gibheer
/
bleve
0
0
Fork 0
Code

scorch fix for TestSortMatchSearch 

The cachedDocs preparation has to happen for all docs in the field,
not just on the currently requested docNum.

Also, as part of this commit, there's a loop optimization where we no
longer use bytes.Split() on the terms buffer, thus avoiding garbage
creation.
This commit is contained in:
Steve Yen 2017-12-14 13:16:06 -08:00
parent 5bca9f06b9
commit a8884e1011
2 changed files with 15 additions and 17 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
index/scorch/snapshot_index.go
View File

@ -413,7 +413,7 @@ func (i *IndexSnapshot) DocumentVisitFieldTerms(id index.IndexInternalID,
ss := i.segment[segmentIndex]
err = ss.cachedDocs.prepareFields(localDocNum, fields, ss)
err = ss.cachedDocs.prepareFields(fields, ss)
if err != nil {
return err
}
@ -421,11 +421,13 @@ func (i *IndexSnapshot) DocumentVisitFieldTerms(id index.IndexInternalID,
for _, field := range fields {
if cachedFieldDocs, exists := ss.cachedDocs.cache[field]; exists {
if tlist, exists := cachedFieldDocs.docs[localDocNum]; exists {
terms := bytes.SplitN(tlist, TermSeparatorSplitSlice, -1)
for _, term := range terms {
if len(term) > 0 {
visitor(field, term)
for {
i := bytes.Index(tlist, TermSeparatorSplitSlice)
if i < 0 {
break
}
visitor(field, tlist[0:i])
tlist = tlist[i+1:]
}
}
}

20
index/scorch/snapshot_segment.go
View File

@ -171,8 +171,7 @@ type cachedFieldDocs struct {
docs map[uint64][]byte // Keyed by localDocNum, value is a list of terms delimited by 0xFF.
}
func (cfd *cachedFieldDocs) prepareFields(docNum uint64, field string,
ss *SegmentSnapshot) {
func (cfd *cachedFieldDocs) prepareFields(field string, ss *SegmentSnapshot) {
defer close(cfd.readyCh)
dict, err := ss.segment.Dictionary(field)
@ -183,7 +182,7 @@ func (cfd *cachedFieldDocs) prepareFields(docNum uint64, field string,
dictItr := dict.Iterator()
next, err := dictItr.Next()
for next != nil && err == nil {
for err == nil && next != nil {
postings, err1 := dict.PostingsList(next.Term, nil)
if err1 != nil {
cfd.err = err1
@ -192,12 +191,10 @@ func (cfd *cachedFieldDocs) prepareFields(docNum uint64, field string,
postingsItr := postings.Iterator()
nextPosting, err2 := postingsItr.Next()
for err2 == nil && nextPosting != nil && nextPosting.Number() <= docNum {
if nextPosting.Number() == docNum {
// got what we're looking for
cfd.docs[docNum] = append(cfd.docs[docNum], []byte(next.Term)...)
cfd.docs[docNum] = append(cfd.docs[docNum], TermSeparator)
}
for err2 == nil && nextPosting != nil {
docNum := nextPosting.Number()
cfd.docs[docNum] = append(cfd.docs[docNum], []byte(next.Term)...)
cfd.docs[docNum] = append(cfd.docs[docNum], TermSeparator)
nextPosting, err2 = postingsItr.Next()
}
@ -220,8 +217,7 @@ type cachedDocs struct {
cache map[string]*cachedFieldDocs // Keyed by field
}
func (c *cachedDocs) prepareFields(docNum uint64, wantedFields []string,
ss *SegmentSnapshot) error {
func (c *cachedDocs) prepareFields(wantedFields []string, ss *SegmentSnapshot) error {
c.m.Lock()
if c.cache == nil {
c.cache = make(map[string]*cachedFieldDocs, len(ss.Fields()))
@ -235,7 +231,7 @@ func (c *cachedDocs) prepareFields(docNum uint64, wantedFields []string,
docs: make(map[uint64][]byte),
}
go c.cache[field].prepareFields(docNum, field, ss)
go c.cache[field].prepareFields(field, ss)
}
}