Upgrade server dependencies, manage them with govendor

This commit is contained in:
Ken-Håvard Lieng 2017-04-18 03:02:51 +02:00
parent ebee2746d6
commit 971278e7e5
1748 changed files with 196165 additions and 194500 deletions

View file

@ -0,0 +1,35 @@
// Copyright (c) 2014 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 searcher
import (
"github.com/blevesearch/bleve/search"
)
type OrderedSearcherList []search.Searcher
// sort.Interface
func (otrl OrderedSearcherList) Len() int {
return len(otrl)
}
func (otrl OrderedSearcherList) Less(i, j int) bool {
return otrl[i].Count() < otrl[j].Count()
}
func (otrl OrderedSearcherList) Swap(i, j int) {
otrl[i], otrl[j] = otrl[j], otrl[i]
}

View file

@ -0,0 +1,391 @@
// Copyright (c) 2014 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 searcher
import (
"math"
"github.com/blevesearch/bleve/index"
"github.com/blevesearch/bleve/search"
"github.com/blevesearch/bleve/search/scorer"
)
type BooleanSearcher struct {
indexReader index.IndexReader
mustSearcher search.Searcher
shouldSearcher search.Searcher
mustNotSearcher search.Searcher
queryNorm float64
currMust *search.DocumentMatch
currShould *search.DocumentMatch
currMustNot *search.DocumentMatch
currentID index.IndexInternalID
min uint64
scorer *scorer.ConjunctionQueryScorer
matches []*search.DocumentMatch
initialized bool
}
func NewBooleanSearcher(indexReader index.IndexReader, mustSearcher search.Searcher, shouldSearcher search.Searcher, mustNotSearcher search.Searcher, options search.SearcherOptions) (*BooleanSearcher, error) {
// build our searcher
rv := BooleanSearcher{
indexReader: indexReader,
mustSearcher: mustSearcher,
shouldSearcher: shouldSearcher,
mustNotSearcher: mustNotSearcher,
scorer: scorer.NewConjunctionQueryScorer(options),
matches: make([]*search.DocumentMatch, 2),
}
rv.computeQueryNorm()
return &rv, nil
}
func (s *BooleanSearcher) computeQueryNorm() {
// first calculate sum of squared weights
sumOfSquaredWeights := 0.0
if s.mustSearcher != nil {
sumOfSquaredWeights += s.mustSearcher.Weight()
}
if s.shouldSearcher != nil {
sumOfSquaredWeights += s.shouldSearcher.Weight()
}
// now compute query norm from this
s.queryNorm = 1.0 / math.Sqrt(sumOfSquaredWeights)
// finally tell all the downstream searchers the norm
if s.mustSearcher != nil {
s.mustSearcher.SetQueryNorm(s.queryNorm)
}
if s.shouldSearcher != nil {
s.shouldSearcher.SetQueryNorm(s.queryNorm)
}
}
func (s *BooleanSearcher) initSearchers(ctx *search.SearchContext) error {
var err error
// get all searchers pointing at their first match
if s.mustSearcher != nil {
if s.currMust != nil {
ctx.DocumentMatchPool.Put(s.currMust)
}
s.currMust, err = s.mustSearcher.Next(ctx)
if err != nil {
return err
}
}
if s.shouldSearcher != nil {
if s.currShould != nil {
ctx.DocumentMatchPool.Put(s.currShould)
}
s.currShould, err = s.shouldSearcher.Next(ctx)
if err != nil {
return err
}
}
if s.mustNotSearcher != nil {
if s.currMustNot != nil {
ctx.DocumentMatchPool.Put(s.currMustNot)
}
s.currMustNot, err = s.mustNotSearcher.Next(ctx)
if err != nil {
return err
}
}
if s.mustSearcher != nil && s.currMust != nil {
s.currentID = s.currMust.IndexInternalID
} else if s.mustSearcher == nil && s.currShould != nil {
s.currentID = s.currShould.IndexInternalID
} else {
s.currentID = nil
}
s.initialized = true
return nil
}
func (s *BooleanSearcher) advanceNextMust(ctx *search.SearchContext, skipReturn *search.DocumentMatch) error {
var err error
if s.mustSearcher != nil {
if s.currMust != skipReturn {
ctx.DocumentMatchPool.Put(s.currMust)
}
s.currMust, err = s.mustSearcher.Next(ctx)
if err != nil {
return err
}
} else {
if s.currShould != skipReturn {
ctx.DocumentMatchPool.Put(s.currShould)
}
s.currShould, err = s.shouldSearcher.Next(ctx)
if err != nil {
return err
}
}
if s.mustSearcher != nil && s.currMust != nil {
s.currentID = s.currMust.IndexInternalID
} else if s.mustSearcher == nil && s.currShould != nil {
s.currentID = s.currShould.IndexInternalID
} else {
s.currentID = nil
}
return nil
}
func (s *BooleanSearcher) Weight() float64 {
var rv float64
if s.mustSearcher != nil {
rv += s.mustSearcher.Weight()
}
if s.shouldSearcher != nil {
rv += s.shouldSearcher.Weight()
}
return rv
}
func (s *BooleanSearcher) SetQueryNorm(qnorm float64) {
if s.mustSearcher != nil {
s.mustSearcher.SetQueryNorm(qnorm)
}
if s.shouldSearcher != nil {
s.shouldSearcher.SetQueryNorm(qnorm)
}
}
func (s *BooleanSearcher) Next(ctx *search.SearchContext) (*search.DocumentMatch, error) {
if !s.initialized {
err := s.initSearchers(ctx)
if err != nil {
return nil, err
}
}
var err error
var rv *search.DocumentMatch
for s.currentID != nil {
if s.currMustNot != nil {
cmp := s.currMustNot.IndexInternalID.Compare(s.currentID)
if cmp < 0 {
ctx.DocumentMatchPool.Put(s.currMustNot)
// advance must not searcher to our candidate entry
s.currMustNot, err = s.mustNotSearcher.Advance(ctx, s.currentID)
if err != nil {
return nil, err
}
if s.currMustNot != nil && s.currMustNot.IndexInternalID.Equals(s.currentID) {
// the candidate is excluded
err = s.advanceNextMust(ctx, nil)
if err != nil {
return nil, err
}
continue
}
} else if cmp == 0 {
// the candidate is excluded
err = s.advanceNextMust(ctx, nil)
if err != nil {
return nil, err
}
continue
}
}
shouldCmpOrNil := 1 // NOTE: shouldCmp will also be 1 when currShould == nil.
if s.currShould != nil {
shouldCmpOrNil = s.currShould.IndexInternalID.Compare(s.currentID)
}
if shouldCmpOrNil < 0 {
ctx.DocumentMatchPool.Put(s.currShould)
// advance should searcher to our candidate entry
s.currShould, err = s.shouldSearcher.Advance(ctx, s.currentID)
if err != nil {
return nil, err
}
if s.currShould != nil && s.currShould.IndexInternalID.Equals(s.currentID) {
// score bonus matches should
var cons []*search.DocumentMatch
if s.currMust != nil {
cons = s.matches
cons[0] = s.currMust
cons[1] = s.currShould
} else {
cons = s.matches[0:1]
cons[0] = s.currShould
}
rv = s.scorer.Score(ctx, cons)
err = s.advanceNextMust(ctx, rv)
if err != nil {
return nil, err
}
break
} else if s.shouldSearcher.Min() == 0 {
// match is OK anyway
cons := s.matches[0:1]
cons[0] = s.currMust
rv = s.scorer.Score(ctx, cons)
err = s.advanceNextMust(ctx, rv)
if err != nil {
return nil, err
}
break
}
} else if shouldCmpOrNil == 0 {
// score bonus matches should
var cons []*search.DocumentMatch
if s.currMust != nil {
cons = s.matches
cons[0] = s.currMust
cons[1] = s.currShould
} else {
cons = s.matches[0:1]
cons[0] = s.currShould
}
rv = s.scorer.Score(ctx, cons)
err = s.advanceNextMust(ctx, rv)
if err != nil {
return nil, err
}
break
} else if s.shouldSearcher == nil || s.shouldSearcher.Min() == 0 {
// match is OK anyway
cons := s.matches[0:1]
cons[0] = s.currMust
rv = s.scorer.Score(ctx, cons)
err = s.advanceNextMust(ctx, rv)
if err != nil {
return nil, err
}
break
}
err = s.advanceNextMust(ctx, nil)
if err != nil {
return nil, err
}
}
return rv, nil
}
func (s *BooleanSearcher) Advance(ctx *search.SearchContext, ID index.IndexInternalID) (*search.DocumentMatch, error) {
if !s.initialized {
err := s.initSearchers(ctx)
if err != nil {
return nil, err
}
}
var err error
if s.mustSearcher != nil {
if s.currMust != nil {
ctx.DocumentMatchPool.Put(s.currMust)
}
s.currMust, err = s.mustSearcher.Advance(ctx, ID)
if err != nil {
return nil, err
}
}
if s.shouldSearcher != nil {
if s.currShould != nil {
ctx.DocumentMatchPool.Put(s.currShould)
}
s.currShould, err = s.shouldSearcher.Advance(ctx, ID)
if err != nil {
return nil, err
}
}
if s.mustNotSearcher != nil {
if s.currMustNot != nil {
ctx.DocumentMatchPool.Put(s.currMustNot)
}
s.currMustNot, err = s.mustNotSearcher.Advance(ctx, ID)
if err != nil {
return nil, err
}
}
if s.mustSearcher != nil && s.currMust != nil {
s.currentID = s.currMust.IndexInternalID
} else if s.mustSearcher == nil && s.currShould != nil {
s.currentID = s.currShould.IndexInternalID
} else {
s.currentID = nil
}
return s.Next(ctx)
}
func (s *BooleanSearcher) Count() uint64 {
// for now return a worst case
var sum uint64
if s.mustSearcher != nil {
sum += s.mustSearcher.Count()
}
if s.shouldSearcher != nil {
sum += s.shouldSearcher.Count()
}
return sum
}
func (s *BooleanSearcher) Close() error {
var err0, err1, err2 error
if s.mustSearcher != nil {
err0 = s.mustSearcher.Close()
}
if s.shouldSearcher != nil {
err1 = s.shouldSearcher.Close()
}
if s.mustNotSearcher != nil {
err2 = s.mustNotSearcher.Close()
}
if err0 != nil {
return err0
}
if err1 != nil {
return err1
}
if err2 != nil {
return err2
}
return nil
}
func (s *BooleanSearcher) Min() int {
return 0
}
func (s *BooleanSearcher) DocumentMatchPoolSize() int {
rv := 3
if s.mustSearcher != nil {
rv += s.mustSearcher.DocumentMatchPoolSize()
}
if s.shouldSearcher != nil {
rv += s.shouldSearcher.DocumentMatchPoolSize()
}
if s.mustNotSearcher != nil {
rv += s.mustNotSearcher.DocumentMatchPoolSize()
}
return rv
}

