implementing the Count() API for the postings list, updated todos

build.go

@@ -24,7 +24,7 @@ import (
 	"github.com/blevesearch/vellum"
 )
 
-const Version uint32 = 16
+const Version uint32 = 15
 
 const Type string = "zap"
 

section_vector_index.go

@@ -141,6 +141,10 @@ func (vo *vectorIndexOpaque) writeVectorIndexes(w *CountHashWriter) (offset uint
 
 		// fixme: this can cause a write amplification. need to improve this.
 		// todo: might need to a reformating to optimize according to mmap needs.
+		// reformating idea: storing all the IDs mapping towards the end of the
+		// section would be help avoiding in paging in this data as part of a page
+		// (which is to load a non-cacheable info like index). this could help the
+		// paging costs
 		for vecID, docIDs := range docIDsMap {
 			// write the vecID
 			_, err := writeUvarints(w, vecID)

vector_search.go

@@ -36,6 +36,9 @@ func (vp *VecPosting) Size() int {
 // the score is actually a float32 value and in order to store it as a uint32 in
 // the bitmap, we use the IEEE 754 floating point format.
 type VecPostingsList struct {
+	// todo: perhaps we don't even need to store a bitmap if there is only
+	// one similar vector the query, but rather store it as a field value
+	// in the struct
 	except   *roaring64.Bitmap
 	postings *roaring64.Bitmap
 }
@@ -90,8 +93,13 @@ func (vpl *VecPostingsList) Size() int {
 	return 0
 }
 
-func (vpl *VecPostingsList) Count() uint64 {
-	return 0
+func (p *VecPostingsList) Count() uint64 {
+	n := p.postings.GetCardinality()
+	var e uint64
+	if p.except != nil {
+		e = p.postings.AndCardinality(p.except)
+	}
+	return n - e
 }
 
 func (vpl *VecPostingsList) ResetBytesRead(val uint64) {