list
- produce a slice containing multiple consecutive chunks (each chunk is varint stream)
- produce a slice remembering offsets of where each chunk starts
- preparation phase:
- for each hit in the posting list
- if this hit is in next chunk close out encoding of last chunk and record offset start of next
- encode field (uint16)
- encode field pos (uint64)
- encode field start (uint64)
- encode field end (uint64)
- encode number of array positions to follow (uint64)
- encode each array position (each uint64)
- file writing phase:
- remember start position for this posting list details
- write out number of chunks that follow (varint uint64)
- write out length of each chunk (each a varint uint64)
- write out the byte slice containing all the chunk data
If you know the doc number you're interested in, this format lets you jump to the correct chunk (docNum/chunkFactor) directly and then seek within that chunk until you find it.
## postings list section
- for each posting list
- preparation phase:
- encode roaring bitmap posting list to bytes (so we know the length)
- file writing phase:
- remember the start position for this posting list
- write freq/norm details offset (remembered from previous, as varint uint64)
- write location details offset (remembered from previous, as varint uint64)
- write length of encoded roaring bitmap
- write the serialized roaring bitmap data
## dictionary
- for each field
- preparation phase:
- encode vellum FST with dictionary data pointing to file offset of posting list (remembered from previous)
- file writing phase:
- remember the start position of this persistDictionary
- write length of vellum data (varint uint64)
- write out vellum data
## fields section
- for each field
- file writing phase:
- remember start offset for each field
- write dictionary address (remembered from previous) (varint uint64)
- write length of field name (varint uint64)
- write field name bytes
## fields idx
- for each field
- file writing phase:
- write big endian uint64 of start offset for each field
NOTE: currently we don't know or record the length of this fields index. Instead we rely on the fact that we know it immediately precedes a footer of known size.
## fields DocValue
- for each field
- preparation phase:
- produce a slice containing multiple consecutive chunks, where each chunk is composed of a meta section followed by compressed columnar field data
- produce a slice remembering the length of each chunk
- file writing phase:
- remember the start position of this first field DocValue offset in the footer
- write out number of chunks that follow (varint uint64)
- write out length of each chunk (each a varint uint64)
- write out the byte slice containing all the chunk data
NOTE: currently the meta header inside each chunk gives clue to the location offsets and size of the data pertaining to a given docID and any
read operation leverage that meta information to extract the document specific data from the file.
## footer
- file writing phase
- write number of docs (big endian uint64)
- write stored field index location (big endian uint64)
- write field index location (big endian uint64)
- write field docValue location (big endian uint64)
- write out chunk factor (big endian uint32)
- write out version (big endian uint32)
- write out file CRC of everything preceding this (big endian uint32)

@ -0,0 +1,156 @@
// Copyright (c) 2017 Couchbase, Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
package zap
import (
"bufio"
"math"
"os"
"github.com/couchbase/vellum"
)
const Version uint32 = 12
const Type string = "zap"
const fieldNotUninverted = math.MaxUint64
func (sb *SegmentBase) Persist(path string) error {
return PersistSegmentBase(sb, path)
}
// PersistSegmentBase persists SegmentBase in the zap file format.
func PersistSegmentBase(sb *SegmentBase, path string) error {
flag := os.O_RDWR | os.O_CREATE
f, err := os.OpenFile(path, flag, 0600)
if err != nil {
return err
}
cleanup := func() {
_ = f.Close()
_ = os.Remove(path)
}
br := bufio.NewWriter(f)
_, err = br.Write(sb.mem)
if err != nil {
cleanup()
return err
}
err = persistFooter(sb.numDocs, sb.storedIndexOffset, sb.fieldsIndexOffset, sb.docValueOffset,
sb.chunkMode, sb.memCRC, br)
if err != nil {
cleanup()
return err
}
err = br.Flush()
if err != nil {
cleanup()
return err
}
err = f.Sync()
if err != nil {
cleanup()
return err
}
err = f.Close()
if err != nil {
cleanup()
return err
}
return nil
}
func persistStoredFieldValues(fieldID int,
storedFieldValues [][]byte, stf []byte, spf [][]uint64,
curr int, metaEncode varintEncoder, data []byte) (
int, []byte, error) {
for i := 0; i < len(storedFieldValues); i++ {
// encode field
_, err := metaEncode(uint64(fieldID))
if err != nil {
return 0, nil, err
}
// encode type
_, err = metaEncode(uint64(stf[i]))
if err != nil {
return 0, nil, err
}
// encode start offset
_, err = metaEncode(uint64(curr))
if err != nil {
return 0, nil, err
}
// end len
_, err = metaEncode(uint64(len(storedFieldValues[i])))
if err != nil {
return 0, nil, err
}
// encode number of array pos
_, err = metaEncode(uint64(len(spf[i])))
if err != nil {
return 0, nil, err
}
// encode all array positions
for _, pos := range spf[i] {
_, err = metaEncode(pos)
if err != nil {
return 0, nil, err
}
}
data = append(data, storedFieldValues[i]...)
curr += len(storedFieldValues[i])
}
return curr, data, nil
}
func InitSegmentBase(mem []byte, memCRC uint32, chunkMode uint32,
fieldsMap map[string]uint16, fieldsInv []string, numDocs uint64,
storedIndexOffset uint64, fieldsIndexOffset uint64, docValueOffset uint64,
dictLocs []uint64) (*SegmentBase, error) {
sb := &SegmentBase{
mem: mem,
memCRC: memCRC,
chunkMode: chunkMode,
fieldsMap: fieldsMap,
fieldsInv: fieldsInv,
numDocs: numDocs,
storedIndexOffset: storedIndexOffset,
fieldsIndexOffset: fieldsIndexOffset,
docValueOffset: docValueOffset,
dictLocs: dictLocs,
fieldDvReaders: make(map[uint16]*docValueReader),
fieldFSTs: make(map[uint16]*vellum.FST),
}
sb.updateSize()
err := sb.loadDvReaders()
if err != nil {
return nil, err
}
return sb, nil
}

@ -0,0 +1,54 @@
// Copyright (c) 2019 Couchbase, Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
package zap
import (
"fmt"
)
// LegacyChunkMode was the original chunk mode (always chunk size 1024)
// this mode is still used for chunking doc values.
var LegacyChunkMode uint32 = 1024
// DefaultChunkMode is the most recent improvement to chunking and should
// be used by default.
var DefaultChunkMode uint32 = 1025
func getChunkSize(chunkMode uint32, cardinality uint64, maxDocs uint64) (uint64, error) {
switch {
// any chunkMode <= 1024 will always chunk with chunkSize=chunkMode
case chunkMode <= 1024:
// legacy chunk size
return uint64(chunkMode), nil
case chunkMode == 1025:
// attempt at simple improvement
// theory - the point of chunking is to put a bound on the maximum number of
// calls to Next() needed to find a random document. ie, you should be able
// to do one jump to the correct chunk, and then walk through at most
// chunk-size items
// previously 1024 was chosen as the chunk size, but this is particularly
// wasteful for low cardinality terms. the observation is that if there
// are less than 1024 items, why not put them all in one chunk,
// this way you'll still achieve the same goal of visiting at most
// chunk-size items.
// no attempt is made to tweak any other case
if cardinality <= 1024 {
return maxDocs, nil
}
return 1024, nil
}
return 0, fmt.Errorf("unknown chunk mode %d", chunkMode)
}

@ -0,0 +1,243 @@
// Copyright (c) 2017 Couchbase, Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
package zap
import (
"bytes"
"encoding/binary"
"io"
"reflect"
"github.com/golang/snappy"
)
var reflectStaticSizeMetaData int
func init() {
var md MetaData
reflectStaticSizeMetaData = int(reflect.TypeOf(md).Size())
}
var termSeparator byte = 0xff
var termSeparatorSplitSlice = []byte{termSeparator}
type chunkedContentCoder struct {
final []byte
chunkSize uint64
currChunk uint64
chunkLens []uint64
w io.Writer
progressiveWrite bool
chunkMetaBuf bytes.Buffer
chunkBuf bytes.Buffer
chunkMeta []MetaData
compressed []byte // temp buf for snappy compression
}
// MetaData represents the data information inside a
// chunk.
type MetaData struct {
DocNum uint64 // docNum of the data inside the chunk
DocDvOffset uint64 // offset of data inside the chunk for the given docid
}
// newChunkedContentCoder returns a new chunk content coder which
// packs data into chunks based on the provided chunkSize
func newChunkedContentCoder(chunkSize uint64, maxDocNum uint64,
w io.Writer, progressiveWrite bool) *chunkedContentCoder {
total := maxDocNum/chunkSize + 1
rv := &chunkedContentCoder{
chunkSize: chunkSize,
chunkLens: make([]uint64, total),
chunkMeta: make([]MetaData, 0, total),
w: w,
progressiveWrite: progressiveWrite,
}
return rv
}
// Reset lets you reuse this chunked content coder. Buffers are reset
// and re used. You cannot change the chunk size.
func (c *chunkedContentCoder) Reset() {
c.currChunk = 0
c.final = c.final[:0]
c.chunkBuf.Reset()
c.chunkMetaBuf.Reset()
for i := range c.chunkLens {
c.chunkLens[i] = 0
}
c.chunkMeta = c.chunkMeta[:0]
}
func (c *chunkedContentCoder) SetChunkSize(chunkSize uint64, maxDocNum uint64) {
total := int(maxDocNum/chunkSize + 1)
c.chunkSize = chunkSize
if cap(c.chunkLens) < total {
c.chunkLens = make([]uint64, total)
} else {
c.chunkLens = c.chunkLens[:total]
}
if cap(c.chunkMeta) < total {
c.chunkMeta = make([]MetaData, 0, total)
}
}
// Close indicates you are done calling Add() this allows
// the final chunk to be encoded.
func (c *chunkedContentCoder) Close() error {
return c.flushContents()
}
func (c *chunkedContentCoder) flushContents() error {
// flush the contents, with meta information at first
buf := make([]byte, binary.MaxVarintLen64)
n := binary.PutUvarint(buf, uint64(len(c.chunkMeta)))
_, err := c.chunkMetaBuf.Write(buf[:n])
if err != nil {
return err
}
// write out the metaData slice
for _, meta := range c.chunkMeta {
_, err := writeUvarints(&c.chunkMetaBuf, meta.DocNum, meta.DocDvOffset)
if err != nil {
return err
}
}
// write the metadata to final data
metaData := c.chunkMetaBuf.Bytes()
c.final = append(c.final, c.chunkMetaBuf.Bytes()...)
// write the compressed data to the final data
c.compressed = snappy.Encode(c.compressed[:cap(c.compressed)], c.chunkBuf.Bytes())
c.final = append(c.final, c.compressed...)
c.chunkLens[c.currChunk] = uint64(len(c.compressed) + len(metaData))
if c.progressiveWrite {
_, err := c.w.Write(c.final)
if err != nil {
return err
}
c.final = c.final[:0]
}
return nil
}
// Add encodes the provided byte slice into the correct chunk for the provided
// doc num. You MUST call Add() with increasing docNums.
func (c *chunkedContentCoder) Add(docNum uint64, vals []byte) error {
chunk := docNum / c.chunkSize
if chunk != c.currChunk {
// flush out the previous chunk details
err := c.flushContents()
if err != nil {
return err
}
// clearing the chunk specific meta for next chunk
c.chunkBuf.Reset()
c.chunkMetaBuf.Reset()
c.chunkMeta = c.chunkMeta[:0]
c.currChunk = chunk
}
// get the starting offset for this doc
dvOffset := c.chunkBuf.Len()
dvSize, err := c.chunkBuf.Write(vals)
if err != nil {
return err
}
c.chunkMeta = append(c.chunkMeta, MetaData{
DocNum: docNum,
DocDvOffset: uint64(dvOffset + dvSize),
})
return nil
}
// Write commits all the encoded chunked contents to the provided writer.
//
// | ..... data ..... | chunk offsets (varints)
// | position of chunk offsets (uint64) | number of offsets (uint64) |
//
func (c *chunkedContentCoder) Write() (int, error) {
var tw int
if c.final != nil {
// write out the data section first
nw, err := c.w.Write(c.final)
tw += nw
if err != nil {
return tw, err
}
}
chunkOffsetsStart := uint64(tw)
if cap(c.final) < binary.MaxVarintLen64 {
c.final = make([]byte, binary.MaxVarintLen64)
} else {
c.final = c.final[0:binary.MaxVarintLen64]
}
chunkOffsets := modifyLengthsToEndOffsets(c.chunkLens)
// write out the chunk offsets
for _, chunkOffset := range chunkOffsets {
n := binary.PutUvarint(c.final, chunkOffset)
nw, err := c.w.Write(c.final[:n])
tw += nw
if err != nil {
return tw, err
}
}
chunkOffsetsLen := uint64(tw) - chunkOffsetsStart
c.final = c.final[0:8]
// write out the length of chunk offsets
binary.BigEndian.PutUint64(c.final, chunkOffsetsLen)
nw, err := c.w.Write(c.final)
tw += nw
if err != nil {
return tw, err
}
// write out the number of chunks
binary.BigEndian.PutUint64(c.final, uint64(len(c.chunkLens)))
nw, err = c.w.Write(c.final)
tw += nw
if err != nil {
return tw, err
}
c.final = c.final[:0]
return tw, nil
}
// ReadDocValueBoundary elicits the start, end offsets from a
// metaData header slice
func ReadDocValueBoundary(chunk int, metaHeaders []MetaData) (uint64, uint64) {
var start uint64
if chunk > 0 {
start = metaHeaders[chunk-1].DocDvOffset
}
return start, metaHeaders[chunk].DocDvOffset
}

@ -0,0 +1,61 @@
// Copyright (c) 2017 Couchbase, Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
package zap
import (
"hash/crc32"
"io"
"github.com/blevesearch/bleve/index/scorch/segment"
)
// CountHashWriter is a wrapper around a Writer which counts the number of
// bytes which have been written and computes a crc32 hash
type CountHashWriter struct {
w io.Writer
crc uint32
n int
s segment.StatsReporter
}
// NewCountHashWriter returns a CountHashWriter which wraps the provided Writer
func NewCountHashWriter(w io.Writer) *CountHashWriter {
return &CountHashWriter{w: w}
}
func NewCountHashWriterWithStatsReporter(w io.Writer, s segment.StatsReporter) *CountHashWriter {
return &CountHashWriter{w: w, s: s}
}
// Write writes the provided bytes to the wrapped writer and counts the bytes
func (c *CountHashWriter) Write(b []byte) (int, error) {
n, err := c.w.Write(b)
c.crc = crc32.Update(c.crc, crc32.IEEETable, b[:n])
c.n += n
if c.s != nil {
c.s.ReportBytesWritten(uint64(n))
}
return n, err
}
// Count returns the number of bytes written
func (c *CountHashWriter) Count() int {
return c.n
}
// Sum32 returns the CRC-32 hash of the content written to this writer
func (c *CountHashWriter) Sum32() uint32 {
return c.crc
}

@ -0,0 +1,263 @@
// Copyright (c) 2017 Couchbase, Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
package zap
import (
"bytes"
"fmt"
"github.com/RoaringBitmap/roaring"
"github.com/blevesearch/bleve/index"
"github.com/blevesearch/bleve/index/scorch/segment"
"github.com/couchbase/vellum"
)
// Dictionary is the zap representation of the term dictionary
type Dictionary struct {
sb *SegmentBase
field string
fieldID uint16
fst *vellum.FST
fstReader *vellum.Reader
}
// PostingsList returns the postings list for the specified term
func (d *Dictionary) PostingsList(term []byte, except *roaring.Bitmap,
prealloc segment.PostingsList) (segment.PostingsList, error) {
var preallocPL *PostingsList
pl, ok := prealloc.(*PostingsList)
if ok && pl != nil {
preallocPL = pl
}
return d.postingsList(term, except, preallocPL)
}
func (d *Dictionary) postingsList(term []byte, except *roaring.Bitmap, rv *PostingsList) (*PostingsList, error) {
if d.fstReader == nil {
if rv == nil || rv == emptyPostingsList {
return emptyPostingsList, nil
}
return d.postingsListInit(rv, except), nil
}
postingsOffset, exists, err := d.fstReader.Get(term)
if err != nil {
return nil, fmt.Errorf("vellum err: %v", err)
}
if !exists {
if rv == nil || rv == emptyPostingsList {
return emptyPostingsList, nil
}
return d.postingsListInit(rv, except), nil
}
return d.postingsListFromOffset(postingsOffset, except, rv)
}
func (d *Dictionary) postingsListFromOffset(postingsOffset uint64, except *roaring.Bitmap, rv *PostingsList) (*PostingsList, error) {
rv = d.postingsListInit(rv, except)
err := rv.read(postingsOffset, d)
if err != nil {
return nil, err
}
return rv, nil
}
func (d *Dictionary) postingsListInit(rv *PostingsList, except *roaring.Bitmap) *PostingsList {
if rv == nil || rv == emptyPostingsList {
rv = &PostingsList{}
} else {
postings := rv.postings
if postings != nil {
postings.Clear()
}
*rv = PostingsList{} // clear the struct
rv.postings = postings
}
rv.sb = d.sb
rv.except = except
return rv
}
func (d *Dictionary) Contains(key []byte) (bool, error) {
return d.fst.Contains(key)
}
// Iterator returns an iterator for this dictionary
func (d *Dictionary) Iterator() segment.DictionaryIterator {
rv := &DictionaryIterator{
d: d,
}
if d.fst != nil {
itr, err := d.fst.Iterator(nil, nil)
if err == nil {
rv.itr = itr
} else if err != vellum.ErrIteratorDone {
rv.err = err
}
}
return rv
}
// PrefixIterator returns an iterator which only visits terms having the
// the specified prefix
func (d *Dictionary) PrefixIterator(prefix string) segment.DictionaryIterator {
rv := &DictionaryIterator{
d: d,
}
kBeg := []byte(prefix)
kEnd := segment.IncrementBytes(kBeg)
if d.fst != nil {
itr, err := d.fst.Iterator(kBeg, kEnd)
if err == nil {
rv.itr = itr
} else if err != vellum.ErrIteratorDone {
rv.err = err
}
}
return rv
}
// RangeIterator returns an iterator which only visits terms between the
// start and end terms. NOTE: bleve.index API specifies the end is inclusive.
func (d *Dictionary) RangeIterator(start, end string) segment.DictionaryIterator {
rv := &DictionaryIterator{
d: d,
}
// need to increment the end position to be inclusive
var endBytes []byte
if len(end) > 0 {
endBytes = []byte(end)
if endBytes[len(endBytes)-1] < 0xff {
endBytes[len(endBytes)-1]++
} else {
endBytes = append(endBytes, 0xff)
}
}
if d.fst != nil {
itr, err := d.fst.Iterator([]byte(start), endBytes)
if err == nil {
rv.itr = itr
} else if err != vellum.ErrIteratorDone {
rv.err = err
}
}
return rv
}
// AutomatonIterator returns an iterator which only visits terms
// having the the vellum automaton and start/end key range
func (d *Dictionary) AutomatonIterator(a vellum.Automaton,
startKeyInclusive, endKeyExclusive []byte) segment.DictionaryIterator {
rv := &DictionaryIterator{
d: d,
}
if d.fst != nil {
itr, err := d.fst.Search(a, startKeyInclusive, endKeyExclusive)
if err == nil {
rv.itr = itr
} else if err != vellum.ErrIteratorDone {
rv.err = err
}
}
return rv
}
func (d *Dictionary) OnlyIterator(onlyTerms [][]byte,
includeCount bool) segment.DictionaryIterator {
rv := &DictionaryIterator{
d: d,
omitCount: !includeCount,
}
var buf bytes.Buffer
builder, err := vellum.New(&buf, nil)
if err != nil {
rv.err = err
return rv
}
for _, term := range onlyTerms {
err = builder.Insert(term, 0)
if err != nil {
rv.err = err
return rv
}
}
err = builder.Close()
if err != nil {
rv.err = err
return rv
}
onlyFST, err := vellum.Load(buf.Bytes())
if err != nil {
rv.err = err
return rv
}
itr, err := d.fst.Search(onlyFST, nil, nil)
if err == nil {
rv.itr = itr
} else if err != vellum.ErrIteratorDone {
rv.err = err
}
return rv
}
// DictionaryIterator is an iterator for term dictionary
type DictionaryIterator struct {
d *Dictionary
itr vellum.Iterator
err error
tmp PostingsList
entry index.DictEntry
omitCount bool
}
// Next returns the next entry in the dictionary
func (i *DictionaryIterator) Next() (*index.DictEntry, error) {
if i.err != nil && i.err != vellum.ErrIteratorDone {
return nil, i.err
} else if i.itr == nil || i.err == vellum.ErrIteratorDone {
return nil, nil
}
term, postingsOffset := i.itr.Current()
i.entry.Term = string(term)
if !i.omitCount {
i.err = i.tmp.read(postingsOffset, i.d)
if i.err != nil {
return nil, i.err
}
i.entry.Count = i.tmp.Count()
}
i.err = i.itr.Next()
return &i.entry, nil
}