View file

@ -0,0 +1,232 @@
// Copyright (c) 2014 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 searcher
import (
"math"
"sort"
"github.com/blevesearch/bleve/index"
"github.com/blevesearch/bleve/search"
"github.com/blevesearch/bleve/search/scorer"
)
type ConjunctionSearcher struct {
indexReader index.IndexReader
searchers OrderedSearcherList
queryNorm float64
currs []*search.DocumentMatch
maxIDIdx int
scorer *scorer.ConjunctionQueryScorer
initialized bool
options search.SearcherOptions
}
func NewConjunctionSearcher(indexReader index.IndexReader, qsearchers []search.Searcher, options search.SearcherOptions) (*ConjunctionSearcher, error) {
// build the downstream searchers
searchers := make(OrderedSearcherList, len(qsearchers))
for i, searcher := range qsearchers {
searchers[i] = searcher
}
// sort the searchers
sort.Sort(searchers)
// build our searcher
rv := ConjunctionSearcher{
indexReader: indexReader,
options: options,
searchers: searchers,
currs: make([]*search.DocumentMatch, len(searchers)),
scorer: scorer.NewConjunctionQueryScorer(options),
}
rv.computeQueryNorm()
return &rv, nil
}
func (s *ConjunctionSearcher) computeQueryNorm() {
// first calculate sum of squared weights
sumOfSquaredWeights := 0.0
for _, termSearcher := range s.searchers {
sumOfSquaredWeights += termSearcher.Weight()
}
// now compute query norm from this
s.queryNorm = 1.0 / math.Sqrt(sumOfSquaredWeights)
// finally tell all the downstream searchers the norm
for _, termSearcher := range s.searchers {
termSearcher.SetQueryNorm(s.queryNorm)
}
}
func (s *ConjunctionSearcher) initSearchers(ctx *search.SearchContext) error {
var err error
// get all searchers pointing at their first match
for i, termSearcher := range s.searchers {
if s.currs[i] != nil {
ctx.DocumentMatchPool.Put(s.currs[i])
}
s.currs[i], err = termSearcher.Next(ctx)
if err != nil {
return err
}
}
s.initialized = true
return nil
}
func (s *ConjunctionSearcher) Weight() float64 {
var rv float64
for _, searcher := range s.searchers {
rv += searcher.Weight()
}
return rv
}
func (s *ConjunctionSearcher) SetQueryNorm(qnorm float64) {
for _, searcher := range s.searchers {
searcher.SetQueryNorm(qnorm)
}
}
func (s *ConjunctionSearcher) Next(ctx *search.SearchContext) (*search.DocumentMatch, error) {
if !s.initialized {
err := s.initSearchers(ctx)
if err != nil {
return nil, err
}
}
var rv *search.DocumentMatch
var err error
OUTER:
for s.currs[s.maxIDIdx] != nil {
maxID := s.currs[s.maxIDIdx].IndexInternalID
i := 0
for i < len(s.currs) {
if s.currs[i] == nil {
return nil, nil
}
if i == s.maxIDIdx {
i++
continue
}
cmp := maxID.Compare(s.currs[i].IndexInternalID)
if cmp == 0 {
i++
continue
}
if cmp < 0 {
// maxID < currs[i], so we found a new maxIDIdx
s.maxIDIdx = i
// advance the positions where [0 <= x < i], since we
// know they were equal to the former max entry
maxID = s.currs[s.maxIDIdx].IndexInternalID
for x := 0; x < i; x++ {
err = s.advanceChild(ctx, x, maxID)
if err != nil {
return nil, err
}
}
continue OUTER
}
// maxID > currs[i], so need to advance searchers[i]
err = s.advanceChild(ctx, i, maxID)
if err != nil {
return nil, err
}
// don't bump i, so that we'll examine the just-advanced
// currs[i] again
}
// if we get here, a doc matched all readers, so score and add it
rv = s.scorer.Score(ctx, s.currs)
// we know all the searchers are pointing at the same thing
// so they all need to be bumped
for i, termSearcher := range s.searchers {
if s.currs[i] != rv {
ctx.DocumentMatchPool.Put(s.currs[i])
}
s.currs[i], err = termSearcher.Next(ctx)
if err != nil {
return nil, err
}
}
// don't continue now, wait for the next call to Next()
break
}
return rv, nil
}
func (s *ConjunctionSearcher) Advance(ctx *search.SearchContext, ID index.IndexInternalID) (*search.DocumentMatch, error) {
if !s.initialized {
err := s.initSearchers(ctx)
if err != nil {
return nil, err
}
}
for i := range s.searchers {
err := s.advanceChild(ctx, i, ID)
if err != nil {
return nil, err
}
}
return s.Next(ctx)
}
func (s *ConjunctionSearcher) advanceChild(ctx *search.SearchContext, i int, ID index.IndexInternalID) (err error) {
if s.currs[i] != nil {
ctx.DocumentMatchPool.Put(s.currs[i])
}
s.currs[i], err = s.searchers[i].Advance(ctx, ID)
return err
}
func (s *ConjunctionSearcher) Count() uint64 {
// for now return a worst case
var sum uint64
for _, searcher := range s.searchers {
sum += searcher.Count()
}
return sum
}
func (s *ConjunctionSearcher) Close() (rv error) {
for _, searcher := range s.searchers {
err := searcher.Close()
if err != nil && rv == nil {
rv = err
}
}
return rv
}
func (s *ConjunctionSearcher) Min() int {
return 0
}
func (s *ConjunctionSearcher) DocumentMatchPoolSize() int {
rv := len(s.currs)
for _, s := range s.searchers {
rv += s.DocumentMatchPoolSize()
}
return rv
}

