gibheer
/
bleve
0
0
Fork 0
Code

optimize DisjunctionSearcher.Next() 

This change simplifies and removes the DisjunctionSearcher.currentID
tracking, and instead utilizes the the matching/matchingIdxs slices
for tracking the required information.

As the core of the optimization, the previous code used two loop
passses to compare the internal ID's to the currentID field.  This
commit instead optimizes to have a single pass to both compare the
internalID's and to also maintain the matching/matchingIdxs arrays.

On my dev box, using a bleve-query benchmark on a wiki dataset, with
query-string of "text:afternoon text:coffee", the previous code had
throughput of 958qps, and this commit has 1174qps.
This commit is contained in:
Steve Yen 2016-09-20 19:22:37 -07:00
parent 75281a1f9f
commit e344582021
1 changed files with 33 additions and 23 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

56
search/searchers/search_disjunction.go
View File

@ -29,7 +29,6 @@ type DisjunctionSearcher struct {
searchers OrderedSearcherList
queryNorm float64
currs []*search.DocumentMatch
currentID index.IndexInternalID
scorer *scorers.DisjunctionQueryScorer
min int
matching []*search.DocumentMatch
@ -100,19 +99,38 @@ func (s *DisjunctionSearcher) initSearchers(ctx *search.SearchContext) error {
}
}
s.currentID = s.nextSmallestID()
s.updateMatches()
s.initialized = true
return nil
}
func (s *DisjunctionSearcher) nextSmallestID() index.IndexInternalID {
var rv index.IndexInternalID
for _, curr := range s.currs {
if curr != nil && (curr.IndexInternalID.Compare(rv) < 0 || rv == nil) {
rv = curr.IndexInternalID
func (s *DisjunctionSearcher) updateMatches() {
matching := s.matching[:0]
matchingIdxs := s.matchingIdxs[:0]
for i, curr := range s.currs {
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)
}
return rv
s.matching = matching
s.matchingIdxs = matchingIdxs
}
func (s *DisjunctionSearcher) Weight() float64 {
@ -140,24 +158,15 @@ func (s *DisjunctionSearcher) Next(ctx *search.SearchContext) (*search.DocumentM
var rv *search.DocumentMatch
found := false
for !found && s.currentID != nil {
matching := s.matching[:0]
matchingIdxs := s.matchingIdxs[:0]
for i, curr := range s.currs {
if curr != nil && curr.IndexInternalID.Equals(s.currentID) {
matching = append(matching, curr)
matchingIdxs = append(matchingIdxs, i)
}
}
if len(matching) >= s.min {
for !found && len(s.matching) > 0 {
if len(s.matching) >= s.min {
found = true
// score this match
rv = s.scorer.Score(ctx, matching, len(matching), len(s.searchers))
rv = s.scorer.Score(ctx, s.matching, len(s.matching), len(s.searchers))
}
// invoke next on all the matching searchers
for _, i := range matchingIdxs {
for _, i := range s.matchingIdxs {
searcher := s.searchers[i]
if s.currs[i] != rv {
ctx.DocumentMatchPool.Put(s.currs[i])
@ -167,7 +176,8 @@ func (s *DisjunctionSearcher) Next(ctx *search.SearchContext) (*search.DocumentM
return nil, err
}
}
s.currentID = s.nextSmallestID()
s.updateMatches()
}
return rv, nil
}
@ -191,7 +201,7 @@ func (s *DisjunctionSearcher) Advance(ctx *search.SearchContext, ID index.IndexI
}
}
s.currentID = s.nextSmallestID()
s.updateMatches()
return s.Next(ctx)
}