@ -0,0 +1,312 @@
// Copyright (c) 2017 Couchbase, Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
package zap
import (
"bytes"
"encoding/binary"
"fmt"
"math"
"reflect"
"sort"
"github.com/blevesearch/bleve/index"
"github.com/blevesearch/bleve/index/scorch/segment"
"github.com/blevesearch/bleve/size"
"github.com/golang/snappy"
)
var reflectStaticSizedocValueReader int
func init() {
var dvi docValueReader
reflectStaticSizedocValueReader = int(reflect.TypeOf(dvi).Size())
}
type docNumTermsVisitor func(docNum uint64, terms []byte) error
type docVisitState struct {
dvrs map[uint16]*docValueReader
segment *SegmentBase
}
type docValueReader struct {
field string
curChunkNum uint64
chunkOffsets []uint64
dvDataLoc uint64
curChunkHeader []MetaData
curChunkData []byte // compressed data cache
uncompressed []byte // temp buf for snappy decompression
}
func (di *docValueReader) size() int {
return reflectStaticSizedocValueReader + size.SizeOfPtr +
len(di.field) +
len(di.chunkOffsets)*size.SizeOfUint64 +
len(di.curChunkHeader)*reflectStaticSizeMetaData +
len(di.curChunkData)
}
func (di *docValueReader) cloneInto(rv *docValueReader) *docValueReader {
if rv == nil {
rv = &docValueReader{}
}
rv.field = di.field
rv.curChunkNum = math.MaxUint64
rv.chunkOffsets = di.chunkOffsets // immutable, so it's sharable
rv.dvDataLoc = di.dvDataLoc
rv.curChunkHeader = rv.curChunkHeader[:0]
rv.curChunkData = nil
rv.uncompressed = rv.uncompressed[:0]
return rv
}
func (di *docValueReader) curChunkNumber() uint64 {
return di.curChunkNum
}
func (s *SegmentBase) loadFieldDocValueReader(field string,
fieldDvLocStart, fieldDvLocEnd uint64) (*docValueReader, error) {
// get the docValue offset for the given fields
if fieldDvLocStart == fieldNotUninverted {
// no docValues found, nothing to do
return nil, nil
}
// read the number of chunks, and chunk offsets position
var numChunks, chunkOffsetsPosition uint64
if fieldDvLocEnd-fieldDvLocStart > 16 {
numChunks = binary.BigEndian.Uint64(s.mem[fieldDvLocEnd-8 : fieldDvLocEnd])
// read the length of chunk offsets
chunkOffsetsLen := binary.BigEndian.Uint64(s.mem[fieldDvLocEnd-16 : fieldDvLocEnd-8])
// acquire position of chunk offsets
chunkOffsetsPosition = (fieldDvLocEnd - 16) - chunkOffsetsLen
} else {
return nil, fmt.Errorf("loadFieldDocValueReader: fieldDvLoc too small: %d-%d", fieldDvLocEnd, fieldDvLocStart)
}
fdvIter := &docValueReader{
curChunkNum: math.MaxUint64,
field: field,
chunkOffsets: make([]uint64, int(numChunks)),
}
// read the chunk offsets
var offset uint64
for i := 0; i < int(numChunks); i++ {
loc, read := binary.Uvarint(s.mem[chunkOffsetsPosition+offset : chunkOffsetsPosition+offset+binary.MaxVarintLen64])
if read <= 0 {
return nil, fmt.Errorf("corrupted chunk offset during segment load")
}
fdvIter.chunkOffsets[i] = loc
offset += uint64(read)
}
// set the data offset
fdvIter.dvDataLoc = fieldDvLocStart
return fdvIter, nil
}
func (di *docValueReader) loadDvChunk(chunkNumber uint64, s *SegmentBase) error {
// advance to the chunk where the docValues
// reside for the given docNum
destChunkDataLoc, curChunkEnd := di.dvDataLoc, di.dvDataLoc
start, end := readChunkBoundary(int(chunkNumber), di.chunkOffsets)
if start >= end {
di.curChunkHeader = di.curChunkHeader[:0]
di.curChunkData = nil
di.curChunkNum = chunkNumber
di.uncompressed = di.uncompressed[:0]
return nil
}
destChunkDataLoc += start
curChunkEnd += end
// read the number of docs reside in the chunk
numDocs, read := binary.Uvarint(s.mem[destChunkDataLoc : destChunkDataLoc+binary.MaxVarintLen64])
if read <= 0 {
return fmt.Errorf("failed to read the chunk")
}
chunkMetaLoc := destChunkDataLoc + uint64(read)
offset := uint64(0)
if cap(di.curChunkHeader) < int(numDocs) {
di.curChunkHeader = make([]MetaData, int(numDocs))
} else {
di.curChunkHeader = di.curChunkHeader[:int(numDocs)]
}
for i := 0; i < int(numDocs); i++ {
di.curChunkHeader[i].DocNum, read = binary.Uvarint(s.mem[chunkMetaLoc+offset : chunkMetaLoc+offset+binary.MaxVarintLen64])
offset += uint64(read)
di.curChunkHeader[i].DocDvOffset, read = binary.Uvarint(s.mem[chunkMetaLoc+offset : chunkMetaLoc+offset+binary.MaxVarintLen64])
offset += uint64(read)
}
compressedDataLoc := chunkMetaLoc + offset
dataLength := curChunkEnd - compressedDataLoc
di.curChunkData = s.mem[compressedDataLoc : compressedDataLoc+dataLength]
di.curChunkNum = chunkNumber
di.uncompressed = di.uncompressed[:0]
return nil
}
func (di *docValueReader) iterateAllDocValues(s *SegmentBase, visitor docNumTermsVisitor) error {
for i := 0; i < len(di.chunkOffsets); i++ {
err := di.loadDvChunk(uint64(i), s)
if err != nil {
return err
}
if di.curChunkData == nil || len(di.curChunkHeader) == 0 {
continue
}
// uncompress the already loaded data
uncompressed, err := snappy.Decode(di.uncompressed[:cap(di.uncompressed)], di.curChunkData)
if err != nil {
return err
}
di.uncompressed = uncompressed
start := uint64(0)
for _, entry := range di.curChunkHeader {
err = visitor(entry.DocNum, uncompressed[start:entry.DocDvOffset])
if err != nil {
return err
}
start = entry.DocDvOffset
}
}
return nil
}
func (di *docValueReader) visitDocValues(docNum uint64,
visitor index.DocumentFieldTermVisitor) error {
// binary search the term locations for the docNum
start, end := di.getDocValueLocs(docNum)
if start == math.MaxUint64 || end == math.MaxUint64 || start == end {
return nil
}
var uncompressed []byte
var err error
// use the uncompressed copy if available
if len(di.uncompressed) > 0 {
uncompressed = di.uncompressed
} else {
// uncompress the already loaded data
uncompressed, err = snappy.Decode(di.uncompressed[:cap(di.uncompressed)], di.curChunkData)
if err != nil {
return err
}
di.uncompressed = uncompressed
}
// pick the terms for the given docNum
uncompressed = uncompressed[start:end]
for {
i := bytes.Index(uncompressed, termSeparatorSplitSlice)
if i < 0 {
break
}
visitor(di.field, uncompressed[0:i])
uncompressed = uncompressed[i+1:]
}
return nil
}
func (di *docValueReader) getDocValueLocs(docNum uint64) (uint64, uint64) {
i := sort.Search(len(di.curChunkHeader), func(i int) bool {
return di.curChunkHeader[i].DocNum >= docNum
})
if i < len(di.curChunkHeader) && di.curChunkHeader[i].DocNum == docNum {
return ReadDocValueBoundary(i, di.curChunkHeader)
}
return math.MaxUint64, math.MaxUint64
}
// VisitDocumentFieldTerms is an implementation of the
// DocumentFieldTermVisitable interface
func (s *SegmentBase) VisitDocumentFieldTerms(localDocNum uint64, fields []string,
visitor index.DocumentFieldTermVisitor, dvsIn segment.DocVisitState) (
segment.DocVisitState, error) {
dvs, ok := dvsIn.(*docVisitState)
if !ok || dvs == nil {
dvs = &docVisitState{}
} else {
if dvs.segment != s {
dvs.segment = s
dvs.dvrs = nil
}
}
var fieldIDPlus1 uint16
if dvs.dvrs == nil {
dvs.dvrs = make(map[uint16]*docValueReader, len(fields))
for _, field := range fields {
if fieldIDPlus1, ok = s.fieldsMap[field]; !ok {
continue
}
fieldID := fieldIDPlus1 - 1
if dvIter, exists := s.fieldDvReaders[fieldID]; exists &&
dvIter != nil {
dvs.dvrs[fieldID] = dvIter.cloneInto(dvs.dvrs[fieldID])
}
}
}
// find the chunkNumber where the docValues are stored
// NOTE: doc values continue to use legacy chunk mode
chunkFactor, err := getChunkSize(LegacyChunkMode, 0, 0)
if err != nil {
return nil, err
}
docInChunk := localDocNum / chunkFactor
var dvr *docValueReader
for _, field := range fields {
if fieldIDPlus1, ok = s.fieldsMap[field]; !ok {
continue
}
fieldID := fieldIDPlus1 - 1
if dvr, ok = dvs.dvrs[fieldID]; ok && dvr != nil {
// check if the chunk is already loaded
if docInChunk != dvr.curChunkNumber() {
err := dvr.loadDvChunk(docInChunk, s)
if err != nil {
return dvs, err
}
}
_ = dvr.visitDocValues(localDocNum, visitor)
}
}
return dvs, nil
}
// VisitableDocValueFields returns the list of fields with
// persisted doc value terms ready to be visitable using the
// VisitDocumentFieldTerms method.
func (s *SegmentBase) VisitableDocValueFields() ([]string, error) {
return s.fieldDvNames, nil
}

@ -0,0 +1,138 @@
// Copyright (c) 2018 Couchbase, Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
package zap
import (
"bytes"
"github.com/couchbase/vellum"
)
// enumerator provides an ordered traversal of multiple vellum
// iterators. Like JOIN of iterators, the enumerator produces a
// sequence of (key, iteratorIndex, value) tuples, sorted by key ASC,
// then iteratorIndex ASC, where the same key might be seen or
// repeated across multiple child iterators.
type enumerator struct {
itrs []vellum.Iterator
currKs [][]byte
currVs []uint64
lowK []byte
lowIdxs []int
lowCurr int
}
// newEnumerator returns a new enumerator over the vellum Iterators
func newEnumerator(itrs []vellum.Iterator) (*enumerator, error) {
rv := &enumerator{
itrs: itrs,
currKs: make([][]byte, len(itrs)),
currVs: make([]uint64, len(itrs)),
lowIdxs: make([]int, 0, len(itrs)),
}
for i, itr := range rv.itrs {
rv.currKs[i], rv.currVs[i] = itr.Current()
}
rv.updateMatches(false)
if rv.lowK == nil && len(rv.lowIdxs) == 0 {
return rv, vellum.ErrIteratorDone
}
return rv, nil
}
// updateMatches maintains the low key matches based on the currKs
func (m *enumerator) updateMatches(skipEmptyKey bool) {
m.lowK = nil
m.lowIdxs = m.lowIdxs[:0]
m.lowCurr = 0
for i, key := range m.currKs {
if (key == nil && m.currVs[i] == 0) || // in case of empty iterator
(len(key) == 0 && skipEmptyKey) { // skip empty keys
continue
}
cmp := bytes.Compare(key, m.lowK)
if cmp < 0 || len(m.lowIdxs) == 0 {
// reached a new low
m.lowK = key
m.lowIdxs = m.lowIdxs[:0]
m.lowIdxs = append(m.lowIdxs, i)
} else if cmp == 0 {
m.lowIdxs = append(m.lowIdxs, i)
}
}
}
// Current returns the enumerator's current key, iterator-index, and
// value. If the enumerator is not pointing at a valid value (because
// Next returned an error previously), Current will return nil,0,0.
func (m *enumerator) Current() ([]byte, int, uint64) {
var i int
var v uint64
if m.lowCurr < len(m.lowIdxs) {
i = m.lowIdxs[m.lowCurr]
v = m.currVs[i]
}
return m.lowK, i, v
}
// GetLowIdxsAndValues will return all of the iterator indices
// which point to the current key, and their corresponding
// values. This can be used by advanced caller which may need
// to peek into these other sets of data before processing.
func (m *enumerator) GetLowIdxsAndValues() ([]int, []uint64) {
values := make([]uint64, 0, len(m.lowIdxs))
for _, idx := range m.lowIdxs {
values = append(values, m.currVs[idx])
}
return m.lowIdxs, values
}
// Next advances the enumerator to the next key/iterator/value result,
// else vellum.ErrIteratorDone is returned.
func (m *enumerator) Next() error {
m.lowCurr += 1
if m.lowCurr >= len(m.lowIdxs) {
// move all the current low iterators forwards
for _, vi := range m.lowIdxs {
err := m.itrs[vi].Next()
if err != nil && err != vellum.ErrIteratorDone {
return err
}
m.currKs[vi], m.currVs[vi] = m.itrs[vi].Current()
}
// can skip any empty keys encountered at this point
m.updateMatches(true)
}
if m.lowK == nil && len(m.lowIdxs) == 0 {
return vellum.ErrIteratorDone
}
return nil
}
// Close all the underlying Iterators. The first error, if any, will
// be returned.
func (m *enumerator) Close() error {
var rv error
for _, itr := range m.itrs {
err := itr.Close()
if rv == nil {
rv = err
}
}
return rv
}

@ -0,0 +1,12 @@
module github.com/blevesearch/zap/v12
go 1.12
require (
github.com/RoaringBitmap/roaring v0.4.21
github.com/blevesearch/bleve v1.0.7
github.com/blevesearch/mmap-go v1.0.2
github.com/couchbase/vellum v1.0.1
github.com/golang/snappy v0.0.1
github.com/spf13/cobra v0.0.5
)

@ -0,0 +1,111 @@
// Copyright (c) 2019 Couchbase, Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
package zap
import (
"encoding/binary"
"fmt"
"github.com/blevesearch/bleve/index/scorch/segment"
)
type chunkedIntDecoder struct {
startOffset uint64
dataStartOffset uint64
chunkOffsets []uint64
curChunkBytes []byte
data []byte
r *segment.MemUvarintReader
}
func newChunkedIntDecoder(buf []byte, offset uint64) *chunkedIntDecoder {
rv := &chunkedIntDecoder{startOffset: offset, data: buf}
var n, numChunks uint64
var read int
if offset == termNotEncoded {
numChunks = 0
} else {
numChunks, read = binary.Uvarint(buf[offset+n : offset+n+binary.MaxVarintLen64])
}
n += uint64(read)
if cap(rv.chunkOffsets) >= int(numChunks) {
rv.chunkOffsets = rv.chunkOffsets[:int(numChunks)]
} else {
rv.chunkOffsets = make([]uint64, int(numChunks))
}
for i := 0; i < int(numChunks); i++ {
rv.chunkOffsets[i], read = binary.Uvarint(buf[offset+n : offset+n+binary.MaxVarintLen64])
n += uint64(read)
}
rv.dataStartOffset = offset + n
return rv
}
func (d *chunkedIntDecoder) loadChunk(chunk int) error {
if d.startOffset == termNotEncoded {
d.r = segment.NewMemUvarintReader([]byte(nil))
return nil
}
if chunk >= len(d.chunkOffsets) {
return fmt.Errorf("tried to load freq chunk that doesn't exist %d/(%d)",
chunk, len(d.chunkOffsets))
}
end, start := d.dataStartOffset, d.dataStartOffset
s, e := readChunkBoundary(chunk, d.chunkOffsets)
start += s
end += e
d.curChunkBytes = d.data[start:end]
if d.r == nil {
d.r = segment.NewMemUvarintReader(d.curChunkBytes)
} else {
d.r.Reset(d.curChunkBytes)
}
return nil
}
func (d *chunkedIntDecoder) reset() {
d.startOffset = 0
d.dataStartOffset = 0
d.chunkOffsets = d.chunkOffsets[:0]
d.curChunkBytes = d.curChunkBytes[:0]
d.data = d.data[:0]
if d.r != nil {
d.r.Reset([]byte(nil))
}
}
func (d *chunkedIntDecoder) isNil() bool {
return d.curChunkBytes == nil
}
func (d *chunkedIntDecoder) readUvarint() (uint64, error) {
return d.r.ReadUvarint()
}
func (d *chunkedIntDecoder) SkipUvarint() {
d.r.SkipUvarint()
}
func (d *chunkedIntDecoder) SkipBytes(count int) {
d.r.SkipBytes(count)
}
func (d *chunkedIntDecoder) Len() int {
return d.r.Len()
}