View file

@ -0,0 +1,271 @@
// Copyright (c) 2014 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 searcher
import (
"fmt"
"math"
"sort"
"github.com/blevesearch/bleve/index"
"github.com/blevesearch/bleve/search"
"github.com/blevesearch/bleve/search/scorer"
)
// DisjunctionMaxClauseCount is a compile time setting that applications can
// adjust to non-zero value to cause the DisjunctionSearcher to return an
// error instead of exeucting searches when the size exceeds this value.
var DisjunctionMaxClauseCount = 0
type DisjunctionSearcher struct {
indexReader index.IndexReader
searchers OrderedSearcherList
numSearchers int
queryNorm float64
currs []*search.DocumentMatch
scorer *scorer.DisjunctionQueryScorer
min int
matching []*search.DocumentMatch
matchingIdxs []int
initialized bool
}
func tooManyClauses(count int) bool {
if DisjunctionMaxClauseCount != 0 && count > DisjunctionMaxClauseCount {
return true
}
return false
}
func tooManyClausesErr() error {
return fmt.Errorf("TooManyClauses[maxClauseCount is set to %d]",
DisjunctionMaxClauseCount)
}
func NewDisjunctionSearcher(indexReader index.IndexReader,
qsearchers []search.Searcher, min float64, options search.SearcherOptions) (
*DisjunctionSearcher, error) {
return newDisjunctionSearcher(indexReader, qsearchers, min, options,
true)
}
func newDisjunctionSearcher(indexReader index.IndexReader,
qsearchers []search.Searcher, min float64, options search.SearcherOptions,
limit bool) (
*DisjunctionSearcher, error) {
if limit && tooManyClauses(len(qsearchers)) {
return nil, tooManyClausesErr()
}
// build the downstream searchers
searchers := make(OrderedSearcherList, len(qsearchers))
for i, searcher := range qsearchers {
searchers[i] = searcher
}
// sort the searchers
sort.Sort(sort.Reverse(searchers))
// build our searcher
rv := DisjunctionSearcher{
indexReader: indexReader,
searchers: searchers,
numSearchers: len(searchers),
currs: make([]*search.DocumentMatch, len(searchers)),
scorer: scorer.NewDisjunctionQueryScorer(options),
min: int(min),
matching: make([]*search.DocumentMatch, len(searchers)),
matchingIdxs: make([]int, len(searchers)),
}
rv.computeQueryNorm()
return &rv, nil
}
func (s *DisjunctionSearcher) computeQueryNorm() {
// first calculate sum of squared weights
sumOfSquaredWeights := 0.0
for _, termSearcher := range s.searchers {
sumOfSquaredWeights += termSearcher.Weight()
}
// now compute query norm from this
s.queryNorm = 1.0 / math.Sqrt(sumOfSquaredWeights)
// finally tell all the downstream searchers the norm
for _, termSearcher := range s.searchers {
termSearcher.SetQueryNorm(s.queryNorm)
}
}
func (s *DisjunctionSearcher) initSearchers(ctx *search.SearchContext) error {
var err error
// get all searchers pointing at their first match
for i, termSearcher := range s.searchers {
if s.currs[i] != nil {
ctx.DocumentMatchPool.Put(s.currs[i])
}
s.currs[i], err = termSearcher.Next(ctx)
if err != nil {
return err
}
}
err = s.updateMatches()
if err != nil {
return err
}
s.initialized = true
return nil
}
func (s *DisjunctionSearcher) updateMatches() error {
matching := s.matching[:0]
matchingIdxs := s.matchingIdxs[:0]
for i := 0; i < len(s.currs); i++ {
curr := s.currs[i]
if curr == nil {
continue
}
if len(matching) > 0 {
cmp := curr.IndexInternalID.Compare(matching[0].IndexInternalID)
if cmp > 0 {
continue
}
if cmp < 0 {
matching = matching[:0]
matchingIdxs = matchingIdxs[:0]
}
}
matching = append(matching, curr)
matchingIdxs = append(matchingIdxs, i)
}
s.matching = matching
s.matchingIdxs = matchingIdxs
return nil
}
func (s *DisjunctionSearcher) Weight() float64 {
var rv float64
for _, searcher := range s.searchers {
rv += searcher.Weight()
}
return rv
}
func (s *DisjunctionSearcher) SetQueryNorm(qnorm float64) {
for _, searcher := range s.searchers {
searcher.SetQueryNorm(qnorm)
}
}
func (s *DisjunctionSearcher) Next(ctx *search.SearchContext) (
*search.DocumentMatch, error) {
if !s.initialized {
err := s.initSearchers(ctx)
if err != nil {
return nil, err
}
}
var err error
var rv *search.DocumentMatch
found := false
for !found && len(s.matching) > 0 {
if len(s.matching) >= s.min {
found = true
// score this match
rv = s.scorer.Score(ctx, s.matching, len(s.matching), s.numSearchers)
}
// invoke next on all the matching searchers
for _, i := range s.matchingIdxs {
searcher := s.searchers[i]
if s.currs[i] != rv {
ctx.DocumentMatchPool.Put(s.currs[i])
}
s.currs[i], err = searcher.Next(ctx)
if err != nil {
return nil, err
}
}
err = s.updateMatches()
if err != nil {
return nil, err
}
}
return rv, nil
}
func (s *DisjunctionSearcher) Advance(ctx *search.SearchContext,
ID index.IndexInternalID) (*search.DocumentMatch, error) {
if !s.initialized {
err := s.initSearchers(ctx)
if err != nil {
return nil, err
}
}
// get all searchers pointing at their first match
var err error
for i, termSearcher := range s.searchers {
if s.currs[i] != nil {
ctx.DocumentMatchPool.Put(s.currs[i])
}
s.currs[i], err = termSearcher.Advance(ctx, ID)
if err != nil {
return nil, err
}
}
err = s.updateMatches()
if err != nil {
return nil, err
}
return s.Next(ctx)
}
func (s *DisjunctionSearcher) Count() uint64 {
// for now return a worst case
var sum uint64
for _, searcher := range s.searchers {
sum += searcher.Count()
}
return sum
}
func (s *DisjunctionSearcher) Close() (rv error) {
for _, searcher := range s.searchers {
err := searcher.Close()
if err != nil && rv == nil {
rv = err
}
}
return rv
}
func (s *DisjunctionSearcher) Min() int {
return s.min
}
func (s *DisjunctionSearcher) DocumentMatchPoolSize() int {
rv := len(s.currs)
for _, s := range s.searchers {
rv += s.DocumentMatchPoolSize()
}
return rv
}

