MB-59633: Disable DocValues Chunking & Compression for Geo Fields

document/field_geopoint.go

@@ -182,8 +182,12 @@
 	prefixCoded := numeric.MustNewPrefixCodedInt64(int64(mhash), 0)
 
 	// docvalues are always enabled for geopoint fields, even if the
-	// indexing options are set to not include them
+	// indexing options are set to not include docvalues.
+	// snappy compression and chunking are always skipped for geopoint
+	// to avoid mem copies and faster lookups.
 	options |= index.DocValues
+	options |= index.SkipDVChunking
+	options |= index.SkipDVCompression
 
 	return &GeoPointField{
 		name:           name,

document/field_geoshape.go

@@ -180,7 +180,11 @@
 
 	// docvalues are always enabled for geoshape fields, even if the
 	// indexing options are set to not include docvalues.
+	// snappy compression and chunking are always skipped for geoshape
+	// to avoid mem copies and faster lookups.
 	options |= index.DocValues
+	options |= index.SkipDVChunking
+	options |= index.SkipDVCompression
 
 	return &GeoShapeField{
 		shape:             shape,
@@ -232,7 +236,11 @@
 
 	// docvalues are always enabled for geoshape fields, even if the
 	// indexing options are set to not include docvalues.
+	// snappy compression and chunking are always skipped for geoshape
+	// to avoid mem copies and faster lookups.
 	options |= index.DocValues
+	options |= index.SkipDVChunking
+	options |= index.SkipDVCompression
 
 	return &GeoShapeField{
 		shape:             shape,

index/scorch/snapshot_segment.go

@@ -26,10 +26,6 @@ import (
 	segment "github.com/blevesearch/scorch_segment_api/v2"
 )
 
-var TermSeparator byte = 0xff
-
-var TermSeparatorSplitSlice = []byte{TermSeparator}
-
 type SegmentSnapshot struct {
 	// this flag is needed to identify whether this
 	// segment was mmaped recently, in which case
@@ -233,7 +229,7 @@ func (cfd *cachedFieldDocs) prepareField(field string, ss *SegmentSnapshot) {
 		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)
+			cfd.docs[docNum] = append(cfd.docs[docNum], index.DocValueTermSeparator)
 			cfd.size += uint64(len(next.Term) + 1) // map value
 			nextPosting, err2 = postingsItr.Next()
 		}
@@ -334,7 +330,7 @@ func (c *cachedDocs) visitDoc(localDocNum uint64,
 
 			if tlist, exists := cachedFieldDocs.docs[localDocNum]; exists {
 				for {
-					i := bytes.Index(tlist, TermSeparatorSplitSlice)
+					i := bytes.IndexByte(tlist, index.DocValueTermSeparator)
 					if i < 0 {
 						break
 					}

search/searcher/search_geoshape.go

@@ -58,11 +58,6 @@ func NewGeoShapeSearcher(ctx context.Context, indexReader index.IndexReader, sha
 	return NewFilteringSearcher(ctx, mSearcher, buildRelationFilterOnShapes(ctx, dvReader, field, relation, shape)), nil
 }
 
-// Using the same term splitter slice used in the doc values in zap.
-// TODO: This needs to be revisited whenever we change the zap
-// implementation of doc values.
-var termSeparatorSplitSlice = []byte{0xff}
-
 func buildRelationFilterOnShapes(ctx context.Context, dvReader index.DocValueReader, field string,
 	relation string, shape index.GeoJSON,
 ) FilterFunc {
@@ -104,7 +99,7 @@ func buildRelationFilterOnShapes(ctx context.Context, dvReader index.DocValueRea
 					finishReading = true
 				}
 
-				dvShapeValue = append(dvShapeValue, termSeparatorSplitSlice...)
+				dvShapeValue = append(dvShapeValue, index.DocValueTermSeparator)
 				dvShapeValue = append(dvShapeValue, term...)
 			}