@ -0,0 +1,203 @@
// Copyright (c) 2017 Couchbase, Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
package zap
import (
"bytes"
"encoding/binary"
"io"
"math"
)
const termNotEncoded = math.MaxUint64
type chunkedIntCoder struct {
final []byte
chunkSize uint64
chunkBuf bytes.Buffer
chunkLens []uint64
currChunk uint64
buf []byte
}
// newChunkedIntCoder returns a new chunk int coder which packs data into
// chunks based on the provided chunkSize and supports up to the specified
// maxDocNum
func newChunkedIntCoder(chunkSize uint64, maxDocNum uint64) *chunkedIntCoder {
total := maxDocNum/chunkSize + 1
rv := &chunkedIntCoder{
chunkSize: chunkSize,
chunkLens: make([]uint64, total),
final: make([]byte, 0, 64),
}
return rv
}
// Reset lets you reuse this chunked int coder. buffers are reset and reused
// from previous use. you cannot change the chunk size or max doc num.
func (c *chunkedIntCoder) Reset() {
c.final = c.final[:0]
c.chunkBuf.Reset()
c.currChunk = 0
for i := range c.chunkLens {
c.chunkLens[i] = 0
}
}
// SetChunkSize changes the chunk size. It is only valid to do so
// with a new chunkedIntCoder, or immediately after calling Reset()
func (c *chunkedIntCoder) SetChunkSize(chunkSize uint64, maxDocNum uint64) {
total := int(maxDocNum/chunkSize + 1)
c.chunkSize = chunkSize
if cap(c.chunkLens) < total {
c.chunkLens = make([]uint64, total)
} else {
c.chunkLens = c.chunkLens[:total]
}
}
// Add encodes the provided integers into the correct chunk for the provided
// doc num. You MUST call Add() with increasing docNums.
func (c *chunkedIntCoder) Add(docNum uint64, vals ...uint64) error {
chunk := docNum / c.chunkSize
if chunk != c.currChunk {
// starting a new chunk
c.Close()
c.chunkBuf.Reset()
c.currChunk = chunk
}
if len(c.buf) < binary.MaxVarintLen64 {
c.buf = make([]byte, binary.MaxVarintLen64)
}
for _, val := range vals {
wb := binary.PutUvarint(c.buf, val)
_, err := c.chunkBuf.Write(c.buf[:wb])
if err != nil {
return err
}
}
return nil
}
func (c *chunkedIntCoder) AddBytes(docNum uint64, buf []byte) error {
chunk := docNum / c.chunkSize
if chunk != c.currChunk {
// starting a new chunk
c.Close()
c.chunkBuf.Reset()
c.currChunk = chunk
}
_, err := c.chunkBuf.Write(buf)
return err
}
// Close indicates you are done calling Add() this allows the final chunk
// to be encoded.
func (c *chunkedIntCoder) Close() {
encodingBytes := c.chunkBuf.Bytes()
c.chunkLens[c.currChunk] = uint64(len(encodingBytes))
c.final = append(c.final, encodingBytes...)
c.currChunk = uint64(cap(c.chunkLens)) // sentinel to detect double close
}
// Write commits all the encoded chunked integers to the provided writer.
func (c *chunkedIntCoder) Write(w io.Writer) (int, error) {
bufNeeded := binary.MaxVarintLen64 * (1 + len(c.chunkLens))
if len(c.buf) < bufNeeded {
c.buf = make([]byte, bufNeeded)
}
buf := c.buf
// convert the chunk lengths into chunk offsets
chunkOffsets := modifyLengthsToEndOffsets(c.chunkLens)
// write out the number of chunks & each chunk offsets
n := binary.PutUvarint(buf, uint64(len(chunkOffsets)))
for _, chunkOffset := range chunkOffsets {
n += binary.PutUvarint(buf[n:], chunkOffset)
}
tw, err := w.Write(buf[:n])
if err != nil {
return tw, err
}
// write out the data
nw, err := w.Write(c.final)
tw += nw
if err != nil {
return tw, err
}
return tw, nil
}
// writeAt commits all the encoded chunked integers to the provided writer
// and returns the starting offset, total bytes written and an error
func (c *chunkedIntCoder) writeAt(w io.Writer) (uint64, int, error) {
startOffset := uint64(termNotEncoded)
if len(c.final) <= 0 {
return startOffset, 0, nil
}
if chw := w.(*CountHashWriter); chw != nil {
startOffset = uint64(chw.Count())
}
tw, err := c.Write(w)
return startOffset, tw, err
}
func (c *chunkedIntCoder) FinalSize() int {
return len(c.final)
}
// modifyLengthsToEndOffsets converts the chunk length array
// to a chunk offset array. The readChunkBoundary
// will figure out the start and end of every chunk from
// these offsets. Starting offset of i'th index is stored
// in i-1'th position except for 0'th index and ending offset
// is stored at i'th index position.
// For 0'th element, starting position is always zero.
// eg:
// Lens -> 5 5 5 5 => 5 10 15 20
// Lens -> 0 5 0 5 => 0 5 5 10
// Lens -> 0 0 0 5 => 0 0 0 5
// Lens -> 5 0 0 0 => 5 5 5 5
// Lens -> 0 5 0 0 => 0 5 5 5
// Lens -> 0 0 5 0 => 0 0 5 5
func modifyLengthsToEndOffsets(lengths []uint64) []uint64 {
var runningOffset uint64
var index, i int
for i = 1; i <= len(lengths); i++ {
runningOffset += lengths[i-1]
lengths[index] = runningOffset
index++
}
return lengths
}
func readChunkBoundary(chunk int, offsets []uint64) (uint64, uint64) {
var start uint64
if chunk > 0 {
start = offsets[chunk-1]
}
return start, offsets[chunk]
}