View file

@ -0,0 +1,93 @@
// Copyright (c) 2015 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 searcher
import (
"github.com/blevesearch/bleve/index"
"github.com/blevesearch/bleve/search"
"github.com/blevesearch/bleve/search/scorer"
)
// DocIDSearcher returns documents matching a predefined set of identifiers.
type DocIDSearcher struct {
reader index.DocIDReader
scorer *scorer.ConstantScorer
count int
}
func NewDocIDSearcher(indexReader index.IndexReader, ids []string, boost float64,
options search.SearcherOptions) (searcher *DocIDSearcher, err error) {
reader, err := indexReader.DocIDReaderOnly(ids)
if err != nil {
return nil, err
}
scorer := scorer.NewConstantScorer(1.0, boost, options)
return &DocIDSearcher{
scorer: scorer,
reader: reader,
count: len(ids),
}, nil
}
func (s *DocIDSearcher) Count() uint64 {
return uint64(s.count)
}
func (s *DocIDSearcher) Weight() float64 {
return s.scorer.Weight()
}
func (s *DocIDSearcher) SetQueryNorm(qnorm float64) {
s.scorer.SetQueryNorm(qnorm)
}
func (s *DocIDSearcher) Next(ctx *search.SearchContext) (*search.DocumentMatch, error) {
docidMatch, err := s.reader.Next()
if err != nil {
return nil, err
}
if docidMatch == nil {
return nil, nil
}
docMatch := s.scorer.Score(ctx, docidMatch)
return docMatch, nil
}
func (s *DocIDSearcher) Advance(ctx *search.SearchContext, ID index.IndexInternalID) (*search.DocumentMatch, error) {
docidMatch, err := s.reader.Advance(ID)
if err != nil {
return nil, err
}
if docidMatch == nil {
return nil, nil
}
docMatch := s.scorer.Score(ctx, docidMatch)
return docMatch, nil
}
func (s *DocIDSearcher) Close() error {
return s.reader.Close()
}
func (s *DocIDSearcher) Min() int {
return 0
}
func (s *DocIDSearcher) DocumentMatchPoolSize() int {
return 1
}

View file

@ -0,0 +1,88 @@
// 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 searcher
import (
"github.com/blevesearch/bleve/index"
"github.com/blevesearch/bleve/search"
)
// FilterFunc defines a function which can filter documents
// returning true means keep the document
// returning false means do not keep the document
type FilterFunc func(d *search.DocumentMatch) bool
// FilteringSearcher wraps any other searcher, but checks any Next/Advance
// call against the supplied FilterFunc
type FilteringSearcher struct {
child search.Searcher
accept FilterFunc
}
func NewFilteringSearcher(s search.Searcher, filter FilterFunc) *FilteringSearcher {
return &FilteringSearcher{
child: s,
accept: filter,
}
}
func (f *FilteringSearcher) Next(ctx *search.SearchContext) (*search.DocumentMatch, error) {
next, err := f.child.Next(ctx)
for next != nil && err == nil {
if f.accept(next) {
return next, nil
}
next, err = f.child.Next(ctx)
}
return nil, err
}
func (f *FilteringSearcher) Advance(ctx *search.SearchContext, ID index.IndexInternalID) (*search.DocumentMatch, error) {
adv, err := f.child.Advance(ctx, ID)
if err != nil {
return nil, err
}
if adv == nil {
return nil, nil
}
if f.accept(adv) {
return adv, nil
}
return f.Next(ctx)
}
func (f *FilteringSearcher) Close() error {
return f.child.Close()
}
func (f *FilteringSearcher) Weight() float64 {
return f.child.Weight()
}
func (f *FilteringSearcher) SetQueryNorm(n float64) {
f.child.SetQueryNorm(n)
}
func (f *FilteringSearcher) Count() uint64 {
return f.child.Count()
}
func (f *FilteringSearcher) Min() int {
return f.child.Min()
}
func (f *FilteringSearcher) DocumentMatchPoolSize() int {
return f.child.DocumentMatchPoolSize()
}

View file

@ -0,0 +1,74 @@
// Copyright (c) 2014 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 searcher
import (
"github.com/blevesearch/bleve/index"
"github.com/blevesearch/bleve/search"
)
func NewFuzzySearcher(indexReader index.IndexReader, term string,
prefix, fuzziness int, field string, boost float64,
options search.SearcherOptions) (search.Searcher, error) {
// Note: we don't byte slice the term for a prefix because of runes.
prefixTerm := ""
for i, r := range term {
if i < prefix {
prefixTerm += string(r)
} else {
break
}
}
candidateTerms, err := findFuzzyCandidateTerms(indexReader, term, fuzziness,
field, prefixTerm)
if err != nil {
return nil, err
}
return NewMultiTermSearcher(indexReader, candidateTerms, field,
boost, options, true)
}
func findFuzzyCandidateTerms(indexReader index.IndexReader, term string,
fuzziness int, field, prefixTerm string) (rv []string, err error) {
rv = make([]string, 0)
var fieldDict index.FieldDict
if len(prefixTerm) > 0 {
fieldDict, err = indexReader.FieldDictPrefix(field, []byte(prefixTerm))
} else {
fieldDict, err = indexReader.FieldDict(field)
}
defer func() {
if cerr := fieldDict.Close(); cerr != nil && err == nil {
err = cerr
}
}()
// enumerate terms and check levenshtein distance
tfd, err := fieldDict.Next()
for err == nil && tfd != nil {
ld, exceeded := search.LevenshteinDistanceMax(term, tfd.Term, fuzziness)
if !exceeded && ld <= fuzziness {
rv = append(rv, tfd.Term)
if tooManyClauses(len(rv)) {
return rv, tooManyClausesErr()
}
}
tfd, err = fieldDict.Next()
}
return rv, err
}

View file

@ -0,0 +1,173 @@
// 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 searcher
import (
"github.com/blevesearch/bleve/document"
"github.com/blevesearch/bleve/geo"
"github.com/blevesearch/bleve/index"
"github.com/blevesearch/bleve/numeric"
"github.com/blevesearch/bleve/search"
)
func NewGeoBoundingBoxSearcher(indexReader index.IndexReader, minLon, minLat,
maxLon, maxLat float64, field string, boost float64,
options search.SearcherOptions, checkBoundaries bool) (
search.Searcher, error) {
// track list of opened searchers, for cleanup on early exit
var openedSearchers []search.Searcher
cleanupOpenedSearchers := func() {
for _, s := range openedSearchers {
_ = s.Close()
}
}
// do math to produce list of terms needed for this search
onBoundaryTerms, notOnBoundaryTerms := ComputeGeoRange(0, (geo.GeoBits<<1)-1,
minLon, minLat, maxLon, maxLat, checkBoundaries)
var onBoundarySearcher search.Searcher
if len(onBoundaryTerms) > 0 {
rawOnBoundarySearcher, err := NewMultiTermSearcherBytes(indexReader,
onBoundaryTerms, field, boost, options, false)
if err != nil {
return nil, err
}
// add filter to check points near the boundary
onBoundarySearcher = NewFilteringSearcher(rawOnBoundarySearcher,
buildRectFilter(indexReader, field, minLon, minLat, maxLon, maxLat))
openedSearchers = append(openedSearchers, onBoundarySearcher)
}
var notOnBoundarySearcher search.Searcher
if len(notOnBoundaryTerms) > 0 {
var err error
notOnBoundarySearcher, err = NewMultiTermSearcherBytes(indexReader,
notOnBoundaryTerms, field, boost, options, false)
if err != nil {
cleanupOpenedSearchers()
return nil, err
}
openedSearchers = append(openedSearchers, notOnBoundarySearcher)
}
if onBoundarySearcher != nil && notOnBoundarySearcher != nil {
rv, err := NewDisjunctionSearcher(indexReader,
[]search.Searcher{
onBoundarySearcher,
notOnBoundarySearcher,
},
0, options)
if err != nil {
cleanupOpenedSearchers()
return nil, err
}
return rv, nil
} else if onBoundarySearcher != nil {
return onBoundarySearcher, nil
} else if notOnBoundarySearcher != nil {
return notOnBoundarySearcher, nil
}
return NewMatchNoneSearcher(indexReader)
}
var geoMaxShift = document.GeoPrecisionStep * 4
var geoDetailLevel = ((geo.GeoBits << 1) - geoMaxShift) / 2
func ComputeGeoRange(term uint64, shift uint,
sminLon, sminLat, smaxLon, smaxLat float64,
checkBoundaries bool) (
onBoundary [][]byte, notOnBoundary [][]byte) {
split := term | uint64(0x1)<<shift
var upperMax uint64
if shift < 63 {
upperMax = term | ((uint64(1) << (shift + 1)) - 1)
} else {
upperMax = 0xffffffffffffffff
}
lowerMax := split - 1
onBoundary, notOnBoundary = relateAndRecurse(term, lowerMax, shift,
sminLon, sminLat, smaxLon, smaxLat, checkBoundaries)
plusOnBoundary, plusNotOnBoundary := relateAndRecurse(split, upperMax, shift,
sminLon, sminLat, smaxLon, smaxLat, checkBoundaries)
onBoundary = append(onBoundary, plusOnBoundary...)
notOnBoundary = append(notOnBoundary, plusNotOnBoundary...)
return
}
func relateAndRecurse(start, end uint64, res uint,
sminLon, sminLat, smaxLon, smaxLat float64,
checkBoundaries bool) (
onBoundary [][]byte, notOnBoundary [][]byte) {
minLon := geo.MortonUnhashLon(start)
minLat := geo.MortonUnhashLat(start)
maxLon := geo.MortonUnhashLon(end)
maxLat := geo.MortonUnhashLat(end)
level := ((geo.GeoBits << 1) - res) >> 1
within := res%document.GeoPrecisionStep == 0 &&
geo.RectWithin(minLon, minLat, maxLon, maxLat,
sminLon, sminLat, smaxLon, smaxLat)
if within || (level == geoDetailLevel &&
geo.RectIntersects(minLon, minLat, maxLon, maxLat,
sminLon, sminLat, smaxLon, smaxLat)) {
if !within && checkBoundaries {
return [][]byte{
numeric.MustNewPrefixCodedInt64(int64(start), res),
}, nil
}
return nil,
[][]byte{
numeric.MustNewPrefixCodedInt64(int64(start), res),
}
} else if level < geoDetailLevel &&
geo.RectIntersects(minLon, minLat, maxLon, maxLat,
sminLon, sminLat, smaxLon, smaxLat) {
return ComputeGeoRange(start, res-1, sminLon, sminLat, smaxLon, smaxLat,
checkBoundaries)
}
return nil, nil
}
func buildRectFilter(indexReader index.IndexReader, field string,
minLon, minLat, maxLon, maxLat float64) FilterFunc {
return func(d *search.DocumentMatch) bool {
var lon, lat float64
var found bool
err := indexReader.DocumentVisitFieldTerms(d.IndexInternalID,
[]string{field}, func(field string, term []byte) {
// only consider the values which are shifted 0
prefixCoded := numeric.PrefixCoded(term)
shift, err := prefixCoded.Shift()
if err == nil && shift == 0 {
var i64 int64
i64, err = prefixCoded.Int64()
if err == nil {
lon = geo.MortonUnhashLon(uint64(i64))
lat = geo.MortonUnhashLat(uint64(i64))
found = true
}
}
})
if err == nil && found {
return geo.BoundingBoxContains(lon, lat,
minLon, minLat, maxLon, maxLat)
}
return false
}
}