@ -0,0 +1,847 @@
// Copyright (c) 2017 Couchbase, Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
package zap
import (
"bufio"
"bytes"
"encoding/binary"
"fmt"
"math"
"os"
"sort"
"github.com/RoaringBitmap/roaring"
seg "github.com/blevesearch/bleve/index/scorch/segment"
"github.com/couchbase/vellum"
"github.com/golang/snappy"
)
var DefaultFileMergerBufferSize = 1024 * 1024
const docDropped = math.MaxUint64 // sentinel docNum to represent a deleted doc
// Merge takes a slice of segments and bit masks describing which
// documents may be dropped, and creates a new segment containing the
// remaining data. This new segment is built at the specified path.
func (*ZapPlugin) Merge(segments []seg.Segment, drops []*roaring.Bitmap, path string,
closeCh chan struct{}, s seg.StatsReporter) (
[][]uint64, uint64, error) {
segmentBases := make([]*SegmentBase, len(segments))
for segmenti, segment := range segments {
switch segmentx := segment.(type) {
case *Segment:
segmentBases[segmenti] = &segmentx.SegmentBase
case *SegmentBase:
segmentBases[segmenti] = segmentx
default:
panic(fmt.Sprintf("oops, unexpected segment type: %T", segment))
}
}
return mergeSegmentBases(segmentBases, drops, path, DefaultChunkMode, closeCh, s)
}
func mergeSegmentBases(segmentBases []*SegmentBase, drops []*roaring.Bitmap, path string,
chunkMode uint32, closeCh chan struct{}, s seg.StatsReporter) (
[][]uint64, uint64, error) {
flag := os.O_RDWR | os.O_CREATE
f, err := os.OpenFile(path, flag, 0600)
if err != nil {
return nil, 0, err
}
cleanup := func() {
_ = f.Close()
_ = os.Remove(path)
}
// buffer the output
br := bufio.NewWriterSize(f, DefaultFileMergerBufferSize)
// wrap it for counting (tracking offsets)
cr := NewCountHashWriterWithStatsReporter(br, s)
newDocNums, numDocs, storedIndexOffset, fieldsIndexOffset, docValueOffset, _, _, _, err :=
MergeToWriter(segmentBases, drops, chunkMode, cr, closeCh)
if err != nil {
cleanup()
return nil, 0, err
}
err = persistFooter(numDocs, storedIndexOffset, fieldsIndexOffset,
docValueOffset, chunkMode, cr.Sum32(), cr)
if err != nil {
cleanup()
return nil, 0, err
}
err = br.Flush()
if err != nil {
cleanup()
return nil, 0, err
}
err = f.Sync()
if err != nil {
cleanup()
return nil, 0, err
}
err = f.Close()
if err != nil {
cleanup()
return nil, 0, err
}
return newDocNums, uint64(cr.Count()), nil
}
func MergeToWriter(segments []*SegmentBase, drops []*roaring.Bitmap,
chunkMode uint32, cr *CountHashWriter, closeCh chan struct{}) (
newDocNums [][]uint64,
numDocs, storedIndexOffset, fieldsIndexOffset, docValueOffset uint64,
dictLocs []uint64, fieldsInv []string, fieldsMap map[string]uint16,
err error) {
docValueOffset = uint64(fieldNotUninverted)
var fieldsSame bool
fieldsSame, fieldsInv = mergeFields(segments)
fieldsMap = mapFields(fieldsInv)
numDocs = computeNewDocCount(segments, drops)
if isClosed(closeCh) {
return nil, 0, 0, 0, 0, nil, nil, nil, seg.ErrClosed
}
if numDocs > 0 {
storedIndexOffset, newDocNums, err = mergeStoredAndRemap(segments, drops,
fieldsMap, fieldsInv, fieldsSame, numDocs, cr, closeCh)
if err != nil {
return nil, 0, 0, 0, 0, nil, nil, nil, err
}
dictLocs, docValueOffset, err = persistMergedRest(segments, drops,
fieldsInv, fieldsMap, fieldsSame,
newDocNums, numDocs, chunkMode, cr, closeCh)
if err != nil {
return nil, 0, 0, 0, 0, nil, nil, nil, err
}
} else {
dictLocs = make([]uint64, len(fieldsInv))
}
fieldsIndexOffset, err = persistFields(fieldsInv, cr, dictLocs)
if err != nil {
return nil, 0, 0, 0, 0, nil, nil, nil, err
}
return newDocNums, numDocs, storedIndexOffset, fieldsIndexOffset, docValueOffset, dictLocs, fieldsInv, fieldsMap, nil
}
// mapFields takes the fieldsInv list and returns a map of fieldName
// to fieldID+1
func mapFields(fields []string) map[string]uint16 {
rv := make(map[string]uint16, len(fields))
for i, fieldName := range fields {
rv[fieldName] = uint16(i) + 1
}
return rv
}
// computeNewDocCount determines how many documents will be in the newly
// merged segment when obsoleted docs are dropped
func computeNewDocCount(segments []*SegmentBase, drops []*roaring.Bitmap) uint64 {
var newDocCount uint64
for segI, segment := range segments {
newDocCount += segment.numDocs
if drops[segI] != nil {
newDocCount -= drops[segI].GetCardinality()
}
}
return newDocCount
}
func persistMergedRest(segments []*SegmentBase, dropsIn []*roaring.Bitmap,
fieldsInv []string, fieldsMap map[string]uint16, fieldsSame bool,
newDocNumsIn [][]uint64, newSegDocCount uint64, chunkMode uint32,
w *CountHashWriter, closeCh chan struct{}) ([]uint64, uint64, error) {
var bufMaxVarintLen64 []byte = make([]byte, binary.MaxVarintLen64)
var bufLoc []uint64
var postings *PostingsList
var postItr *PostingsIterator
rv := make([]uint64, len(fieldsInv))
fieldDvLocsStart := make([]uint64, len(fieldsInv))
fieldDvLocsEnd := make([]uint64, len(fieldsInv))
// these int coders are initialized with chunk size 1024
// however this will be reset to the correct chunk size
// while processing each individual field-term section
tfEncoder := newChunkedIntCoder(1024, newSegDocCount-1)
locEncoder := newChunkedIntCoder(1024, newSegDocCount-1)
var vellumBuf bytes.Buffer
newVellum, err := vellum.New(&vellumBuf, nil)
if err != nil {
return nil, 0, err
}
newRoaring := roaring.NewBitmap()
// for each field
for fieldID, fieldName := range fieldsInv {
// collect FST iterators from all active segments for this field
var newDocNums [][]uint64
var drops []*roaring.Bitmap
var dicts []*Dictionary
var itrs []vellum.Iterator
var segmentsInFocus []*SegmentBase
for segmentI, segment := range segments {
// check for the closure in meantime
if isClosed(closeCh) {
return nil, 0, seg.ErrClosed
}
dict, err2 := segment.dictionary(fieldName)
if err2 != nil {
return nil, 0, err2
}
if dict != nil && dict.fst != nil {
itr, err2 := dict.fst.Iterator(nil, nil)
if err2 != nil && err2 != vellum.ErrIteratorDone {
return nil, 0, err2
}
if itr != nil {
newDocNums = append(newDocNums, newDocNumsIn[segmentI])
if dropsIn[segmentI] != nil && !dropsIn[segmentI].IsEmpty() {
drops = append(drops, dropsIn[segmentI])
} else {
drops = append(drops, nil)
}
dicts = append(dicts, dict)
itrs = append(itrs, itr)
segmentsInFocus = append(segmentsInFocus, segment)
}
}
}
var prevTerm []byte
newRoaring.Clear()
var lastDocNum, lastFreq, lastNorm uint64
// determines whether to use "1-hit" encoding optimization
// when a term appears in only 1 doc, with no loc info,
// has freq of 1, and the docNum fits into 31-bits
use1HitEncoding := func(termCardinality uint64) (bool, uint64, uint64) {
if termCardinality == uint64(1) && locEncoder.FinalSize() <= 0 {
docNum := uint64(newRoaring.Minimum())
if under32Bits(docNum) && docNum == lastDocNum && lastFreq == 1 {
return true, docNum, lastNorm
}
}
return false, 0, 0
}
finishTerm := func(term []byte) error {
tfEncoder.Close()
locEncoder.Close()
postingsOffset, err := writePostings(newRoaring,
tfEncoder, locEncoder, use1HitEncoding, w, bufMaxVarintLen64)
if err != nil {
return err
}
if postingsOffset > 0 {
err = newVellum.Insert(term, postingsOffset)
if err != nil {
return err
}
}
newRoaring.Clear()
tfEncoder.Reset()
locEncoder.Reset()
lastDocNum = 0
lastFreq = 0
lastNorm = 0
return nil
}
enumerator, err := newEnumerator(itrs)
for err == nil {
term, itrI, postingsOffset := enumerator.Current()
if !bytes.Equal(prevTerm, term) {
// check for the closure in meantime
if isClosed(closeCh) {
return nil, 0, seg.ErrClosed
}
// if the term changed, write out the info collected
// for the previous term
err = finishTerm(prevTerm)
if err != nil {
return nil, 0, err
}
}
if !bytes.Equal(prevTerm, term) || prevTerm == nil {
// compute cardinality of field-term in new seg
var newCard uint64
lowItrIdxs, lowItrVals := enumerator.GetLowIdxsAndValues()
for i, idx := range lowItrIdxs {
pl, err := dicts[idx].postingsListFromOffset(lowItrVals[i], drops[idx], nil)
if err != nil {
return nil, 0, err
}
newCard += pl.Count()
}
// compute correct chunk size with this
chunkSize, err := getChunkSize(chunkMode, newCard, newSegDocCount)
if err != nil {
return nil, 0, err
}
// update encoders chunk
tfEncoder.SetChunkSize(chunkSize, newSegDocCount-1)
locEncoder.SetChunkSize(chunkSize, newSegDocCount-1)
}
postings, err = dicts[itrI].postingsListFromOffset(
postingsOffset, drops[itrI], postings)
if err != nil {
return nil, 0, err
}
postItr = postings.iterator(true, true, true, postItr)
// can no longer optimize by copying, since chunk factor could have changed
lastDocNum, lastFreq, lastNorm, bufLoc, err = mergeTermFreqNormLocs(
fieldsMap, term, postItr, newDocNums[itrI], newRoaring,
tfEncoder, locEncoder, bufLoc)
if err != nil {
return nil, 0, err
}
prevTerm = prevTerm[:0] // copy to prevTerm in case Next() reuses term mem
prevTerm = append(prevTerm, term...)
err = enumerator.Next()
}
if err != vellum.ErrIteratorDone {
return nil, 0, err
}
err = finishTerm(prevTerm)
if err != nil {
return nil, 0, err
}
dictOffset := uint64(w.Count())
err = newVellum.Close()
if err != nil {
return nil, 0, err
}
vellumData := vellumBuf.Bytes()
// write out the length of the vellum data
n := binary.PutUvarint(bufMaxVarintLen64, uint64(len(vellumData)))
_, err = w.Write(bufMaxVarintLen64[:n])
if err != nil {
return nil, 0, err
}
// write this vellum to disk
_, err = w.Write(vellumData)
if err != nil {
return nil, 0, err
}
rv[fieldID] = dictOffset
// get the field doc value offset (start)
fieldDvLocsStart[fieldID] = uint64(w.Count())
// update the field doc values
// NOTE: doc values continue to use legacy chunk mode
chunkSize, err := getChunkSize(LegacyChunkMode, 0, 0)
if err != nil {
return nil, 0, err
}
fdvEncoder := newChunkedContentCoder(chunkSize, newSegDocCount-1, w, true)
fdvReadersAvailable := false
var dvIterClone *docValueReader
for segmentI, segment := range segmentsInFocus {
// check for the closure in meantime
if isClosed(closeCh) {
return nil, 0, seg.ErrClosed
}
fieldIDPlus1 := uint16(segment.fieldsMap[fieldName])
if dvIter, exists := segment.fieldDvReaders[fieldIDPlus1-1]; exists &&
dvIter != nil {
fdvReadersAvailable = true
dvIterClone = dvIter.cloneInto(dvIterClone)
err = dvIterClone.iterateAllDocValues(segment, func(docNum uint64, terms []byte) error {
if newDocNums[segmentI][docNum] == docDropped {
return nil
}
err := fdvEncoder.Add(newDocNums[segmentI][docNum], terms)
if err != nil {
return err
}
return nil
})
if err != nil {
return nil, 0, err
}
}
}
if fdvReadersAvailable {
err = fdvEncoder.Close()
if err != nil {
return nil, 0, err
}
// persist the doc value details for this field
_, err = fdvEncoder.Write()
if err != nil {
return nil, 0, err
}
// get the field doc value offset (end)
fieldDvLocsEnd[fieldID] = uint64(w.Count())
} else {
fieldDvLocsStart[fieldID] = fieldNotUninverted
fieldDvLocsEnd[fieldID] = fieldNotUninverted
}
// reset vellum buffer and vellum builder
vellumBuf.Reset()
err = newVellum.Reset(&vellumBuf)
if err != nil {
return nil, 0, err
}
}
fieldDvLocsOffset := uint64(w.Count())
buf := bufMaxVarintLen64
for i := 0; i < len(fieldDvLocsStart); i++ {
n := binary.PutUvarint(buf, fieldDvLocsStart[i])
_, err := w.Write(buf[:n])
if err != nil {
return nil, 0, err
}
n = binary.PutUvarint(buf, fieldDvLocsEnd[i])
_, err = w.Write(buf[:n])
if err != nil {
return nil, 0, err
}
}
return rv, fieldDvLocsOffset, nil
}
func mergeTermFreqNormLocs(fieldsMap map[string]uint16, term []byte, postItr *PostingsIterator,
newDocNums []uint64, newRoaring *roaring.Bitmap,
tfEncoder *chunkedIntCoder, locEncoder *chunkedIntCoder, bufLoc []uint64) (
lastDocNum uint64, lastFreq uint64, lastNorm uint64, bufLocOut []uint64, err error) {
next, err := postItr.Next()
for next != nil && err == nil {
hitNewDocNum := newDocNums[next.Number()]
if hitNewDocNum == docDropped {
return 0, 0, 0, nil, fmt.Errorf("see hit with dropped docNum")
}
newRoaring.Add(uint32(hitNewDocNum))
nextFreq := next.Frequency()
nextNorm := uint64(math.Float32bits(float32(next.Norm())))
locs := next.Locations()
err = tfEncoder.Add(hitNewDocNum,
encodeFreqHasLocs(nextFreq, len(locs) > 0), nextNorm)
if err != nil {
return 0, 0, 0, nil, err
}
if len(locs) > 0 {
numBytesLocs := 0
for _, loc := range locs {
ap := loc.ArrayPositions()
numBytesLocs += totalUvarintBytes(uint64(fieldsMap[loc.Field()]-1),
loc.Pos(), loc.Start(), loc.End(), uint64(len(ap)), ap)
}
err = locEncoder.Add(hitNewDocNum, uint64(numBytesLocs))
if err != nil {
return 0, 0, 0, nil, err
}
for _, loc := range locs {
ap := loc.ArrayPositions()
if cap(bufLoc) < 5+len(ap) {
bufLoc = make([]uint64, 0, 5+len(ap))
}
args := bufLoc[0:5]
args[0] = uint64(fieldsMap[loc.Field()] - 1)
args[1] = loc.Pos()
args[2] = loc.Start()
args[3] = loc.End()
args[4] = uint64(len(ap))
args = append(args, ap...)
err = locEncoder.Add(hitNewDocNum, args...)
if err != nil {
return 0, 0, 0, nil, err
}
}
}
lastDocNum = hitNewDocNum
lastFreq = nextFreq
lastNorm = nextNorm
next, err = postItr.Next()
}
return lastDocNum, lastFreq, lastNorm, bufLoc, err
}
func writePostings(postings *roaring.Bitmap, tfEncoder, locEncoder *chunkedIntCoder,
use1HitEncoding func(uint64) (bool, uint64, uint64),
w *CountHashWriter, bufMaxVarintLen64 []byte) (
offset uint64, err error) {
termCardinality := postings.GetCardinality()
if termCardinality <= 0 {
return 0, nil
}
if use1HitEncoding != nil {
encodeAs1Hit, docNum1Hit, normBits1Hit := use1HitEncoding(termCardinality)
if encodeAs1Hit {
return FSTValEncode1Hit(docNum1Hit, normBits1Hit), nil
}
}
var tfOffset uint64
tfOffset, _, err = tfEncoder.writeAt(w)
if err != nil {
return 0, err
}
var locOffset uint64
locOffset, _, err = locEncoder.writeAt(w)
if err != nil {
return 0, err
}
postingsOffset := uint64(w.Count())
n := binary.PutUvarint(bufMaxVarintLen64, tfOffset)
_, err = w.Write(bufMaxVarintLen64[:n])
if err != nil {
return 0, err
}
n = binary.PutUvarint(bufMaxVarintLen64, locOffset)
_, err = w.Write(bufMaxVarintLen64[:n])
if err != nil {
return 0, err
}
_, err = writeRoaringWithLen(postings, w, bufMaxVarintLen64)
if err != nil {
return 0, err
}
return postingsOffset, nil
}
type varintEncoder func(uint64) (int, error)
func mergeStoredAndRemap(segments []*SegmentBase, drops []*roaring.Bitmap,
fieldsMap map[string]uint16, fieldsInv []string, fieldsSame bool, newSegDocCount uint64,
w *CountHashWriter, closeCh chan struct{}) (uint64, [][]uint64, error) {
var rv [][]uint64 // The remapped or newDocNums for each segment.
var newDocNum uint64
var curr int
var data, compressed []byte
var metaBuf bytes.Buffer
varBuf := make([]byte, binary.MaxVarintLen64)
metaEncode := func(val uint64) (int, error) {
wb := binary.PutUvarint(varBuf, val)
return metaBuf.Write(varBuf[:wb])
}
vals := make([][][]byte, len(fieldsInv))
typs := make([][]byte, len(fieldsInv))
poss := make([][][]uint64, len(fieldsInv))
var posBuf []uint64
docNumOffsets := make([]uint64, newSegDocCount)
vdc := visitDocumentCtxPool.Get().(*visitDocumentCtx)
defer visitDocumentCtxPool.Put(vdc)
// for each segment
for segI, segment := range segments {
// check for the closure in meantime
if isClosed(closeCh) {
return 0, nil, seg.ErrClosed
}
segNewDocNums := make([]uint64, segment.numDocs)
dropsI := drops[segI]
// optimize when the field mapping is the same across all
// segments and there are no deletions, via byte-copying
// of stored docs bytes directly to the writer
if fieldsSame && (dropsI == nil || dropsI.GetCardinality() == 0) {
err := segment.copyStoredDocs(newDocNum, docNumOffsets, w)
if err != nil {
return 0, nil, err
}
for i := uint64(0); i < segment.numDocs; i++ {
segNewDocNums[i] = newDocNum
newDocNum++
}
rv = append(rv, segNewDocNums)
continue
}
// for each doc num
for docNum := uint64(0); docNum < segment.numDocs; docNum++ {
// TODO: roaring's API limits docNums to 32-bits?
if dropsI != nil && dropsI.Contains(uint32(docNum)) {
segNewDocNums[docNum] = docDropped
continue
}
segNewDocNums[docNum] = newDocNum
curr = 0
metaBuf.Reset()
data = data[:0]
posTemp := posBuf
// collect all the data
for i := 0; i < len(fieldsInv); i++ {
vals[i] = vals[i][:0]
typs[i] = typs[i][:0]
poss[i] = poss[i][:0]
}
err := segment.visitDocument(vdc, docNum, func(field string, typ byte, value []byte, pos []uint64) bool {
fieldID := int(fieldsMap[field]) - 1
vals[fieldID] = append(vals[fieldID], value)
typs[fieldID] = append(typs[fieldID], typ)
// copy array positions to preserve them beyond the scope of this callback
var curPos []uint64
if len(pos) > 0 {
if cap(posTemp) < len(pos) {
posBuf = make([]uint64, len(pos)*len(fieldsInv))
posTemp = posBuf
}
curPos = posTemp[0:len(pos)]
copy(curPos, pos)
posTemp = posTemp[len(pos):]
}
poss[fieldID] = append(poss[fieldID], curPos)
return true
})
if err != nil {
return 0, nil, err
}
// _id field special case optimizes ExternalID() lookups
idFieldVal := vals[uint16(0)][0]
_, err = metaEncode(uint64(len(idFieldVal)))
if err != nil {
return 0, nil, err
}
// now walk the non-"_id" fields in order
for fieldID := 1; fieldID < len(fieldsInv); fieldID++ {
storedFieldValues := vals[fieldID]
stf := typs[fieldID]
spf := poss[fieldID]
var err2 error
curr, data, err2 = persistStoredFieldValues(fieldID,
storedFieldValues, stf, spf, curr, metaEncode, data)
if err2 != nil {
return 0, nil, err2
}
}
metaBytes := metaBuf.Bytes()
compressed = snappy.Encode(compressed[:cap(compressed)], data)
// record where we're about to start writing
docNumOffsets[newDocNum] = uint64(w.Count())
// write out the meta len and compressed data len
_, err = writeUvarints(w,
uint64(len(metaBytes)),
uint64(len(idFieldVal)+len(compressed)))
if err != nil {
return 0, nil, err
}
// now write the meta
_, err = w.Write(metaBytes)
if err != nil {
return 0, nil, err
}
// now write the _id field val (counted as part of the 'compressed' data)
_, err = w.Write(idFieldVal)
if err != nil {
return 0, nil, err
}
// now write the compressed data
_, err = w.Write(compressed)
if err != nil {
return 0, nil, err
}
newDocNum++
}
rv = append(rv, segNewDocNums)
}
// return value is the start of the stored index
storedIndexOffset := uint64(w.Count())
// now write out the stored doc index
for _, docNumOffset := range docNumOffsets {
err := binary.Write(w, binary.BigEndian, docNumOffset)
if err != nil {
return 0, nil, err
}
}
return storedIndexOffset, rv, nil
}
// copyStoredDocs writes out a segment's stored doc info, optimized by
// using a single Write() call for the entire set of bytes. The
// newDocNumOffsets is filled with the new offsets for each doc.
func (s *SegmentBase) copyStoredDocs(newDocNum uint64, newDocNumOffsets []uint64,
w *CountHashWriter) error {
if s.numDocs <= 0 {
return nil
}
indexOffset0, storedOffset0, _, _, _ :=
s.getDocStoredOffsets(0) // the segment's first doc
indexOffsetN, storedOffsetN, readN, metaLenN, dataLenN :=
s.getDocStoredOffsets(s.numDocs - 1) // the segment's last doc
storedOffset0New := uint64(w.Count())
storedBytes := s.mem[storedOffset0 : storedOffsetN+readN+metaLenN+dataLenN]
_, err := w.Write(storedBytes)
if err != nil {
return err
}
// remap the storedOffset's for the docs into new offsets relative
// to storedOffset0New, filling the given docNumOffsetsOut array
for indexOffset := indexOffset0; indexOffset <= indexOffsetN; indexOffset += 8 {
storedOffset := binary.BigEndian.Uint64(s.mem[indexOffset : indexOffset+8])
storedOffsetNew := storedOffset - storedOffset0 + storedOffset0New
newDocNumOffsets[newDocNum] = storedOffsetNew
newDocNum += 1
}
return nil
}
// mergeFields builds a unified list of fields used across all the
// input segments, and computes whether the fields are the same across
// segments (which depends on fields to be sorted in the same way
// across segments)
func mergeFields(segments []*SegmentBase) (bool, []string) {
fieldsSame := true
var segment0Fields []string
if len(segments) > 0 {
segment0Fields = segments[0].Fields()
}
fieldsExist := map[string]struct{}{}
for _, segment := range segments {
fields := segment.Fields()
for fieldi, field := range fields {
fieldsExist[field] = struct{}{}
if len(segment0Fields) != len(fields) || segment0Fields[fieldi] != field {
fieldsSame = false
}
}
}
rv := make([]string, 0, len(fieldsExist))
// ensure _id stays first
rv = append(rv, "_id")
for k := range fieldsExist {
if k != "_id" {
rv = append(rv, k)
}
}
sort.Strings(rv[1:]) // leave _id as first
return fieldsSame, rv
}
func isClosed(closeCh chan struct{}) bool {
select {
case <-closeCh:
return true
default:
return false
}
}

@ -0,0 +1,860 @@
// Copyright (c) 2018 Couchbase, Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
package zap
import (
"bytes"
"encoding/binary"
"math"
"sort"
"sync"
"github.com/RoaringBitmap/roaring"
"github.com/blevesearch/bleve/analysis"
"github.com/blevesearch/bleve/document"
"github.com/blevesearch/bleve/index"
"github.com/blevesearch/bleve/index/scorch/segment"
"github.com/couchbase/vellum"
"github.com/golang/snappy"
)
var NewSegmentBufferNumResultsBump int = 100
var NewSegmentBufferNumResultsFactor float64 = 1.0
var NewSegmentBufferAvgBytesPerDocFactor float64 = 1.0
// ValidateDocFields can be set by applications to perform additional checks
// on fields in a document being added to a new segment, by default it does
// nothing.
// This API is experimental and may be removed at any time.
var ValidateDocFields = func(field document.Field) error {
return nil
}
// AnalysisResultsToSegmentBase produces an in-memory zap-encoded
// SegmentBase from analysis results
func (z *ZapPlugin) New(results []*index.AnalysisResult) (
segment.Segment, uint64, error) {
return z.newWithChunkMode(results, DefaultChunkMode)
}
func (*ZapPlugin) newWithChunkMode(results []*index.AnalysisResult,
chunkMode uint32) (segment.Segment, uint64, error) {
s := interimPool.Get().(*interim)
var br bytes.Buffer
if s.lastNumDocs > 0 {
// use previous results to initialize the buf with an estimate
// size, but note that the interim instance comes from a
// global interimPool, so multiple scorch instances indexing
// different docs can lead to low quality estimates
estimateAvgBytesPerDoc := int(float64(s.lastOutSize/s.lastNumDocs) *
NewSegmentBufferNumResultsFactor)
estimateNumResults := int(float64(len(results)+NewSegmentBufferNumResultsBump) *
NewSegmentBufferAvgBytesPerDocFactor)
br.Grow(estimateAvgBytesPerDoc * estimateNumResults)
}
s.results = results
s.chunkMode = chunkMode
s.w = NewCountHashWriter(&br)
storedIndexOffset, fieldsIndexOffset, fdvIndexOffset, dictOffsets,
err := s.convert()
if err != nil {
return nil, uint64(0), err
}
sb, err := InitSegmentBase(br.Bytes(), s.w.Sum32(), chunkMode,
s.FieldsMap, s.FieldsInv, uint64(len(results)),
storedIndexOffset, fieldsIndexOffset, fdvIndexOffset, dictOffsets)
if err == nil && s.reset() == nil {
s.lastNumDocs = len(results)
s.lastOutSize = len(br.Bytes())
interimPool.Put(s)
}
return sb, uint64(len(br.Bytes())), err
}
var interimPool = sync.Pool{New: func() interface{} { return &interim{} }}
// interim holds temporary working data used while converting from
// analysis results to a zap-encoded segment
type interim struct {
results []*index.AnalysisResult
chunkMode uint32
w *CountHashWriter
// FieldsMap adds 1 to field id to avoid zero value issues
// name -> field id + 1
FieldsMap map[string]uint16
// FieldsInv is the inverse of FieldsMap
// field id -> name
FieldsInv []string
// Term dictionaries for each field
// field id -> term -> postings list id + 1
Dicts []map[string]uint64
// Terms for each field, where terms are sorted ascending
// field id -> []term
DictKeys [][]string
// Fields whose IncludeDocValues is true
// field id -> bool
IncludeDocValues []bool
// postings id -> bitmap of docNums
Postings []*roaring.Bitmap
// postings id -> freq/norm's, one for each docNum in postings
FreqNorms [][]interimFreqNorm
freqNormsBacking []interimFreqNorm
// postings id -> locs, one for each freq
Locs [][]interimLoc
locsBacking []interimLoc
numTermsPerPostingsList []int // key is postings list id
numLocsPerPostingsList []int // key is postings list id
builder *vellum.Builder
builderBuf bytes.Buffer
metaBuf bytes.Buffer
tmp0 []byte
tmp1 []byte
lastNumDocs int
lastOutSize int
}
func (s *interim) reset() (err error) {
s.results = nil
s.chunkMode = 0
s.w = nil
s.FieldsMap = nil
s.FieldsInv = nil
for i := range s.Dicts {
s.Dicts[i] = nil
}
s.Dicts = s.Dicts[:0]
for i := range s.DictKeys {
s.DictKeys[i] = s.DictKeys[i][:0]
}
s.DictKeys = s.DictKeys[:0]
for i := range s.IncludeDocValues {
s.IncludeDocValues[i] = false
}
s.IncludeDocValues = s.IncludeDocValues[:0]
for _, idn := range s.Postings {
idn.Clear()
}
s.Postings = s.Postings[:0]
s.FreqNorms = s.FreqNorms[:0]
for i := range s.freqNormsBacking {
s.freqNormsBacking[i] = interimFreqNorm{}
}
s.freqNormsBacking = s.freqNormsBacking[:0]
s.Locs = s.Locs[:0]
for i := range s.locsBacking {
s.locsBacking[i] = interimLoc{}
}
s.locsBacking = s.locsBacking[:0]
s.numTermsPerPostingsList = s.numTermsPerPostingsList[:0]
s.numLocsPerPostingsList = s.numLocsPerPostingsList[:0]
s.builderBuf.Reset()
if s.builder != nil {
err = s.builder.Reset(&s.builderBuf)
}
s.metaBuf.Reset()
s.tmp0 = s.tmp0[:0]
s.tmp1 = s.tmp1[:0]
s.lastNumDocs = 0
s.lastOutSize = 0
return err
}
func (s *interim) grabBuf(size int) []byte {
buf := s.tmp0
if cap(buf) < size {
buf = make([]byte, size)
s.tmp0 = buf
}
return buf[0:size]
}
type interimStoredField struct {
vals [][]byte
typs []byte
arrayposs [][]uint64 // array positions
}
type interimFreqNorm struct {
freq uint64
norm float32
numLocs int
}
type interimLoc struct {
fieldID uint16
pos uint64
start uint64
end uint64
arrayposs []uint64
}
func (s *interim) convert() (uint64, uint64, uint64, []uint64, error) {
s.FieldsMap = map[string]uint16{}
s.getOrDefineField("_id") // _id field is fieldID 0
for _, result := range s.results {
for _, field := range result.Document.CompositeFields {
s.getOrDefineField(field.Name())
}
for _, field := range result.Document.Fields {
s.getOrDefineField(field.Name())
}
}
sort.Strings(s.FieldsInv[1:]) // keep _id as first field
for fieldID, fieldName := range s.FieldsInv {
s.FieldsMap[fieldName] = uint16(fieldID + 1)
}
if cap(s.IncludeDocValues) >= len(s.FieldsInv) {
s.IncludeDocValues = s.IncludeDocValues[:len(s.FieldsInv)]
} else {
s.IncludeDocValues = make([]bool, len(s.FieldsInv))
}
s.prepareDicts()
for _, dict := range s.DictKeys {
sort.Strings(dict)
}
s.processDocuments()
storedIndexOffset, err := s.writeStoredFields()
if err != nil {
return 0, 0, 0, nil, err
}
var fdvIndexOffset uint64
var dictOffsets []uint64
if len(s.results) > 0 {
fdvIndexOffset, dictOffsets, err = s.writeDicts()
if err != nil {
return 0, 0, 0, nil, err
}
} else {
dictOffsets = make([]uint64, len(s.FieldsInv))
}
fieldsIndexOffset, err := persistFields(s.FieldsInv, s.w, dictOffsets)
if err != nil {
return 0, 0, 0, nil, err
}
return storedIndexOffset, fieldsIndexOffset, fdvIndexOffset, dictOffsets, nil
}
func (s *interim) getOrDefineField(fieldName string) int {
fieldIDPlus1, exists := s.FieldsMap[fieldName]
if !exists {
fieldIDPlus1 = uint16(len(s.FieldsInv) + 1)
s.FieldsMap[fieldName] = fieldIDPlus1
s.FieldsInv = append(s.FieldsInv, fieldName)
s.Dicts = append(s.Dicts, make(map[string]uint64))
n := len(s.DictKeys)
if n < cap(s.DictKeys) {
s.DictKeys = s.DictKeys[:n+1]
s.DictKeys[n] = s.DictKeys[n][:0]
} else {
s.DictKeys = append(s.DictKeys, []string(nil))
}
}
return int(fieldIDPlus1 - 1)
}
// fill Dicts and DictKeys from analysis results
func (s *interim) prepareDicts() {
var pidNext int
var totTFs int
var totLocs int
visitField := func(fieldID uint16, tfs analysis.TokenFrequencies) {
dict := s.Dicts[fieldID]
dictKeys := s.DictKeys[fieldID]
for term, tf := range tfs {
pidPlus1, exists := dict[term]
if !exists {
pidNext++
pidPlus1 = uint64(pidNext)
dict[term] = pidPlus1
dictKeys = append(dictKeys, term)
s.numTermsPerPostingsList = append(s.numTermsPerPostingsList, 0)
s.numLocsPerPostingsList = append(s.numLocsPerPostingsList, 0)
}
pid := pidPlus1 - 1
s.numTermsPerPostingsList[pid] += 1
s.numLocsPerPostingsList[pid] += len(tf.Locations)
totLocs += len(tf.Locations)
}
totTFs += len(tfs)
s.DictKeys[fieldID] = dictKeys
}
for _, result := range s.results {
// walk each composite field
for _, field := range result.Document.CompositeFields {
fieldID := uint16(s.getOrDefineField(field.Name()))
_, tf := field.Analyze()
visitField(fieldID, tf)
}
// walk each field
for i, field := range result.Document.Fields {
fieldID := uint16(s.getOrDefineField(field.Name()))
tf := result.Analyzed[i]
visitField(fieldID, tf)
}
}
numPostingsLists := pidNext
if cap(s.Postings) >= numPostingsLists {
s.Postings = s.Postings[:numPostingsLists]
} else {
postings := make([]*roaring.Bitmap, numPostingsLists)
copy(postings, s.Postings[:cap(s.Postings)])
for i := 0; i < numPostingsLists; i++ {
if postings[i] == nil {
postings[i] = roaring.New()
}
}
s.Postings = postings
}
if cap(s.FreqNorms) >= numPostingsLists {
s.FreqNorms = s.FreqNorms[:numPostingsLists]
} else {
s.FreqNorms = make([][]interimFreqNorm, numPostingsLists)
}
if cap(s.freqNormsBacking) >= totTFs {
s.freqNormsBacking = s.freqNormsBacking[:totTFs]
} else {
s.freqNormsBacking = make([]interimFreqNorm, totTFs)
}
freqNormsBacking := s.freqNormsBacking
for pid, numTerms := range s.numTermsPerPostingsList {
s.FreqNorms[pid] = freqNormsBacking[0:0]
freqNormsBacking = freqNormsBacking[numTerms:]
}
if cap(s.Locs) >= numPostingsLists {
s.Locs = s.Locs[:numPostingsLists]
} else {
s.Locs = make([][]interimLoc, numPostingsLists)
}
if cap(s.locsBacking) >= totLocs {
s.locsBacking = s.locsBacking[:totLocs]
} else {
s.locsBacking = make([]interimLoc, totLocs)
}
locsBacking := s.locsBacking
for pid, numLocs := range s.numLocsPerPostingsList {
s.Locs[pid] = locsBacking[0:0]
locsBacking = locsBacking[numLocs:]
}
}
func (s *interim) processDocuments() {
numFields := len(s.FieldsInv)
reuseFieldLens := make([]int, numFields)
reuseFieldTFs := make([]analysis.TokenFrequencies, numFields)
for docNum, result := range s.results {
for i := 0; i < numFields; i++ { // clear these for reuse
reuseFieldLens[i] = 0
reuseFieldTFs[i] = nil
}
s.processDocument(uint64(docNum), result,
reuseFieldLens, reuseFieldTFs)
}
}
func (s *interim) processDocument(docNum uint64,
result *index.AnalysisResult,
fieldLens []int, fieldTFs []analysis.TokenFrequencies) {
visitField := func(fieldID uint16, fieldName string,
ln int, tf analysis.TokenFrequencies) {
fieldLens[fieldID] += ln
existingFreqs := fieldTFs[fieldID]
if existingFreqs != nil {
existingFreqs.MergeAll(fieldName, tf)
} else {
fieldTFs[fieldID] = tf
}
}
// walk each composite field
for _, field := range result.Document.CompositeFields {
fieldID := uint16(s.getOrDefineField(field.Name()))
ln, tf := field.Analyze()
visitField(fieldID, field.Name(), ln, tf)
}
// walk each field
for i, field := range result.Document.Fields {
fieldID := uint16(s.getOrDefineField(field.Name()))
ln := result.Length[i]
tf := result.Analyzed[i]
visitField(fieldID, field.Name(), ln, tf)
}
// now that it's been rolled up into fieldTFs, walk that
for fieldID, tfs := range fieldTFs {
dict := s.Dicts[fieldID]
norm := float32(1.0 / math.Sqrt(float64(fieldLens[fieldID])))
for term, tf := range tfs {
pid := dict[term] - 1
bs := s.Postings[pid]
bs.Add(uint32(docNum))
s.FreqNorms[pid] = append(s.FreqNorms[pid],
interimFreqNorm{
freq: uint64(tf.Frequency()),
norm: norm,
numLocs: len(tf.Locations),
})
if len(tf.Locations) > 0 {
locs := s.Locs[pid]
for _, loc := range tf.Locations {
var locf = uint16(fieldID)
if loc.Field != "" {
locf = uint16(s.getOrDefineField(loc.Field))
}
var arrayposs []uint64
if len(loc.ArrayPositions) > 0 {
arrayposs = loc.ArrayPositions
}
locs = append(locs, interimLoc{
fieldID: locf,
pos: uint64(loc.Position),
start: uint64(loc.Start),
end: uint64(loc.End),
arrayposs: arrayposs,
})
}
s.Locs[pid] = locs
}
}
}
}
func (s *interim) writeStoredFields() (
storedIndexOffset uint64, err error) {
varBuf := make([]byte, binary.MaxVarintLen64)
metaEncode := func(val uint64) (int, error) {
wb := binary.PutUvarint(varBuf, val)
return s.metaBuf.Write(varBuf[:wb])
}
data, compressed := s.tmp0[:0], s.tmp1[:0]
defer func() { s.tmp0, s.tmp1 = data, compressed }()
// keyed by docNum
docStoredOffsets := make([]uint64, len(s.results))
// keyed by fieldID, for the current doc in the loop
docStoredFields := map[uint16]interimStoredField{}
for docNum, result := range s.results {
for fieldID := range docStoredFields { // reset for next doc
delete(docStoredFields, fieldID)
}
for _, field := range result.Document.Fields {
fieldID := uint16(s.getOrDefineField(field.Name()))
opts := field.Options()
if opts.IsStored() {
isf := docStoredFields[fieldID]
isf.vals = append(isf.vals, field.Value())
isf.typs = append(isf.typs, encodeFieldType(field))
isf.arrayposs = append(isf.arrayposs, field.ArrayPositions())
docStoredFields[fieldID] = isf
}
if opts.IncludeDocValues() {
s.IncludeDocValues[fieldID] = true
}
err := ValidateDocFields(field)
if err != nil {
return 0, err
}
}
var curr int
s.metaBuf.Reset()
data = data[:0]
// _id field special case optimizes ExternalID() lookups
idFieldVal := docStoredFields[uint16(0)].vals[0]
_, err = metaEncode(uint64(len(idFieldVal)))
if err != nil {
return 0, err
}
// handle non-"_id" fields
for fieldID := 1; fieldID < len(s.FieldsInv); fieldID++ {
isf, exists := docStoredFields[uint16(fieldID)]
if exists {
curr, data, err = persistStoredFieldValues(
fieldID, isf.vals, isf.typs, isf.arrayposs,
curr, metaEncode, data)
if err != nil {
return 0, err
}
}
}
metaBytes := s.metaBuf.Bytes()
compressed = snappy.Encode(compressed[:cap(compressed)], data)
docStoredOffsets[docNum] = uint64(s.w.Count())
_, err := writeUvarints(s.w,
uint64(len(metaBytes)),
uint64(len(idFieldVal)+len(compressed)))
if err != nil {
return 0, err
}
_, err = s.w.Write(metaBytes)
if err != nil {
return 0, err
}
_, err = s.w.Write(idFieldVal)
if err != nil {
return 0, err
}
_, err = s.w.Write(compressed)
if err != nil {
return 0, err
}
}
storedIndexOffset = uint64(s.w.Count())
for _, docStoredOffset := range docStoredOffsets {
err = binary.Write(s.w, binary.BigEndian, docStoredOffset)
if err != nil {
return 0, err
}
}
return storedIndexOffset, nil
}
func (s *interim) writeDicts() (fdvIndexOffset uint64, dictOffsets []uint64, err error) {
dictOffsets = make([]uint64, len(s.FieldsInv))
fdvOffsetsStart := make([]uint64, len(s.FieldsInv))
fdvOffsetsEnd := make([]uint64, len(s.FieldsInv))
buf := s.grabBuf(binary.MaxVarintLen64)
// these int coders are initialized with chunk size 1024
// however this will be reset to the correct chunk size
// while processing each individual field-term section
tfEncoder := newChunkedIntCoder(1024, uint64(len(s.results)-1))
locEncoder := newChunkedIntCoder(1024, uint64(len(s.results)-1))
var docTermMap [][]byte
if s.builder == nil {
s.builder, err = vellum.New(&s.builderBuf, nil)
if err != nil {
return 0, nil, err
}
}
for fieldID, terms := range s.DictKeys {
if cap(docTermMap) < len(s.results) {
docTermMap = make([][]byte, len(s.results))
} else {
docTermMap = docTermMap[0:len(s.results)]
for docNum := range docTermMap { // reset the docTermMap
docTermMap[docNum] = docTermMap[docNum][:0]
}
}
dict := s.Dicts[fieldID]
for _, term := range terms { // terms are already sorted
pid := dict[term] - 1
postingsBS := s.Postings[pid]
freqNorms := s.FreqNorms[pid]
freqNormOffset := 0
locs := s.Locs[pid]
locOffset := 0
chunkSize, err := getChunkSize(s.chunkMode, postingsBS.GetCardinality(), uint64(len(s.results)))
if err != nil {
return 0, nil, err
}
tfEncoder.SetChunkSize(chunkSize, uint64(len(s.results)-1))
locEncoder.SetChunkSize(chunkSize, uint64(len(s.results)-1))
postingsItr := postingsBS.Iterator()
for postingsItr.HasNext() {
docNum := uint64(postingsItr.Next())
freqNorm := freqNorms[freqNormOffset]
err = tfEncoder.Add(docNum,
encodeFreqHasLocs(freqNorm.freq, freqNorm.numLocs > 0),
uint64(math.Float32bits(freqNorm.norm)))
if err != nil {
return 0, nil, err
}
if freqNorm.numLocs > 0 {
numBytesLocs := 0
for _, loc := range locs[locOffset : locOffset+freqNorm.numLocs] {
numBytesLocs += totalUvarintBytes(
uint64(loc.fieldID), loc.pos, loc.start, loc.end,
uint64(len(loc.arrayposs)), loc.arrayposs)
}
err = locEncoder.Add(docNum, uint64(numBytesLocs))
if err != nil {
return 0, nil, err
}
for _, loc := range locs[locOffset : locOffset+freqNorm.numLocs] {
err = locEncoder.Add(docNum,
uint64(loc.fieldID), loc.pos, loc.start, loc.end,
uint64(len(loc.arrayposs)))
if err != nil {
return 0, nil, err
}
err = locEncoder.Add(docNum, loc.arrayposs...)
if err != nil {
return 0, nil, err
}
}
locOffset += freqNorm.numLocs
}
freqNormOffset++
docTermMap[docNum] = append(
append(docTermMap[docNum], term...),
termSeparator)
}
tfEncoder.Close()
locEncoder.Close()
postingsOffset, err :=
writePostings(postingsBS, tfEncoder, locEncoder, nil, s.w, buf)
if err != nil {
return 0, nil, err
}
if postingsOffset > uint64(0) {
err = s.builder.Insert([]byte(term), postingsOffset)
if err != nil {
return 0, nil, err
}
}
tfEncoder.Reset()
locEncoder.Reset()
}
err = s.builder.Close()
if err != nil {
return 0, nil, err
}
// record where this dictionary starts
dictOffsets[fieldID] = uint64(s.w.Count())
vellumData := s.builderBuf.Bytes()
// write out the length of the vellum data
n := binary.PutUvarint(buf, uint64(len(vellumData)))
_, err = s.w.Write(buf[:n])
if err != nil {
return 0, nil, err
}
// write this vellum to disk
_, err = s.w.Write(vellumData)
if err != nil {
return 0, nil, err
}
// reset vellum for reuse
s.builderBuf.Reset()
err = s.builder.Reset(&s.builderBuf)
if err != nil {
return 0, nil, err
}
// write the field doc values
// NOTE: doc values continue to use legacy chunk mode
chunkSize, err := getChunkSize(LegacyChunkMode, 0, 0)
if err != nil {
return 0, nil, err
}
fdvEncoder := newChunkedContentCoder(chunkSize, uint64(len(s.results)-1), s.w, false)
if s.IncludeDocValues[fieldID] {
for docNum, docTerms := range docTermMap {
if len(docTerms) > 0 {
err = fdvEncoder.Add(uint64(docNum), docTerms)
if err != nil {
return 0, nil, err
}
}
}
err = fdvEncoder.Close()
if err != nil {
return 0, nil, err
}
fdvOffsetsStart[fieldID] = uint64(s.w.Count())
_, err = fdvEncoder.Write()
if err != nil {
return 0, nil, err
}
fdvOffsetsEnd[fieldID] = uint64(s.w.Count())
fdvEncoder.Reset()
} else {
fdvOffsetsStart[fieldID] = fieldNotUninverted
fdvOffsetsEnd[fieldID] = fieldNotUninverted
}
}
fdvIndexOffset = uint64(s.w.Count())
for i := 0; i < len(fdvOffsetsStart); i++ {
n := binary.PutUvarint(buf, fdvOffsetsStart[i])
_, err := s.w.Write(buf[:n])
if err != nil {
return 0, nil, err
}
n = binary.PutUvarint(buf, fdvOffsetsEnd[i])
_, err = s.w.Write(buf[:n])
if err != nil {
return 0, nil, err
}
}
return fdvIndexOffset, dictOffsets, nil
}
func encodeFieldType(f document.Field) byte {
fieldType := byte('x')
switch f.(type) {
case *document.TextField:
fieldType = 't'
case *document.NumericField:
fieldType = 'n'
case *document.DateTimeField:
fieldType = 'd'
case *document.BooleanField:
fieldType = 'b'
case *document.GeoPointField:
fieldType = 'g'
case *document.CompositeField:
fieldType = 'c'
}
return fieldType
}
// returns the total # of bytes needed to encode the given uint64's
// into binary.PutUVarint() encoding
func totalUvarintBytes(a, b, c, d, e uint64, more []uint64) (n int) {
n = numUvarintBytes(a)
n += numUvarintBytes(b)
n += numUvarintBytes(c)
n += numUvarintBytes(d)
n += numUvarintBytes(e)
for _, v := range more {
n += numUvarintBytes(v)
}
return n
}
// returns # of bytes needed to encode x in binary.PutUvarint() encoding
func numUvarintBytes(x uint64) (n int) {
for x >= 0x80 {
x >>= 7
n++
}
return n + 1
}