View file

@ -0,0 +1,115 @@
// 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 searcher
import (
"github.com/blevesearch/bleve/geo"
"github.com/blevesearch/bleve/index"
"github.com/blevesearch/bleve/numeric"
"github.com/blevesearch/bleve/search"
)
func NewGeoPointDistanceSearcher(indexReader index.IndexReader, centerLon,
centerLat, dist float64, field string, boost float64,
options search.SearcherOptions) (search.Searcher, error) {
// compute bounding box containing the circle
topLeftLon, topLeftLat, bottomRightLon, bottomRightLat :=
geo.ComputeBoundingBox(centerLon, centerLat, dist)
// build a searcher for the box
boxSearcher, err := boxSearcher(indexReader,
topLeftLon, topLeftLat, bottomRightLon, bottomRightLat,
field, boost, options)
if err != nil {
return nil, err
}
// wrap it in a filtering searcher which checks the actual distance
return NewFilteringSearcher(boxSearcher,
buildDistFilter(indexReader, field, centerLon, centerLat, dist)), nil
}
// boxSearcher builds a searcher for the described bounding box
// if the desired box crosses the dateline, it is automatically split into
// two boxes joined through a disjunction searcher
func boxSearcher(indexReader index.IndexReader,
topLeftLon, topLeftLat, bottomRightLon, bottomRightLat float64,
field string, boost float64, options search.SearcherOptions) (
search.Searcher, error) {
if bottomRightLon < topLeftLon {
// cross date line, rewrite as two parts
leftSearcher, err := NewGeoBoundingBoxSearcher(indexReader,
-180, bottomRightLat, bottomRightLon, topLeftLat,
field, boost, options, false)
if err != nil {
return nil, err
}
rightSearcher, err := NewGeoBoundingBoxSearcher(indexReader,
topLeftLon, bottomRightLat, 180, topLeftLat, field, boost, options, false)
if err != nil {
_ = leftSearcher.Close()
return nil, err
}
boxSearcher, err := NewDisjunctionSearcher(indexReader,
[]search.Searcher{leftSearcher, rightSearcher}, 0, options)
if err != nil {
_ = leftSearcher.Close()
_ = rightSearcher.Close()
return nil, err
}
return boxSearcher, nil
}
// build geoboundinggox searcher for that bounding box
boxSearcher, err := NewGeoBoundingBoxSearcher(indexReader,
topLeftLon, bottomRightLat, bottomRightLon, topLeftLat, field, boost,
options, false)
if err != nil {
return nil, err
}
return boxSearcher, nil
}
func buildDistFilter(indexReader index.IndexReader, field string,
centerLon, centerLat, maxDist float64) FilterFunc {
return func(d *search.DocumentMatch) bool {
var lon, lat float64
var found bool
err := indexReader.DocumentVisitFieldTerms(d.IndexInternalID,
[]string{field}, func(field string, term []byte) {
// only consider the values which are shifted 0
prefixCoded := numeric.PrefixCoded(term)
shift, err := prefixCoded.Shift()
if err == nil && shift == 0 {
i64, err := prefixCoded.Int64()
if err == nil {
lon = geo.MortonUnhashLon(uint64(i64))
lat = geo.MortonUnhashLat(uint64(i64))
found = true
}
}
})
if err == nil && found {
dist := geo.Haversin(lon, lat, centerLon, centerLat)
if dist <= maxDist/1000 {
return true
}
}
return false
}
}

View file

@ -0,0 +1,105 @@
// Copyright (c) 2014 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 searcher
import (
"github.com/blevesearch/bleve/index"
"github.com/blevesearch/bleve/search"
"github.com/blevesearch/bleve/search/scorer"
)
type MatchAllSearcher struct {
indexReader index.IndexReader
reader index.DocIDReader
scorer *scorer.ConstantScorer
count uint64
}
func NewMatchAllSearcher(indexReader index.IndexReader, boost float64, options search.SearcherOptions) (*MatchAllSearcher, error) {
reader, err := indexReader.DocIDReaderAll()
if err != nil {
return nil, err
}
count, err := indexReader.DocCount()
if err != nil {
_ = reader.Close()
return nil, err
}
scorer := scorer.NewConstantScorer(1.0, boost, options)
return &MatchAllSearcher{
indexReader: indexReader,
reader: reader,
scorer: scorer,
count: count,
}, nil
}
func (s *MatchAllSearcher) Count() uint64 {
return s.count
}
func (s *MatchAllSearcher) Weight() float64 {
return s.scorer.Weight()
}
func (s *MatchAllSearcher) SetQueryNorm(qnorm float64) {
s.scorer.SetQueryNorm(qnorm)
}
func (s *MatchAllSearcher) Next(ctx *search.SearchContext) (*search.DocumentMatch, error) {
id, err := s.reader.Next()
if err != nil {
return nil, err
}
if id == nil {
return nil, nil
}
// score match
docMatch := s.scorer.Score(ctx, id)
// return doc match
return docMatch, nil
}
func (s *MatchAllSearcher) Advance(ctx *search.SearchContext, ID index.IndexInternalID) (*search.DocumentMatch, error) {
id, err := s.reader.Advance(ID)
if err != nil {
return nil, err
}
if id == nil {
return nil, nil
}
// score match
docMatch := s.scorer.Score(ctx, id)
// return doc match
return docMatch, nil
}
func (s *MatchAllSearcher) Close() error {
return s.reader.Close()
}
func (s *MatchAllSearcher) Min() int {
return 0
}
func (s *MatchAllSearcher) DocumentMatchPoolSize() int {
return 1
}