@ -0,0 +1,37 @@
// Copyright (c) 2020 Couchbase, Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
package zap
import (
"github.com/blevesearch/bleve/index/scorch/segment"
)
// ZapPlugin implements the Plugin interface of
// the blevesearch/bleve/index/scorch/segment pkg
type ZapPlugin struct{}
func (*ZapPlugin) Type() string {
return Type
}
func (*ZapPlugin) Version() uint32 {
return Version
}
// Plugin returns an instance segment.Plugin for use
// by the Scorch indexing scheme
func Plugin() segment.Plugin {
return &ZapPlugin{}
}

@ -0,0 +1,798 @@
// Copyright (c) 2017 Couchbase, Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
package zap
import (
"encoding/binary"
"fmt"
"math"
"reflect"
"github.com/RoaringBitmap/roaring"
"github.com/blevesearch/bleve/index/scorch/segment"
"github.com/blevesearch/bleve/size"
)
var reflectStaticSizePostingsList int
var reflectStaticSizePostingsIterator int
var reflectStaticSizePosting int
var reflectStaticSizeLocation int
func init() {
var pl PostingsList
reflectStaticSizePostingsList = int(reflect.TypeOf(pl).Size())
var pi PostingsIterator
reflectStaticSizePostingsIterator = int(reflect.TypeOf(pi).Size())
var p Posting
reflectStaticSizePosting = int(reflect.TypeOf(p).Size())
var l Location
reflectStaticSizeLocation = int(reflect.TypeOf(l).Size())
}
// FST or vellum value (uint64) encoding is determined by the top two
// highest-order or most significant bits...
//
// encoding : MSB
// name : 63 62 61...to...bit #0 (LSB)
// ----------+---+---+---------------------------------------------------
// general : 0 | 0 | 62-bits of postingsOffset.
// ~ : 0 | 1 | reserved for future.
// 1-hit : 1 | 0 | 31-bits of positive float31 norm | 31-bits docNum.
// ~ : 1 | 1 | reserved for future.
//
// Encoding "general" is able to handle all cases, where the
// postingsOffset points to more information about the postings for
// the term.
//
// Encoding "1-hit" is used to optimize a commonly seen case when a
// term has only a single hit. For example, a term in the _id field
// will have only 1 hit. The "1-hit" encoding is used for a term
// in a field when...
//
// - term vector info is disabled for that field;
// - and, the term appears in only a single doc for that field;
// - and, the term's freq is exactly 1 in that single doc for that field;
// - and, the docNum must fit into 31-bits;
//
// Otherwise, the "general" encoding is used instead.
//
// In the "1-hit" encoding, the field in that single doc may have
// other terms, which is supported in the "1-hit" encoding by the
// positive float31 norm.
const FSTValEncodingMask = uint64(0xc000000000000000)
const FSTValEncodingGeneral = uint64(0x0000000000000000)
const FSTValEncoding1Hit = uint64(0x8000000000000000)
func FSTValEncode1Hit(docNum uint64, normBits uint64) uint64 {
return FSTValEncoding1Hit | ((mask31Bits & normBits) << 31) | (mask31Bits & docNum)
}
func FSTValDecode1Hit(v uint64) (docNum uint64, normBits uint64) {
return (mask31Bits & v), (mask31Bits & (v >> 31))
}
const mask31Bits = uint64(0x000000007fffffff)
func under32Bits(x uint64) bool {
return x <= mask31Bits
}
const DocNum1HitFinished = math.MaxUint64
var NormBits1Hit = uint64(math.Float32bits(float32(1)))
// PostingsList is an in-memory representation of a postings list
type PostingsList struct {
sb *SegmentBase
postingsOffset uint64
freqOffset uint64
locOffset uint64
postings *roaring.Bitmap
except *roaring.Bitmap
// when normBits1Hit != 0, then this postings list came from a
// 1-hit encoding, and only the docNum1Hit & normBits1Hit apply
docNum1Hit uint64
normBits1Hit uint64
}
// represents an immutable, empty postings list
var emptyPostingsList = &PostingsList{}
func (p *PostingsList) Size() int {
sizeInBytes := reflectStaticSizePostingsList + size.SizeOfPtr
if p.except != nil {
sizeInBytes += int(p.except.GetSizeInBytes())
}
return sizeInBytes
}
func (p *PostingsList) OrInto(receiver *roaring.Bitmap) {
if p.normBits1Hit != 0 {
receiver.Add(uint32(p.docNum1Hit))
return
}
if p.postings != nil {
receiver.Or(p.postings)
}
}
// Iterator returns an iterator for this postings list
func (p *PostingsList) Iterator(includeFreq, includeNorm, includeLocs bool,
prealloc segment.PostingsIterator) segment.PostingsIterator {
if p.normBits1Hit == 0 && p.postings == nil {
return emptyPostingsIterator
}
var preallocPI *PostingsIterator
pi, ok := prealloc.(*PostingsIterator)
if ok && pi != nil {
preallocPI = pi
}
if preallocPI == emptyPostingsIterator {
preallocPI = nil
}
return p.iterator(includeFreq, includeNorm, includeLocs, preallocPI)
}
func (p *PostingsList) iterator(includeFreq, includeNorm, includeLocs bool,
rv *PostingsIterator) *PostingsIterator {
if rv == nil {
rv = &PostingsIterator{}
} else {
freqNormReader := rv.freqNormReader
if freqNormReader != nil {
freqNormReader.reset()
}
locReader := rv.locReader
if locReader != nil {
locReader.reset()
}
nextLocs := rv.nextLocs[:0]
nextSegmentLocs := rv.nextSegmentLocs[:0]
buf := rv.buf
*rv = PostingsIterator{} // clear the struct
rv.freqNormReader = freqNormReader
rv.locReader = locReader
rv.nextLocs = nextLocs
rv.nextSegmentLocs = nextSegmentLocs
rv.buf = buf
}
rv.postings = p
rv.includeFreqNorm = includeFreq || includeNorm || includeLocs
rv.includeLocs = includeLocs
if p.normBits1Hit != 0 {
// "1-hit" encoding
rv.docNum1Hit = p.docNum1Hit
rv.normBits1Hit = p.normBits1Hit
if p.except != nil && p.except.Contains(uint32(rv.docNum1Hit)) {
rv.docNum1Hit = DocNum1HitFinished
}
return rv
}
// "general" encoding, check if empty
if p.postings == nil {
return rv
}
// initialize freq chunk reader
if rv.includeFreqNorm {
rv.freqNormReader = newChunkedIntDecoder(p.sb.mem, p.freqOffset)
}
// initialize the loc chunk reader
if rv.includeLocs {
rv.locReader = newChunkedIntDecoder(p.sb.mem, p.locOffset)
}
rv.all = p.postings.Iterator()
if p.except != nil {
rv.ActualBM = roaring.AndNot(p.postings, p.except)
rv.Actual = rv.ActualBM.Iterator()
} else {
rv.ActualBM = p.postings
rv.Actual = rv.all // Optimize to use same iterator for all & Actual.
}
return rv
}
// Count returns the number of items on this postings list
func (p *PostingsList) Count() uint64 {
var n, e uint64
if p.normBits1Hit != 0 {
n = 1
if p.except != nil && p.except.Contains(uint32(p.docNum1Hit)) {
e = 1
}
} else if p.postings != nil {
n = p.postings.GetCardinality()
if p.except != nil {
e = p.postings.AndCardinality(p.except)
}
}
return n - e
}
func (rv *PostingsList) read(postingsOffset uint64, d *Dictionary) error {
rv.postingsOffset = postingsOffset
// handle "1-hit" encoding special case
if rv.postingsOffset&FSTValEncodingMask == FSTValEncoding1Hit {
return rv.init1Hit(postingsOffset)
}
// read the location of the freq/norm details
var n uint64
var read int
rv.freqOffset, read = binary.Uvarint(d.sb.mem[postingsOffset+n : postingsOffset+binary.MaxVarintLen64])
n += uint64(read)
rv.locOffset, read = binary.Uvarint(d.sb.mem[postingsOffset+n : postingsOffset+n+binary.MaxVarintLen64])
n += uint64(read)
var postingsLen uint64
postingsLen, read = binary.Uvarint(d.sb.mem[postingsOffset+n : postingsOffset+n+binary.MaxVarintLen64])
n += uint64(read)
roaringBytes := d.sb.mem[postingsOffset+n : postingsOffset+n+postingsLen]
if rv.postings == nil {
rv.postings = roaring.NewBitmap()
}
_, err := rv.postings.FromBuffer(roaringBytes)
if err != nil {
return fmt.Errorf("error loading roaring bitmap: %v", err)
}
return nil
}
func (rv *PostingsList) init1Hit(fstVal uint64) error {
docNum, normBits := FSTValDecode1Hit(fstVal)
rv.docNum1Hit = docNum
rv.normBits1Hit = normBits
return nil
}
// PostingsIterator provides a way to iterate through the postings list
type PostingsIterator struct {
postings *PostingsList
all roaring.IntPeekable
Actual roaring.IntPeekable
ActualBM *roaring.Bitmap
currChunk uint32
freqNormReader *chunkedIntDecoder
locReader *chunkedIntDecoder
next Posting // reused across Next() calls
nextLocs []Location // reused across Next() calls
nextSegmentLocs []segment.Location // reused across Next() calls
docNum1Hit uint64
normBits1Hit uint64
buf []byte
includeFreqNorm bool
includeLocs bool
}
var emptyPostingsIterator = &PostingsIterator{}
func (i *PostingsIterator) Size() int {
sizeInBytes := reflectStaticSizePostingsIterator + size.SizeOfPtr +
i.next.Size()
// account for freqNormReader, locReader if we start using this.
for _, entry := range i.nextLocs {
sizeInBytes += entry.Size()
}
return sizeInBytes
}
func (i *PostingsIterator) loadChunk(chunk int) error {
if i.includeFreqNorm {
err := i.freqNormReader.loadChunk(chunk)
if err != nil {
return err
}
}
if i.includeLocs {
err := i.locReader.loadChunk(chunk)
if err != nil {
return err
}
}
i.currChunk = uint32(chunk)
return nil
}
func (i *PostingsIterator) readFreqNormHasLocs() (uint64, uint64, bool, error) {
if i.normBits1Hit != 0 {
return 1, i.normBits1Hit, false, nil
}
freqHasLocs, err := i.freqNormReader.readUvarint()
if err != nil {
return 0, 0, false, fmt.Errorf("error reading frequency: %v", err)
}
freq, hasLocs := decodeFreqHasLocs(freqHasLocs)
normBits, err := i.freqNormReader.readUvarint()
if err != nil {
return 0, 0, false, fmt.Errorf("error reading norm: %v", err)
}
return freq, normBits, hasLocs, nil
}
func (i *PostingsIterator) skipFreqNormReadHasLocs() (bool, error) {
if i.normBits1Hit != 0 {
return false, nil
}
freqHasLocs, err := i.freqNormReader.readUvarint()
if err != nil {
return false, fmt.Errorf("error reading freqHasLocs: %v", err)
}
i.freqNormReader.SkipUvarint() // Skip normBits.
return freqHasLocs&0x01 != 0, nil // See decodeFreqHasLocs() / hasLocs.
}
func encodeFreqHasLocs(freq uint64, hasLocs bool) uint64 {
rv := freq << 1
if hasLocs {
rv = rv | 0x01 // 0'th LSB encodes whether there are locations
}
return rv
}
func decodeFreqHasLocs(freqHasLocs uint64) (uint64, bool) {
freq := freqHasLocs >> 1
hasLocs := freqHasLocs&0x01 != 0
return freq, hasLocs
}
// readLocation processes all the integers on the stream representing a single
// location.
func (i *PostingsIterator) readLocation(l *Location) error {
// read off field
fieldID, err := i.locReader.readUvarint()
if err != nil {
return fmt.Errorf("error reading location field: %v", err)
}
// read off pos
pos, err := i.locReader.readUvarint()
if err != nil {
return fmt.Errorf("error reading location pos: %v", err)
}
// read off start
start, err := i.locReader.readUvarint()
if err != nil {
return fmt.Errorf("error reading location start: %v", err)
}
// read off end
end, err := i.locReader.readUvarint()
if err != nil {
return fmt.Errorf("error reading location end: %v", err)
}
// read off num array pos
numArrayPos, err := i.locReader.readUvarint()
if err != nil {
return fmt.Errorf("error reading location num array pos: %v", err)
}
l.field = i.postings.sb.fieldsInv[fieldID]
l.pos = pos
l.start = start
l.end = end
if cap(l.ap) < int(numArrayPos) {
l.ap = make([]uint64, int(numArrayPos))
} else {
l.ap = l.ap[:int(numArrayPos)]
}
// read off array positions
for k := 0; k < int(numArrayPos); k++ {
ap, err := i.locReader.readUvarint()
if err != nil {
return fmt.Errorf("error reading array position: %v", err)
}
l.ap[k] = ap
}
return nil
}
// Next returns the next posting on the postings list, or nil at the end
func (i *PostingsIterator) Next() (segment.Posting, error) {
return i.nextAtOrAfter(0)
}
// Advance returns the posting at the specified docNum or it is not present
// the next posting, or if the end is reached, nil
func (i *PostingsIterator) Advance(docNum uint64) (segment.Posting, error) {
return i.nextAtOrAfter(docNum)
}
// Next returns the next posting on the postings list, or nil at the end
func (i *PostingsIterator) nextAtOrAfter(atOrAfter uint64) (segment.Posting, error) {
docNum, exists, err := i.nextDocNumAtOrAfter(atOrAfter)
if err != nil || !exists {
return nil, err
}
i.next = Posting{} // clear the struct
rv := &i.next
rv.docNum = docNum
if !i.includeFreqNorm {
return rv, nil
}
var normBits uint64
var hasLocs bool
rv.freq, normBits, hasLocs, err = i.readFreqNormHasLocs()
if err != nil {
return nil, err
}
rv.norm = math.Float32frombits(uint32(normBits))
if i.includeLocs && hasLocs {
// prepare locations into reused slices, where we assume
// rv.freq >= "number of locs", since in a composite field,
// some component fields might have their IncludeTermVector
// flags disabled while other component fields are enabled
if cap(i.nextLocs) >= int(rv.freq) {
i.nextLocs = i.nextLocs[0:rv.freq]
} else {
i.nextLocs = make([]Location, rv.freq, rv.freq*2)
}
if cap(i.nextSegmentLocs) < int(rv.freq) {
i.nextSegmentLocs = make([]segment.Location, rv.freq, rv.freq*2)
}
rv.locs = i.nextSegmentLocs[:0]
numLocsBytes, err := i.locReader.readUvarint()
if err != nil {
return nil, fmt.Errorf("error reading location numLocsBytes: %v", err)
}
j := 0
startBytesRemaining := i.locReader.Len() // # bytes remaining in the locReader
for startBytesRemaining-i.locReader.Len() < int(numLocsBytes) {
err := i.readLocation(&i.nextLocs[j])
if err != nil {
return nil, err
}
rv.locs = append(rv.locs, &i.nextLocs[j])
j++
}
}
return rv, nil
}
// nextDocNum returns the next docNum on the postings list, and also
// sets up the currChunk / loc related fields of the iterator.
func (i *PostingsIterator) nextDocNumAtOrAfter(atOrAfter uint64) (uint64, bool, error) {
if i.normBits1Hit != 0 {
if i.docNum1Hit == DocNum1HitFinished {
return 0, false, nil
}
if i.docNum1Hit < atOrAfter {
// advanced past our 1-hit
i.docNum1Hit = DocNum1HitFinished // consume our 1-hit docNum
return 0, false, nil
}
docNum := i.docNum1Hit
i.docNum1Hit = DocNum1HitFinished // consume our 1-hit docNum
return docNum, true, nil
}
if i.Actual == nil || !i.Actual.HasNext() {
return 0, false, nil
}
if i.postings == nil || i.postings.postings == i.ActualBM {
return i.nextDocNumAtOrAfterClean(atOrAfter)
}
i.Actual.AdvanceIfNeeded(uint32(atOrAfter))
if !i.Actual.HasNext() {
// couldn't find anything
return 0, false, nil
}
n := i.Actual.Next()
allN := i.all.Next()
chunkSize, err := getChunkSize(i.postings.sb.chunkMode, i.postings.postings.GetCardinality(), i.postings.sb.numDocs)
if err != nil {
return 0, false, err
}
nChunk := n / uint32(chunkSize)
// when allN becomes >= to here, then allN is in the same chunk as nChunk.
allNReachesNChunk := nChunk * uint32(chunkSize)
// n is the next actual hit (excluding some postings), and
// allN is the next hit in the full postings, and
// if they don't match, move 'all' forwards until they do
for allN != n {
// we've reached same chunk, so move the freq/norm/loc decoders forward
if i.includeFreqNorm && allN >= allNReachesNChunk {
err := i.currChunkNext(nChunk)
if err != nil {
return 0, false, err
}
}
allN = i.all.Next()
}
if i.includeFreqNorm && (i.currChunk != nChunk || i.freqNormReader.isNil()) {
err := i.loadChunk(int(nChunk))
if err != nil {
return 0, false, fmt.Errorf("error loading chunk: %v", err)
}
}
return uint64(n), true, nil
}
// optimization when the postings list is "clean" (e.g., no updates &
// no deletions) where the all bitmap is the same as the actual bitmap
func (i *PostingsIterator) nextDocNumAtOrAfterClean(
atOrAfter uint64) (uint64, bool, error) {
if !i.includeFreqNorm {
i.Actual.AdvanceIfNeeded(uint32(atOrAfter))
if !i.Actual.HasNext() {
return 0, false, nil // couldn't find anything
}
return uint64(i.Actual.Next()), true, nil
}
chunkSize, err := getChunkSize(i.postings.sb.chunkMode, i.postings.postings.GetCardinality(), i.postings.sb.numDocs)
if err != nil {
return 0, false, err
}
// freq-norm's needed, so maintain freq-norm chunk reader
sameChunkNexts := 0 // # of times we called Next() in the same chunk
n := i.Actual.Next()
nChunk := n / uint32(chunkSize)
for uint64(n) < atOrAfter && i.Actual.HasNext() {
n = i.Actual.Next()
nChunkPrev := nChunk
nChunk = n / uint32(chunkSize)
if nChunk != nChunkPrev {
sameChunkNexts = 0
} else {
sameChunkNexts += 1
}
}
if uint64(n) < atOrAfter {
// couldn't find anything
return 0, false, nil
}
for j := 0; j < sameChunkNexts; j++ {
err := i.currChunkNext(nChunk)
if err != nil {
return 0, false, fmt.Errorf("error optimized currChunkNext: %v", err)
}
}
if i.currChunk != nChunk || i.freqNormReader.isNil() {
err := i.loadChunk(int(nChunk))
if err != nil {
return 0, false, fmt.Errorf("error loading chunk: %v", err)
}
}
return uint64(n), true, nil
}
func (i *PostingsIterator) currChunkNext(nChunk uint32) error {
if i.currChunk != nChunk || i.freqNormReader.isNil() {
err := i.loadChunk(int(nChunk))
if err != nil {
return fmt.Errorf("error loading chunk: %v", err)
}
}
// read off freq/offsets even though we don't care about them
hasLocs, err := i.skipFreqNormReadHasLocs()
if err != nil {
return err
}
if i.includeLocs && hasLocs {
numLocsBytes, err := i.locReader.readUvarint()
if err != nil {
return fmt.Errorf("error reading location numLocsBytes: %v", err)
}
// skip over all the location bytes
i.locReader.SkipBytes(int(numLocsBytes))
}
return nil
}
// DocNum1Hit returns the docNum and true if this is "1-hit" optimized
// and the docNum is available.
func (p *PostingsIterator) DocNum1Hit() (uint64, bool) {
if p.normBits1Hit != 0 && p.docNum1Hit != DocNum1HitFinished {
return p.docNum1Hit, true
}
return 0, false
}
// ActualBitmap returns the underlying actual bitmap
// which can be used up the stack for optimizations
func (p *PostingsIterator) ActualBitmap() *roaring.Bitmap {
return p.ActualBM
}
// ReplaceActual replaces the ActualBM with the provided
// bitmap
func (p *PostingsIterator) ReplaceActual(abm *roaring.Bitmap) {
p.ActualBM = abm
p.Actual = abm.Iterator()
}
// PostingsIteratorFromBitmap constructs a PostingsIterator given an
// "actual" bitmap.
func PostingsIteratorFromBitmap(bm *roaring.Bitmap,
includeFreqNorm, includeLocs bool) (segment.PostingsIterator, error) {
return &PostingsIterator{
ActualBM: bm,
Actual: bm.Iterator(),
includeFreqNorm: includeFreqNorm,
includeLocs: includeLocs,
}, nil
}
// PostingsIteratorFrom1Hit constructs a PostingsIterator given a
// 1-hit docNum.
func PostingsIteratorFrom1Hit(docNum1Hit uint64,
includeFreqNorm, includeLocs bool) (segment.PostingsIterator, error) {
return &PostingsIterator{
docNum1Hit: docNum1Hit,
normBits1Hit: NormBits1Hit,
includeFreqNorm: includeFreqNorm,
includeLocs: includeLocs,
}, nil
}
// Posting is a single entry in a postings list
type Posting struct {
docNum uint64
freq uint64
norm float32
locs []segment.Location
}
func (p *Posting) Size() int {
sizeInBytes := reflectStaticSizePosting
for _, entry := range p.locs {
sizeInBytes += entry.Size()
}
return sizeInBytes
}
// Number returns the document number of this posting in this segment
func (p *Posting) Number() uint64 {
return p.docNum
}
// Frequency returns the frequencies of occurrence of this term in this doc/field
func (p *Posting) Frequency() uint64 {
return p.freq
}
// Norm returns the normalization factor for this posting
func (p *Posting) Norm() float64 {
return float64(p.norm)
}
// Locations returns the location information for each occurrence
func (p *Posting) Locations() []segment.Location {
return p.locs
}
// Location represents the location of a single occurrence
type Location struct {
field string
pos uint64
start uint64
end uint64
ap []uint64
}
func (l *Location) Size() int {
return reflectStaticSizeLocation +
len(l.field) +
len(l.ap)*size.SizeOfUint64
}
// Field returns the name of the field (useful in composite fields to know
// which original field the value came from)
func (l *Location) Field() string {
return l.field
}
// Start returns the start byte offset of this occurrence
func (l *Location) Start() uint64 {
return l.start
}
// End returns the end byte offset of this occurrence
func (l *Location) End() uint64 {
return l.end
}
// Pos returns the 1-based phrase position of this occurrence
func (l *Location) Pos() uint64 {
return l.pos
}
// ArrayPositions returns the array position vector associated with this occurrence
func (l *Location) ArrayPositions() []uint64 {
return l.ap
}