View file

@ -0,0 +1,62 @@
// Copyright (c) 2014 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 searcher
import (
"github.com/blevesearch/bleve/index"
"github.com/blevesearch/bleve/search"
)
type MatchNoneSearcher struct {
indexReader index.IndexReader
}
func NewMatchNoneSearcher(indexReader index.IndexReader) (*MatchNoneSearcher, error) {
return &MatchNoneSearcher{
indexReader: indexReader,
}, nil
}
func (s *MatchNoneSearcher) Count() uint64 {
return uint64(0)
}
func (s *MatchNoneSearcher) Weight() float64 {
return 0.0
}
func (s *MatchNoneSearcher) SetQueryNorm(qnorm float64) {
}
func (s *MatchNoneSearcher) Next(ctx *search.SearchContext) (*search.DocumentMatch, error) {
return nil, nil
}
func (s *MatchNoneSearcher) Advance(ctx *search.SearchContext, ID index.IndexInternalID) (*search.DocumentMatch, error) {
return nil, nil
}
func (s *MatchNoneSearcher) Close() error {
return nil
}
func (s *MatchNoneSearcher) Min() int {
return 0
}
func (s *MatchNoneSearcher) DocumentMatchPoolSize() int {
return 0
}

View file

@ -0,0 +1,85 @@
// 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 searcher
import (
"github.com/blevesearch/bleve/index"
"github.com/blevesearch/bleve/search"
)
func NewMultiTermSearcher(indexReader index.IndexReader, terms []string,
field string, boost float64, options search.SearcherOptions, limit bool) (
search.Searcher, error) {
qsearchers := make([]search.Searcher, len(terms))
qsearchersClose := func() {
for _, searcher := range qsearchers {
if searcher != nil {
_ = searcher.Close()
}
}
}
for i, term := range terms {
var err error
qsearchers[i], err = NewTermSearcher(indexReader, term, field, boost, options)
if err != nil {
qsearchersClose()
return nil, err
}
}
// build disjunction searcher of these ranges
return newMultiTermSearcherBytes(indexReader, qsearchers, field, boost,
options, limit)
}
func NewMultiTermSearcherBytes(indexReader index.IndexReader, terms [][]byte,
field string, boost float64, options search.SearcherOptions, limit bool) (
search.Searcher, error) {
qsearchers := make([]search.Searcher, len(terms))
qsearchersClose := func() {
for _, searcher := range qsearchers {
if searcher != nil {
_ = searcher.Close()
}
}
}
for i, term := range terms {
var err error
qsearchers[i], err = NewTermSearcherBytes(indexReader, term, field, boost, options)
if err != nil {
qsearchersClose()
return nil, err
}
}
return newMultiTermSearcherBytes(indexReader, qsearchers, field, boost,
options, limit)
}
func newMultiTermSearcherBytes(indexReader index.IndexReader,
searchers []search.Searcher, field string, boost float64,
options search.SearcherOptions, limit bool) (
search.Searcher, error) {
// build disjunction searcher of these ranges
searcher, err := newDisjunctionSearcher(indexReader, searchers, 0, options,
limit)
if err != nil {
for _, s := range searchers {
_ = s.Close()
}
return nil, err
}
return searcher, nil
}

View file

@ -0,0 +1,169 @@
// Copyright (c) 2014 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 searcher
import (
"bytes"
"math"
"github.com/blevesearch/bleve/index"
"github.com/blevesearch/bleve/numeric"
"github.com/blevesearch/bleve/search"
)
func NewNumericRangeSearcher(indexReader index.IndexReader,
min *float64, max *float64, inclusiveMin, inclusiveMax *bool, field string,
boost float64, options search.SearcherOptions) (search.Searcher, error) {
// account for unbounded edges
if min == nil {
negInf := math.Inf(-1)
min = &negInf
}
if max == nil {
Inf := math.Inf(1)
max = &Inf
}
if inclusiveMin == nil {
defaultInclusiveMin := true
inclusiveMin = &defaultInclusiveMin
}
if inclusiveMax == nil {
defaultInclusiveMax := false
inclusiveMax = &defaultInclusiveMax
}
// find all the ranges
minInt64 := numeric.Float64ToInt64(*min)
if !*inclusiveMin && minInt64 != math.MaxInt64 {
minInt64++
}
maxInt64 := numeric.Float64ToInt64(*max)
if !*inclusiveMax && maxInt64 != math.MinInt64 {
maxInt64--
}
// FIXME hard-coded precision, should match field declaration
termRanges := splitInt64Range(minInt64, maxInt64, 4)
terms := termRanges.Enumerate()
if tooManyClauses(len(terms)) {
return nil, tooManyClausesErr()
}
return NewMultiTermSearcherBytes(indexReader, terms, field, boost, options,
true)
}
type termRange struct {
startTerm []byte
endTerm []byte
}
func (t *termRange) Enumerate() [][]byte {
var rv [][]byte
next := t.startTerm
for bytes.Compare(next, t.endTerm) <= 0 {
rv = append(rv, next)
next = incrementBytes(next)
}
return rv
}
func incrementBytes(in []byte) []byte {
rv := make([]byte, len(in))
copy(rv, in)
for i := len(rv) - 1; i >= 0; i-- {
rv[i] = rv[i] + 1
if rv[i] != 0 {
// didn't overflow, so stop
break
}
}
return rv
}
type termRanges []*termRange
func (tr termRanges) Enumerate() [][]byte {
var rv [][]byte
for _, tri := range tr {
trie := tri.Enumerate()
rv = append(rv, trie...)
}
return rv
}
func splitInt64Range(minBound, maxBound int64, precisionStep uint) termRanges {
rv := make(termRanges, 0)
if minBound > maxBound {
return rv
}
for shift := uint(0); ; shift += precisionStep {
diff := int64(1) << (shift + precisionStep)
mask := ((int64(1) << precisionStep) - int64(1)) << shift
hasLower := (minBound & mask) != int64(0)
hasUpper := (maxBound & mask) != mask
var nextMinBound int64
if hasLower {
nextMinBound = (minBound + diff) &^ mask
} else {
nextMinBound = minBound &^ mask
}
var nextMaxBound int64
if hasUpper {
nextMaxBound = (maxBound - diff) &^ mask
} else {
nextMaxBound = maxBound &^ mask
}
lowerWrapped := nextMinBound < minBound
upperWrapped := nextMaxBound > maxBound
if shift+precisionStep >= 64 || nextMinBound > nextMaxBound ||
lowerWrapped || upperWrapped {
// We are in the lowest precision or the next precision is not available.
rv = append(rv, newRange(minBound, maxBound, shift))
// exit the split recursion loop
break
}
if hasLower {
rv = append(rv, newRange(minBound, minBound|mask, shift))
}
if hasUpper {
rv = append(rv, newRange(maxBound&^mask, maxBound, shift))
}
// recurse to next precision
minBound = nextMinBound
maxBound = nextMaxBound
}
return rv
}
func newRange(minBound, maxBound int64, shift uint) *termRange {
maxBound |= (int64(1) << shift) - int64(1)
minBytes := numeric.MustNewPrefixCodedInt64(minBound, shift)
maxBytes := numeric.MustNewPrefixCodedInt64(maxBound, shift)
return newRangeBytes(minBytes, maxBytes)
}
func newRangeBytes(minBytes, maxBytes []byte) *termRange {
return &termRange{
startTerm: minBytes,
endTerm: maxBytes,
}
}