@ -0,0 +1,43 @@
// Copyright (c) 2017 Couchbase, Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
package zap
import "encoding/binary"
func (s *SegmentBase) getDocStoredMetaAndCompressed(docNum uint64) ([]byte, []byte) {
_, storedOffset, n, metaLen, dataLen := s.getDocStoredOffsets(docNum)
meta := s.mem[storedOffset+n : storedOffset+n+metaLen]
data := s.mem[storedOffset+n+metaLen : storedOffset+n+metaLen+dataLen]
return meta, data
}
func (s *SegmentBase) getDocStoredOffsets(docNum uint64) (
uint64, uint64, uint64, uint64, uint64) {
indexOffset := s.storedIndexOffset + (8 * docNum)
storedOffset := binary.BigEndian.Uint64(s.mem[indexOffset : indexOffset+8])
var n uint64
metaLen, read := binary.Uvarint(s.mem[storedOffset : storedOffset+binary.MaxVarintLen64])
n += uint64(read)
dataLen, read := binary.Uvarint(s.mem[storedOffset+n : storedOffset+n+binary.MaxVarintLen64])
n += uint64(read)
return indexOffset, storedOffset, n, metaLen, dataLen
}

@ -0,0 +1,572 @@
// Copyright (c) 2017 Couchbase, Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
package zap
import (
"bytes"
"encoding/binary"
"fmt"
"io"
"os"
"sync"
"unsafe"
"github.com/RoaringBitmap/roaring"
"github.com/blevesearch/bleve/index/scorch/segment"
"github.com/blevesearch/bleve/size"
"github.com/couchbase/vellum"
mmap "github.com/blevesearch/mmap-go"
"github.com/golang/snappy"
)
var reflectStaticSizeSegmentBase int
func init() {
var sb SegmentBase
reflectStaticSizeSegmentBase = int(unsafe.Sizeof(sb))
}
// Open returns a zap impl of a segment
func (*ZapPlugin) Open(path string) (segment.Segment, error) {
f, err := os.Open(path)
if err != nil {
return nil, err
}
mm, err := mmap.Map(f, mmap.RDONLY, 0)
if err != nil {
// mmap failed, try to close the file
_ = f.Close()
return nil, err
}
rv := &Segment{
SegmentBase: SegmentBase{
mem: mm[0 : len(mm)-FooterSize],
fieldsMap: make(map[string]uint16),
fieldDvReaders: make(map[uint16]*docValueReader),
fieldFSTs: make(map[uint16]*vellum.FST),
},
f: f,
mm: mm,
path: path,
refs: 1,
}
rv.SegmentBase.updateSize()
err = rv.loadConfig()
if err != nil {
_ = rv.Close()
return nil, err
}
err = rv.loadFields()
if err != nil {
_ = rv.Close()
return nil, err
}
err = rv.loadDvReaders()
if err != nil {
_ = rv.Close()
return nil, err
}
return rv, nil
}
// SegmentBase is a memory only, read-only implementation of the
// segment.Segment interface, using zap's data representation.
type SegmentBase struct {
mem []byte
memCRC uint32
chunkMode uint32
fieldsMap map[string]uint16 // fieldName -> fieldID+1
fieldsInv []string // fieldID -> fieldName
numDocs uint64
storedIndexOffset uint64
fieldsIndexOffset uint64
docValueOffset uint64
dictLocs []uint64
fieldDvReaders map[uint16]*docValueReader // naive chunk cache per field
fieldDvNames []string // field names cached in fieldDvReaders
size uint64
m sync.Mutex
fieldFSTs map[uint16]*vellum.FST
}
func (sb *SegmentBase) Size() int {
return int(sb.size)
}
func (sb *SegmentBase) updateSize() {
sizeInBytes := reflectStaticSizeSegmentBase +
cap(sb.mem)
// fieldsMap
for k := range sb.fieldsMap {
sizeInBytes += (len(k) + size.SizeOfString) + size.SizeOfUint16
}
// fieldsInv, dictLocs
for _, entry := range sb.fieldsInv {
sizeInBytes += len(entry) + size.SizeOfString
}
sizeInBytes += len(sb.dictLocs) * size.SizeOfUint64
// fieldDvReaders
for _, v := range sb.fieldDvReaders {
sizeInBytes += size.SizeOfUint16 + size.SizeOfPtr
if v != nil {
sizeInBytes += v.size()
}
}
sb.size = uint64(sizeInBytes)
}
func (sb *SegmentBase) AddRef() {}
func (sb *SegmentBase) DecRef() (err error) { return nil }
func (sb *SegmentBase) Close() (err error) { return nil }
// Segment implements a persisted segment.Segment interface, by
// embedding an mmap()'ed SegmentBase.
type Segment struct {
SegmentBase
f *os.File
mm mmap.MMap
path string
version uint32
crc uint32
m sync.Mutex // Protects the fields that follow.
refs int64
}
func (s *Segment) Size() int {
// 8 /* size of file pointer */
// 4 /* size of version -> uint32 */
// 4 /* size of crc -> uint32 */
sizeOfUints := 16
sizeInBytes := (len(s.path) + size.SizeOfString) + sizeOfUints
// mutex, refs -> int64
sizeInBytes += 16
// do not include the mmap'ed part
return sizeInBytes + s.SegmentBase.Size() - cap(s.mem)
}
func (s *Segment) AddRef() {
s.m.Lock()
s.refs++
s.m.Unlock()
}
func (s *Segment) DecRef() (err error) {
s.m.Lock()
s.refs--
if s.refs == 0 {
err = s.closeActual()
}
s.m.Unlock()
return err
}
func (s *Segment) loadConfig() error {
crcOffset := len(s.mm) - 4
s.crc = binary.BigEndian.Uint32(s.mm[crcOffset : crcOffset+4])
verOffset := crcOffset - 4
s.version = binary.BigEndian.Uint32(s.mm[verOffset : verOffset+4])
if s.version != Version {
return fmt.Errorf("unsupported version %d", s.version)
}
chunkOffset := verOffset - 4
s.chunkMode = binary.BigEndian.Uint32(s.mm[chunkOffset : chunkOffset+4])
docValueOffset := chunkOffset - 8
s.docValueOffset = binary.BigEndian.Uint64(s.mm[docValueOffset : docValueOffset+8])
fieldsIndexOffset := docValueOffset - 8
s.fieldsIndexOffset = binary.BigEndian.Uint64(s.mm[fieldsIndexOffset : fieldsIndexOffset+8])
storedIndexOffset := fieldsIndexOffset - 8
s.storedIndexOffset = binary.BigEndian.Uint64(s.mm[storedIndexOffset : storedIndexOffset+8])
numDocsOffset := storedIndexOffset - 8
s.numDocs = binary.BigEndian.Uint64(s.mm[numDocsOffset : numDocsOffset+8])
return nil
}
func (s *SegmentBase) loadFields() error {
// NOTE for now we assume the fields index immediately precedes
// the footer, and if this changes, need to adjust accordingly (or
// store explicit length), where s.mem was sliced from s.mm in Open().
fieldsIndexEnd := uint64(len(s.mem))
// iterate through fields index
var fieldID uint64
for s.fieldsIndexOffset+(8*fieldID) < fieldsIndexEnd {
addr := binary.BigEndian.Uint64(s.mem[s.fieldsIndexOffset+(8*fieldID) : s.fieldsIndexOffset+(8*fieldID)+8])
dictLoc, read := binary.Uvarint(s.mem[addr:fieldsIndexEnd])
n := uint64(read)
s.dictLocs = append(s.dictLocs, dictLoc)
var nameLen uint64
nameLen, read = binary.Uvarint(s.mem[addr+n : fieldsIndexEnd])
n += uint64(read)
name := string(s.mem[addr+n : addr+n+nameLen])
s.fieldsInv = append(s.fieldsInv, name)
s.fieldsMap[name] = uint16(fieldID + 1)
fieldID++
}
return nil
}
// Dictionary returns the term dictionary for the specified field
func (s *SegmentBase) Dictionary(field string) (segment.TermDictionary, error) {
dict, err := s.dictionary(field)
if err == nil && dict == nil {
return &segment.EmptyDictionary{}, nil
}
return dict, err
}
func (sb *SegmentBase) dictionary(field string) (rv *Dictionary, err error) {
fieldIDPlus1 := sb.fieldsMap[field]
if fieldIDPlus1 > 0 {
rv = &Dictionary{
sb: sb,
field: field,
fieldID: fieldIDPlus1 - 1,
}
dictStart := sb.dictLocs[rv.fieldID]
if dictStart > 0 {
var ok bool
sb.m.Lock()
if rv.fst, ok = sb.fieldFSTs[rv.fieldID]; !ok {
// read the length of the vellum data
vellumLen, read := binary.Uvarint(sb.mem[dictStart : dictStart+binary.MaxVarintLen64])
fstBytes := sb.mem[dictStart+uint64(read) : dictStart+uint64(read)+vellumLen]
rv.fst, err = vellum.Load(fstBytes)
if err != nil {
sb.m.Unlock()
return nil, fmt.Errorf("dictionary field %s vellum err: %v", field, err)
}
sb.fieldFSTs[rv.fieldID] = rv.fst
}
sb.m.Unlock()
rv.fstReader, err = rv.fst.Reader()
if err != nil {
return nil, fmt.Errorf("dictionary field %s vellum reader err: %v", field, err)
}
}
}
return rv, nil
}
// visitDocumentCtx holds data structures that are reusable across
// multiple VisitDocument() calls to avoid memory allocations
type visitDocumentCtx struct {
buf []byte
reader bytes.Reader
arrayPos []uint64
}
var visitDocumentCtxPool = sync.Pool{
New: func() interface{} {
reuse := &visitDocumentCtx{}
return reuse
},
}
// VisitDocument invokes the DocFieldValueVistor for each stored field
// for the specified doc number
func (s *SegmentBase) VisitDocument(num uint64, visitor segment.DocumentFieldValueVisitor) error {
vdc := visitDocumentCtxPool.Get().(*visitDocumentCtx)
defer visitDocumentCtxPool.Put(vdc)
return s.visitDocument(vdc, num, visitor)
}
func (s *SegmentBase) visitDocument(vdc *visitDocumentCtx, num uint64,
visitor segment.DocumentFieldValueVisitor) error {
// first make sure this is a valid number in this segment
if num < s.numDocs {
meta, compressed := s.getDocStoredMetaAndCompressed(num)
vdc.reader.Reset(meta)
// handle _id field special case
idFieldValLen, err := binary.ReadUvarint(&vdc.reader)
if err != nil {
return err
}
idFieldVal := compressed[:idFieldValLen]
keepGoing := visitor("_id", byte('t'), idFieldVal, nil)
if !keepGoing {
visitDocumentCtxPool.Put(vdc)
return nil
}
// handle non-"_id" fields
compressed = compressed[idFieldValLen:]
uncompressed, err := snappy.Decode(vdc.buf[:cap(vdc.buf)], compressed)
if err != nil {
return err
}
for keepGoing {
field, err := binary.ReadUvarint(&vdc.reader)
if err == io.EOF {
break
}
if err != nil {
return err
}
typ, err := binary.ReadUvarint(&vdc.reader)
if err != nil {
return err
}
offset, err := binary.ReadUvarint(&vdc.reader)
if err != nil {
return err
}
l, err := binary.ReadUvarint(&vdc.reader)
if err != nil {
return err
}
numap, err := binary.ReadUvarint(&vdc.reader)
if err != nil {
return err
}
var arrayPos []uint64
if numap > 0 {
if cap(vdc.arrayPos) < int(numap) {
vdc.arrayPos = make([]uint64, numap)
}
arrayPos = vdc.arrayPos[:numap]
for i := 0; i < int(numap); i++ {
ap, err := binary.ReadUvarint(&vdc.reader)
if err != nil {
return err
}
arrayPos[i] = ap
}
}
value := uncompressed[offset : offset+l]
keepGoing = visitor(s.fieldsInv[field], byte(typ), value, arrayPos)
}
vdc.buf = uncompressed
}
return nil
}
// DocID returns the value of the _id field for the given docNum
func (s *SegmentBase) DocID(num uint64) ([]byte, error) {
if num >= s.numDocs {
return nil, nil
}
vdc := visitDocumentCtxPool.Get().(*visitDocumentCtx)
meta, compressed := s.getDocStoredMetaAndCompressed(num)
vdc.reader.Reset(meta)
// handle _id field special case
idFieldValLen, err := binary.ReadUvarint(&vdc.reader)
if err != nil {
return nil, err
}
idFieldVal := compressed[:idFieldValLen]
visitDocumentCtxPool.Put(vdc)
return idFieldVal, nil
}
// Count returns the number of documents in this segment.
func (s *SegmentBase) Count() uint64 {
return s.numDocs
}
// DocNumbers returns a bitset corresponding to the doc numbers of all the
// provided _id strings
func (s *SegmentBase) DocNumbers(ids []string) (*roaring.Bitmap, error) {
rv := roaring.New()
if len(s.fieldsMap) > 0 {
idDict, err := s.dictionary("_id")
if err != nil {
return nil, err
}
postingsList := emptyPostingsList
sMax, err := idDict.fst.GetMaxKey()
if err != nil {
return nil, err
}
sMaxStr := string(sMax)
filteredIds := make([]string, 0, len(ids))
for _, id := range ids {
if id <= sMaxStr {
filteredIds = append(filteredIds, id)
}
}
for _, id := range filteredIds {
postingsList, err = idDict.postingsList([]byte(id), nil, postingsList)
if err != nil {
return nil, err
}
postingsList.OrInto(rv)
}
}
return rv, nil
}
// Fields returns the field names used in this segment
func (s *SegmentBase) Fields() []string {
return s.fieldsInv
}
// Path returns the path of this segment on disk
func (s *Segment) Path() string {
return s.path
}
// Close releases all resources associated with this segment
func (s *Segment) Close() (err error) {
return s.DecRef()
}
func (s *Segment) closeActual() (err error) {
if s.mm != nil {
err = s.mm.Unmap()
}
// try to close file even if unmap failed
if s.f != nil {
err2 := s.f.Close()
if err == nil {
// try to return first error
err = err2
}
}
return
}
// some helpers i started adding for the command-line utility
// Data returns the underlying mmaped data slice
func (s *Segment) Data() []byte {
return s.mm
}
// CRC returns the CRC value stored in the file footer
func (s *Segment) CRC() uint32 {
return s.crc
}
// Version returns the file version in the file footer
func (s *Segment) Version() uint32 {
return s.version
}
// ChunkFactor returns the chunk factor in the file footer
func (s *Segment) ChunkMode() uint32 {
return s.chunkMode
}
// FieldsIndexOffset returns the fields index offset in the file footer
func (s *Segment) FieldsIndexOffset() uint64 {
return s.fieldsIndexOffset
}
// StoredIndexOffset returns the stored value index offset in the file footer
func (s *Segment) StoredIndexOffset() uint64 {
return s.storedIndexOffset
}
// DocValueOffset returns the docValue offset in the file footer
func (s *Segment) DocValueOffset() uint64 {
return s.docValueOffset
}
// NumDocs returns the number of documents in the file footer
func (s *Segment) NumDocs() uint64 {
return s.numDocs
}
// DictAddr is a helper function to compute the file offset where the
// dictionary is stored for the specified field.
func (s *Segment) DictAddr(field string) (uint64, error) {
fieldIDPlus1, ok := s.fieldsMap[field]
if !ok {
return 0, fmt.Errorf("no such field '%s'", field)
}
return s.dictLocs[fieldIDPlus1-1], nil
}
func (s *SegmentBase) loadDvReaders() error {
if s.docValueOffset == fieldNotUninverted || s.numDocs == 0 {
return nil
}
var read uint64
for fieldID, field := range s.fieldsInv {
var fieldLocStart, fieldLocEnd uint64
var n int
fieldLocStart, n = binary.Uvarint(s.mem[s.docValueOffset+read : s.docValueOffset+read+binary.MaxVarintLen64])
if n <= 0 {
return fmt.Errorf("loadDvReaders: failed to read the docvalue offset start for field %d", fieldID)
}
read += uint64(n)
fieldLocEnd, n = binary.Uvarint(s.mem[s.docValueOffset+read : s.docValueOffset+read+binary.MaxVarintLen64])
if n <= 0 {
return fmt.Errorf("loadDvReaders: failed to read the docvalue offset end for field %d", fieldID)
}
read += uint64(n)
fieldDvReader, err := s.loadFieldDocValueReader(field, fieldLocStart, fieldLocEnd)
if err != nil {
return err
}
if fieldDvReader != nil {
s.fieldDvReaders[uint16(fieldID)] = fieldDvReader
s.fieldDvNames = append(s.fieldDvNames, field)
}
}
return nil
}