View file

@ -0,0 +1,341 @@
// Copyright (c) 2014 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 searcher
import (
"fmt"
"math"
"github.com/blevesearch/bleve/index"
"github.com/blevesearch/bleve/search"
)
type PhraseSearcher struct {
indexReader index.IndexReader
mustSearcher *ConjunctionSearcher
queryNorm float64
currMust *search.DocumentMatch
slop int
terms [][]string
initialized bool
}
func NewPhraseSearcher(indexReader index.IndexReader, terms []string, field string, options search.SearcherOptions) (*PhraseSearcher, error) {
// turn flat terms []string into [][]string
mterms := make([][]string, len(terms))
for i, term := range terms {
mterms[i] = []string{term}
}
return NewMultiPhraseSearcher(indexReader, mterms, field, options)
}
func NewMultiPhraseSearcher(indexReader index.IndexReader, terms [][]string, field string, options search.SearcherOptions) (*PhraseSearcher, error) {
options.IncludeTermVectors = true
var termPositionSearchers []search.Searcher
for _, termPos := range terms {
if len(termPos) == 1 && termPos[0] != "" {
// single term
ts, err := NewTermSearcher(indexReader, termPos[0], field, 1.0, options)
if err != nil {
// close any searchers already opened
for _, ts := range termPositionSearchers {
_ = ts.Close()
}
return nil, fmt.Errorf("phrase searcher error building term searcher: %v", err)
}
termPositionSearchers = append(termPositionSearchers, ts)
} else if len(termPos) > 1 {
// multiple terms
var termSearchers []search.Searcher
for _, term := range termPos {
if term == "" {
continue
}
ts, err := NewTermSearcher(indexReader, term, field, 1.0, options)
if err != nil {
// close any searchers already opened
for _, ts := range termPositionSearchers {
_ = ts.Close()
}
return nil, fmt.Errorf("phrase searcher error building term searcher: %v", err)
}
termSearchers = append(termSearchers, ts)
}
disjunction, err := NewDisjunctionSearcher(indexReader, termSearchers, 1, options)
if err != nil {
// close any searchers already opened
for _, ts := range termPositionSearchers {
_ = ts.Close()
}
return nil, fmt.Errorf("phrase searcher error building term position disjunction searcher: %v", err)
}
termPositionSearchers = append(termPositionSearchers, disjunction)
}
}
mustSearcher, err := NewConjunctionSearcher(indexReader, termPositionSearchers, options)
if err != nil {
// close any searchers already opened
for _, ts := range termPositionSearchers {
_ = ts.Close()
}
return nil, fmt.Errorf("phrase searcher error building conjunction searcher: %v", err)
}
// build our searcher
rv := PhraseSearcher{
indexReader: indexReader,
mustSearcher: mustSearcher,
terms: terms,
}
rv.computeQueryNorm()
return &rv, nil
}
func (s *PhraseSearcher) computeQueryNorm() {
// first calculate sum of squared weights
sumOfSquaredWeights := 0.0
if s.mustSearcher != nil {
sumOfSquaredWeights += s.mustSearcher.Weight()
}
// now compute query norm from this
s.queryNorm = 1.0 / math.Sqrt(sumOfSquaredWeights)
// finally tell all the downstream searchers the norm
if s.mustSearcher != nil {
s.mustSearcher.SetQueryNorm(s.queryNorm)
}
}
func (s *PhraseSearcher) initSearchers(ctx *search.SearchContext) error {
err := s.advanceNextMust(ctx)
if err != nil {
return err
}
s.initialized = true
return nil
}
func (s *PhraseSearcher) advanceNextMust(ctx *search.SearchContext) error {
var err error
if s.mustSearcher != nil {
s.currMust, err = s.mustSearcher.Next(ctx)
if err != nil {
return err
}
}
return nil
}
func (s *PhraseSearcher) Weight() float64 {
return s.mustSearcher.Weight()
}
func (s *PhraseSearcher) SetQueryNorm(qnorm float64) {
s.mustSearcher.SetQueryNorm(qnorm)
}
func (s *PhraseSearcher) Next(ctx *search.SearchContext) (*search.DocumentMatch, error) {
if !s.initialized {
err := s.initSearchers(ctx)
if err != nil {
return nil, err
}
}
for s.currMust != nil {
// check this match against phrase constraints
rv := s.checkCurrMustMatch(ctx)
// prepare for next iteration (either loop or subsequent call to Next())
err := s.advanceNextMust(ctx)
if err != nil {
return nil, err
}
// if match satisfied phrase constraints return it as a hit
if rv != nil {
return rv, nil
}
}
return nil, nil
}
// checkCurrMustMatch is soley concerned with determining if the DocumentMatch
// pointed to by s.currMust (which satisifies the pre-condition searcher)
// also satisfies the phase constraints. if so, it returns a DocumentMatch
// for this document, otherwise nil
func (s *PhraseSearcher) checkCurrMustMatch(ctx *search.SearchContext) *search.DocumentMatch {
rvftlm := make(search.FieldTermLocationMap, 0)
freq := 0
// typically we would expect there to only actually be results in
// one field, but we allow for this to not be the case
// but, we note that phrase constraints can only be satisfied within
// a single field, so we can check them each independently
for field, tlm := range s.currMust.Locations {
f, rvtlm := s.checkCurrMustMatchField(ctx, tlm)
if f > 0 {
freq += f
rvftlm[field] = rvtlm
}
}
if freq > 0 {
// return match
rv := s.currMust
rv.Locations = rvftlm
return rv
}
return nil
}
// checkCurrMustMatchField is soley concerned with determining if one particular
// field within the currMust DocumentMatch Locations satisfies the phase
// constraints (possibly more than once). if so, the number of times it was
// satisfied, and these locations are returned. otherwise 0 and either
// a nil or empty TermLocationMap
func (s *PhraseSearcher) checkCurrMustMatchField(ctx *search.SearchContext, tlm search.TermLocationMap) (int, search.TermLocationMap) {
paths := findPhrasePaths(0, nil, s.terms, tlm, nil, 0)
rv := make(search.TermLocationMap, len(s.terms))
for _, p := range paths {
p.MergeInto(rv)
}
return len(paths), rv
}
type phrasePart struct {
term string
loc *search.Location
}
type phrasePath []*phrasePart
func (p phrasePath) MergeInto(in search.TermLocationMap) {
for _, pp := range p {
in[pp.term] = append(in[pp.term], pp.loc)
}
}
// findPhrasePaths is a function to identify phase matches from a set of known
// term locations. the implementation is recursive, so care must be taken
// with arguments and return values.
//
// prev - the previous location, nil on first invocation
// phraseTerms - slice containing the phrase terms themselves
// may contain empty string as placeholder (don't care)
// tlm - the Term Location Map containing all relevant term locations
// offset - the offset from the previous that this next term must match
// p - the current path being explored (appended to in recursive calls)
// this is the primary state being built during the traversal
//
// returns slice of paths, or nil if invocation did not find any successul paths
func findPhrasePaths(prevPos uint64, ap search.ArrayPositions, phraseTerms [][]string, tlm search.TermLocationMap, p phrasePath, remainingSlop int) []phrasePath {
// no more terms
if len(phraseTerms) < 1 {
return []phrasePath{p}
}
car := phraseTerms[0]
cdr := phraseTerms[1:]
// empty term is treated as match (continue)
if len(car) == 0 || (len(car) == 1 && car[0] == "") {
nextPos := prevPos + 1
if prevPos == 0 {
// if prevPos was 0, don't set it to 1 (as thats not a real abs pos)
nextPos = 0 // don't advance nextPos if prevPos was 0
}
return findPhrasePaths(nextPos, ap, cdr, tlm, p, remainingSlop)
}
var rv []phrasePath
// locations for this term
for _, carTerm := range car {
locations := tlm[carTerm]
for _, loc := range locations {
if prevPos != 0 && !loc.ArrayPositions.Equals(ap) {
// if the array positions are wrong, can't match, try next location
continue
}
// compute distance from previous phrase term
dist := 0
if prevPos != 0 {
dist = editDistance(prevPos+1, loc.Pos)
}
// if enough slop reamining, continue recursively
if prevPos == 0 || (remainingSlop-dist) >= 0 {
// this location works, add it to the path (but not for empty term)
px := append(p, &phrasePart{term: carTerm, loc: loc})
rv = append(rv, findPhrasePaths(loc.Pos, loc.ArrayPositions, cdr, tlm, px, remainingSlop-dist)...)
}
}
}
return rv
}
func editDistance(p1, p2 uint64) int {
dist := int(p1 - p2)
if dist < 0 {
return -dist
}
return dist
}
func (s *PhraseSearcher) Advance(ctx *search.SearchContext, ID index.IndexInternalID) (*search.DocumentMatch, error) {
if !s.initialized {
err := s.initSearchers(ctx)
if err != nil {
return nil, err
}
}
var err error
s.currMust, err = s.mustSearcher.Advance(ctx, ID)
if err != nil {
return nil, err
}
return s.Next(ctx)
}
func (s *PhraseSearcher) Count() uint64 {
// for now return a worst case
return s.mustSearcher.Count()
}
func (s *PhraseSearcher) Close() error {
if s.mustSearcher != nil {
err := s.mustSearcher.Close()
if err != nil {
return err
}
}
return nil
}
func (s *PhraseSearcher) Min() int {
return 0
}
func (s *PhraseSearcher) DocumentMatchPoolSize() int {
return s.mustSearcher.DocumentMatchPoolSize() + 1
}

View file

@ -0,0 +1,80 @@
// Copyright (c) 2015 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 searcher
import (
"regexp"
"github.com/blevesearch/bleve/index"
"github.com/blevesearch/bleve/search"
)
// NewRegexpSearcher creates a searcher which will match documents that
// contain terms which match the pattern regexp. The match must be EXACT
// matching the entire term. The provided regexp SHOULD NOT start with ^
// or end with $ as this can intefere with the implementation. Separately,
// matches will be checked to ensure they match the entire term.
func NewRegexpSearcher(indexReader index.IndexReader, pattern *regexp.Regexp,
field string, boost float64, options search.SearcherOptions) (
search.Searcher, error) {
prefixTerm, complete := pattern.LiteralPrefix()
var candidateTerms []string
if complete {
// there is no pattern
candidateTerms = []string{prefixTerm}
} else {
var err error
candidateTerms, err = findRegexpCandidateTerms(indexReader, pattern, field,
prefixTerm)
if err != nil {
return nil, err
}
}
return NewMultiTermSearcher(indexReader, candidateTerms, field, boost,
options, true)
}
func findRegexpCandidateTerms(indexReader index.IndexReader,
pattern *regexp.Regexp, field, prefixTerm string) (rv []string, err error) {
rv = make([]string, 0)
var fieldDict index.FieldDict
if len(prefixTerm) > 0 {
fieldDict, err = indexReader.FieldDictPrefix(field, []byte(prefixTerm))
} else {
fieldDict, err = indexReader.FieldDict(field)
}
defer func() {
if cerr := fieldDict.Close(); cerr != nil && err == nil {
err = cerr
}
}()
// enumerate the terms and check against regexp
tfd, err := fieldDict.Next()
for err == nil && tfd != nil {
matchPos := pattern.FindStringIndex(tfd.Term)
if matchPos != nil && matchPos[0] == 0 && matchPos[1] == len(tfd.Term) {
rv = append(rv, tfd.Term)
if tooManyClauses(len(rv)) {
return rv, tooManyClausesErr()
}
}
tfd, err = fieldDict.Next()
}
return rv, err
}

View file

@ -0,0 +1,122 @@
// Copyright (c) 2014 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 searcher
import (
"github.com/blevesearch/bleve/index"
"github.com/blevesearch/bleve/search"
"github.com/blevesearch/bleve/search/scorer"
)
type TermSearcher struct {
indexReader index.IndexReader
reader index.TermFieldReader
scorer *scorer.TermQueryScorer
tfd index.TermFieldDoc
}
func NewTermSearcher(indexReader index.IndexReader, term string, field string, boost float64, options search.SearcherOptions) (*TermSearcher, error) {
reader, err := indexReader.TermFieldReader([]byte(term), field, true, true, options.IncludeTermVectors)
if err != nil {
return nil, err
}
count, err := indexReader.DocCount()
if err != nil {
_ = reader.Close()
return nil, err
}
scorer := scorer.NewTermQueryScorer([]byte(term), field, boost, count, reader.Count(), options)
return &TermSearcher{
indexReader: indexReader,
reader: reader,
scorer: scorer,
}, nil
}
func NewTermSearcherBytes(indexReader index.IndexReader, term []byte, field string, boost float64, options search.SearcherOptions) (*TermSearcher, error) {
reader, err := indexReader.TermFieldReader(term, field, true, true, options.IncludeTermVectors)
if err != nil {
return nil, err
}
count, err := indexReader.DocCount()
if err != nil {
_ = reader.Close()
return nil, err
}
scorer := scorer.NewTermQueryScorer(term, field, boost, count, reader.Count(), options)
return &TermSearcher{
indexReader: indexReader,
reader: reader,
scorer: scorer,
}, nil
}
func (s *TermSearcher) Count() uint64 {
return s.reader.Count()
}
func (s *TermSearcher) Weight() float64 {
return s.scorer.Weight()
}
func (s *TermSearcher) SetQueryNorm(qnorm float64) {
s.scorer.SetQueryNorm(qnorm)
}
func (s *TermSearcher) Next(ctx *search.SearchContext) (*search.DocumentMatch, error) {
termMatch, err := s.reader.Next(s.tfd.Reset())
if err != nil {
return nil, err
}
if termMatch == nil {
return nil, nil
}
// score match
docMatch := s.scorer.Score(ctx, termMatch)
// return doc match
return docMatch, nil
}
func (s *TermSearcher) Advance(ctx *search.SearchContext, ID index.IndexInternalID) (*search.DocumentMatch, error) {
termMatch, err := s.reader.Advance(ID, s.tfd.Reset())
if err != nil {
return nil, err
}
if termMatch == nil {
return nil, nil
}
// score match
docMatch := s.scorer.Score(ctx, termMatch)
// return doc match
return docMatch, nil
}
func (s *TermSearcher) Close() error {
return s.reader.Close()
}
func (s *TermSearcher) Min() int {
return 0
}
func (s *TermSearcher) DocumentMatchPoolSize() int {
return 1
}

View file

@ -0,0 +1,39 @@
// Copyright (c) 2014 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 searcher
import (
"github.com/blevesearch/bleve/index"
"github.com/blevesearch/bleve/search"
)
func NewTermPrefixSearcher(indexReader index.IndexReader, prefix string,
field string, boost float64, options search.SearcherOptions) (
search.Searcher, error) {
// find the terms with this prefix
fieldDict, err := indexReader.FieldDictPrefix(field, []byte(prefix))
if err != nil {
return nil, err
}
var terms []string
tfd, err := fieldDict.Next()
for err == nil && tfd != nil {
terms = append(terms, tfd.Term)
tfd, err = fieldDict.Next()
}
return NewMultiTermSearcher(indexReader, terms, field, boost, options, true)
}

View file

@ -0,0 +1,75 @@
// 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 searcher
import (
"github.com/blevesearch/bleve/index"
"github.com/blevesearch/bleve/search"
)
func NewTermRangeSearcher(indexReader index.IndexReader,
min, max []byte, inclusiveMin, inclusiveMax *bool, field string,
boost float64, options search.SearcherOptions) (search.Searcher, error) {
if inclusiveMin == nil {
defaultInclusiveMin := true
inclusiveMin = &defaultInclusiveMin
}
if inclusiveMax == nil {
defaultInclusiveMax := false
inclusiveMax = &defaultInclusiveMax
}
if min == nil {
min = []byte{}
}
rangeMax := max
if rangeMax != nil {
// the term dictionary range end has an unfortunate implementation
rangeMax = append(rangeMax, 0)
}
// find the terms with this prefix
fieldDict, err := indexReader.FieldDictRange(field, min, rangeMax)
if err != nil {
return nil, err
}
var terms []string
tfd, err := fieldDict.Next()
for err == nil && tfd != nil {
terms = append(terms, tfd.Term)
tfd, err = fieldDict.Next()
}
if err != nil {
return nil, err
}
if len(terms) < 1 {
return NewMatchNoneSearcher(indexReader)
}
if !*inclusiveMin && min != nil && string(min) == terms[0] {
terms = terms[1:]
}
// if our term list included the max, it would be the last item
if !*inclusiveMax && max != nil && string(max) == terms[len(terms)-1] {
terms = terms[:len(terms)-1]
}
return NewMultiTermSearcher(indexReader, terms, field, boost, options, true)
}