@ -0,0 +1,145 @@
// Copyright (c) 2017 Couchbase, Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
package zap
import (
"encoding/binary"
"io"
"github.com/RoaringBitmap/roaring"
)
// writes out the length of the roaring bitmap in bytes as varint
// then writes out the roaring bitmap itself
func writeRoaringWithLen(r *roaring.Bitmap, w io.Writer,
reuseBufVarint []byte) (int, error) {
buf, err := r.ToBytes()
if err != nil {
return 0, err
}
var tw int
// write out the length
n := binary.PutUvarint(reuseBufVarint, uint64(len(buf)))
nw, err := w.Write(reuseBufVarint[:n])
tw += nw
if err != nil {
return tw, err
}
// write out the roaring bytes
nw, err = w.Write(buf)
tw += nw
if err != nil {
return tw, err
}
return tw, nil
}
func persistFields(fieldsInv []string, w *CountHashWriter, dictLocs []uint64) (uint64, error) {
var rv uint64
var fieldsOffsets []uint64
for fieldID, fieldName := range fieldsInv {
// record start of this field
fieldsOffsets = append(fieldsOffsets, uint64(w.Count()))
// write out the dict location and field name length
_, err := writeUvarints(w, dictLocs[fieldID], uint64(len(fieldName)))
if err != nil {
return 0, err
}
// write out the field name
_, err = w.Write([]byte(fieldName))
if err != nil {
return 0, err
}
}
// now write out the fields index
rv = uint64(w.Count())
for fieldID := range fieldsInv {
err := binary.Write(w, binary.BigEndian, fieldsOffsets[fieldID])
if err != nil {
return 0, err
}
}
return rv, nil
}
// FooterSize is the size of the footer record in bytes
// crc + ver + chunk + field offset + stored offset + num docs + docValueOffset
const FooterSize = 4 + 4 + 4 + 8 + 8 + 8 + 8
func persistFooter(numDocs, storedIndexOffset, fieldsIndexOffset, docValueOffset uint64,
chunkMode uint32, crcBeforeFooter uint32, writerIn io.Writer) error {
w := NewCountHashWriter(writerIn)
w.crc = crcBeforeFooter
// write out the number of docs
err := binary.Write(w, binary.BigEndian, numDocs)
if err != nil {
return err
}
// write out the stored field index location:
err = binary.Write(w, binary.BigEndian, storedIndexOffset)
if err != nil {
return err
}
// write out the field index location
err = binary.Write(w, binary.BigEndian, fieldsIndexOffset)
if err != nil {
return err
}
// write out the fieldDocValue location
err = binary.Write(w, binary.BigEndian, docValueOffset)
if err != nil {
return err
}
// write out 32-bit chunk factor
err = binary.Write(w, binary.BigEndian, chunkMode)
if err != nil {
return err
}
// write out 32-bit version
err = binary.Write(w, binary.BigEndian, Version)
if err != nil {
return err
}
// write out CRC-32 of everything upto but not including this CRC
err = binary.Write(w, binary.BigEndian, w.crc)
if err != nil {
return err
}
return nil
}
func writeUvarints(w io.Writer, vals ...uint64) (tw int, err error) {
buf := make([]byte, binary.MaxVarintLen64)
for _, val := range vals {
n := binary.PutUvarint(buf, val)
var nw int
nw, err = w.Write(buf[:n])
tw += nw
if err != nil {
return tw, err
}
}
return tw, err
}

@ -0,0 +1,177 @@
# ZAP File Format
## Legend
### Sections
|========|
| | section
|========|
### Fixed-size fields
|--------| |----| |--| |-|
| | uint64 | | uint32 | | uint16 | | uint8
|--------| |----| |--| |-|
### Varints
|~~~~~~~~|
| | varint(up to uint64)
|~~~~~~~~|
### Arbitrary-length fields
|--------...---|
| | arbitrary-length field (string, vellum, roaring bitmap)
|--------...---|
### Chunked data
[--------]
[ ]
[--------]
## Overview
Footer section describes the configuration of particular ZAP file. The format of footer is version-dependent, so it is necessary to check `V` field before the parsing.
|==================================================|
| Stored Fields |
|==================================================|
|-----> | Stored Fields Index |
| |==================================================|
| | Dictionaries + Postings + DocValues |
| |==================================================|
| |---> | DocValues Index |
| | |==================================================|
| | | Fields |
| | |==================================================|
| | |-> | Fields Index |
| | | |========|========|========|========|====|====|====|
| | | | D# | SF | F | FDV | CF | V | CC | (Footer)
| | | |========|====|===|====|===|====|===|====|====|====|
| | | | | |
|-+-+-----------------| | |
| |--------------------------| |
|-------------------------------------|
D#. Number of Docs.
SF. Stored Fields Index Offset.
F. Field Index Offset.
FDV. Field DocValue Offset.
CF. Chunk Factor.
V. Version.
CC. CRC32.
## Stored Fields
Stored Fields Index is `D#` consecutive 64-bit unsigned integers - offsets, where relevant Stored Fields Data records are located.
0 [SF] [SF + D# * 8]
| Stored Fields | Stored Fields Index |
|================================|==================================|
| | |
| |--------------------| ||--------|--------|. . .|--------||
| |-> | Stored Fields Data | || 0 | 1 | | D# - 1 ||
| | |--------------------| ||--------|----|---|. . .|--------||
| | | | |
|===|============================|==============|===================|
| |
|-------------------------------------------|
Stored Fields Data is an arbitrary size record, which consists of metadata and [Snappy](https://github.com/golang/snappy)-compressed data.
Stored Fields Data
|~~~~~~~~|~~~~~~~~|~~~~~~~~...~~~~~~~~|~~~~~~~~...~~~~~~~~|
| MDS | CDS | MD | CD |
|~~~~~~~~|~~~~~~~~|~~~~~~~~...~~~~~~~~|~~~~~~~~...~~~~~~~~|
MDS. Metadata size.
CDS. Compressed data size.
MD. Metadata.
CD. Snappy-compressed data.
## Fields
Fields Index section located between addresses `F` and `len(file) - len(footer)` and consist of `uint64` values (`F1`, `F2`, ...) which are offsets to records in Fields section. We have `F# = (len(file) - len(footer) - F) / sizeof(uint64)` fields.
(...) [F] [F + F#]
| Fields | Fields Index. |
|================================|================================|
| | |
| |~~~~~~~~|~~~~~~~~|---...---|||--------|--------|...|--------||
||->| Dict | Length | Name ||| 0 | 1 | | F# - 1 ||
|| |~~~~~~~~|~~~~~~~~|---...---|||--------|----|---|...|--------||
|| | | |
||===============================|==============|=================|
| |
|----------------------------------------------|
## Dictionaries + Postings
Each of fields has its own dictionary, encoded in [Vellum](https://github.com/couchbase/vellum) format. Dictionary consists of pairs `(term, offset)`, where `offset` indicates the position of postings (list of documents) for this particular term.
|================================================================|- Dictionaries +
| | Postings +
| | DocValues
| Freq/Norm (chunked) |
| [~~~~~~|~~~~~~~~~~~~~~~~~~~~~~~~~~~~~] |
| |->[ Freq | Norm (float32 under varint) ] |
| | [~~~~~~|~~~~~~~~~~~~~~~~~~~~~~~~~~~~~] |
| | |
| |------------------------------------------------------------| |
| Location Details (chunked) | |
| [~~~~~~|~~~~~|~~~~~~~|~~~~~|~~~~~~|~~~~~~~~|~~~~~] | |
| |->[ Size | Pos | Start | End | Arr# | ArrPos | ... ] | |
| | [~~~~~~|~~~~~|~~~~~~~|~~~~~|~~~~~~|~~~~~~~~|~~~~~] | |
| | | |
| |----------------------| | |
| Postings List | | |
| |~~~~~~~~|~~~~~|~~|~~~~~~~~|-----------...--| | |
| |->| F/N | LD | Length | ROARING BITMAP | | |
| | |~~~~~|~~|~~~~~~~~|~~~~~~~~|-----------...--| | |
| | |----------------------------------------------| |
| |--------------------------------------| |
| Dictionary | |
| |~~~~~~~~|--------------------------|-...-| |
| |->| Length | VELLUM DATA : (TERM -> OFFSET) | |
| | |~~~~~~~~|----------------------------...-| |
| | |
|======|=========================================================|- DocValues Index
| | |
|======|=========================================================|- Fields
| | |
| |~~~~|~~~|~~~~~~~~|---...---| |
| | Dict | Length | Name | |
| |~~~~~~~~|~~~~~~~~|---...---| |
| |
|================================================================|
## DocValues
DocValues Index is `F#` pairs of varints, one pair per field. Each pair of varints indicates start and end point of DocValues slice.
|================================================================|
| |------...--| |
| |->| DocValues |<-| |
| | |------...--| | |
|==|=================|===========================================|- DocValues Index
||~|~~~~~~~~~|~~~~~~~|~~| |~~~~~~~~~~~~~~|~~~~~~~~~~~~||
|| DV1 START | DV1 STOP | . . . . . | DV(F#) START | DV(F#) END ||
||~~~~~~~~~~~|~~~~~~~~~~| |~~~~~~~~~~~~~~|~~~~~~~~~~~~||
|================================================================|
DocValues is chunked Snappy-compressed values for each document and field.
[~~~~~~~~~~~~~~~|~~~~~~|~~~~~~~~~|-...-|~~~~~~|~~~~~~~~~|--------------------...-]
[ Doc# in Chunk | Doc1 | Offset1 | ... | DocN | OffsetN | SNAPPY COMPRESSED DATA ]
[~~~~~~~~~~~~~~~|~~~~~~|~~~~~~~~~|-...-|~~~~~~|~~~~~~~~~|--------------------...-]
Last 16 bytes are description of chunks.
|~~~~~~~~~~~~...~|----------------|----------------|
| Chunk Sizes | Chunk Size Arr | Chunk# |
|~~~~~~~~~~~~...~|----------------|----------------|

@ -3,7 +3,7 @@ module github.com/couchbase/vellum
go 1.12
require (
github.com/edsrzf/mmap-go v1.0.0
github.com/blevesearch/mmap-go v1.0.2
github.com/spf13/cobra v0.0.5
github.com/willf/bitset v1.1.10
golang.org/x/sys v0.0.0-20190813064441-fde4db37ae7a // indirect

@ -1,12 +1,12 @@
github.com/BurntSushi/toml v0.3.1/go.mod h1:xHWCNGjB5oqiDr8zfno3MHue2Ht5sIBksp03qcyfWMU=
github.com/armon/consul-api v0.0.0-20180202201655-eb2c6b5be1b6/go.mod h1:grANhF5doyWs3UAsr3K4I6qtAmlQcZDesFNEHPZAzj8=
github.com/blevesearch/mmap-go v1.0.2 h1:JtMHb+FgQCTTYIhtMvimw15dJwu1Y5lrZDMOFXVWPk0=
github.com/blevesearch/mmap-go v1.0.2/go.mod h1:ol2qBqYaOUsGdm7aRMRrYGgPvnwLe6Y+7LMvAB5IbSA=
github.com/coreos/etcd v3.3.10+incompatible/go.mod h1:uF7uidLiAD3TWHmW31ZFd/JWoc32PjwdhPthX9715RE=
github.com/coreos/go-etcd v2.0.0+incompatible/go.mod h1:Jez6KQU2B/sWsbdaef3ED8NzMklzPG4d5KIOhIy30Tk=
github.com/coreos/go-semver v0.2.0/go.mod h1:nnelYz7RCh+5ahJtPPxZlU+153eP4D4r3EedlOD2RNk=
github.com/cpuguy83/go-md2man v1.0.10/go.mod h1:SmD6nW6nTyfqj6ABTjUi3V3JVMnlJmwcJI5acqYI6dE=
github.com/davecgh/go-spew v1.1.1/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
github.com/edsrzf/mmap-go v1.0.0 h1:CEBF7HpRnUCSJgGUb5h1Gm7e3VkmVDrR8lvWVLtrOFw=
github.com/edsrzf/mmap-go v1.0.0/go.mod h1:YO35OhQPt3KJa3ryjFM5Bs14WD66h8eGKpfaBNrHW5M=
github.com/fsnotify/fsnotify v1.4.7/go.mod h1:jwhsz4b93w/PPRr/qN1Yymfu8t87LnFCMoQvtojpjFo=
github.com/hashicorp/hcl v1.0.0/go.mod h1:E5yfLk+7swimpb2L/Alb/PJmXilQ/rhwaUYs4T20WEQ=
github.com/inconshreveable/mousetrap v1.0.0 h1:Z8tu5sraLXCXIcARxBp/8cbvlwVa7Z1NHg9XEKhtSvM=
@ -32,6 +32,7 @@ github.com/willf/bitset v1.1.10/go.mod h1:RjeCKbqT1RxIR/KWY6phxZiaY1IyutSBfGjNPy
github.com/xordataexchange/crypt v0.0.3-0.20170626215501-b2862e3d0a77/go.mod h1:aYKd//L2LvnjZzWKhF00oedf4jCCReLcmhLdhm1A27Q=
golang.org/x/crypto v0.0.0-20181203042331-505ab145d0a9/go.mod h1:6SG95UA2DQfeDnfUPMdvaQW0Q7yPrPDi9nlGo2tz2b4=
golang.org/x/sys v0.0.0-20181205085412-a5c9d58dba9a/go.mod h1:STP8DvDyc/dI5b8T5hshtkjS+E42TnysNCUPdjciGhY=
golang.org/x/sys v0.0.0-20181221143128-b4a75ba826a6/go.mod h1:STP8DvDyc/dI5b8T5hshtkjS+E42TnysNCUPdjciGhY=
golang.org/x/sys v0.0.0-20190813064441-fde4db37ae7a h1:aYOabOQFp6Vj6W1F80affTUvO9UxmJRx8K0gsfABByQ=
golang.org/x/sys v0.0.0-20190813064441-fde4db37ae7a/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
golang.org/x/text v0.3.0/go.mod h1:NqM8EUOU14njkJ3fqMW+pc6Ldnwhi/IjpwHt7yyuwOQ=

@ -19,7 +19,7 @@ package vellum
import (
"os"
mmap "github.com/edsrzf/mmap-go"
mmap "github.com/blevesearch/mmap-go"
)
type mmapWrapper struct {

@ -1,28 +0,0 @@
package bbolt
import "unsafe"
// maxMapSize represents the largest mmap size supported by Bolt.
const maxMapSize = 0x7FFFFFFF // 2GB
// maxAllocSize is the size used when creating array pointers.
const maxAllocSize = 0xFFFFFFF
// Are unaligned load/stores broken on this arch?
var brokenUnaligned bool
func init() {
// Simple check to see whether this arch handles unaligned load/stores
// correctly.
// ARM9 and older devices require load/stores to be from/to aligned
// addresses. If not, the lower 2 bits are cleared and that address is
// read in a jumbled up order.
// See http://infocenter.arm.com/help/index.jsp?topic=/com.arm.doc.faqs/ka15414.html
raw := [6]byte{0xfe, 0xef, 0x11, 0x22, 0x22, 0x11}
val := *(*uint32)(unsafe.Pointer(uintptr(unsafe.Pointer(&raw)) + 2))
brokenUnaligned = val != 0x11222211
}

Some files were not shown because too many files have changed in this diff Show More

Loading…
Cancel
Save