From 6540b197d495053d66b48141bb1525e9a86f61a5 Mon Sep 17 00:00:00 2001
From: Steve Yen <steve.yen@gmail.com>
Date: Sat, 17 Mar 2018 10:40:34 -0700
Subject: [PATCH] scorch zap provide full buffer capacity to snappy
 Encode/Decode()

The snappy Encode/Decode() API's accept an optional destination buffer
param where their encoded/decoded output results will be placed, but
they only check that the buffer has enough len() rather than enough
capacity before deciding to allocate a new buffer.
---
 index/scorch/segment/zap/contentcoder.go | 8 +++++---
 index/scorch/segment/zap/docvalues.go    | 4 +++-
 index/scorch/segment/zap/merge.go        | 3 +--
 index/scorch/segment/zap/new.go          | 3 +--
 4 files changed, 10 insertions(+), 8 deletions(-)

diff --git a/index/scorch/segment/zap/contentcoder.go b/index/scorch/segment/zap/contentcoder.go
index 1e7a785c..2148d1d4 100644
--- a/index/scorch/segment/zap/contentcoder.go
+++ b/index/scorch/segment/zap/contentcoder.go
@@ -42,6 +42,8 @@ type chunkedContentCoder struct {
 	chunkBuf     bytes.Buffer
 
 	chunkMeta []MetaData
+
+	compressed []byte // temp buf for snappy compression
 }
 
 // MetaData represents the data information inside a
@@ -105,10 +107,10 @@ func (c *chunkedContentCoder) flushContents() error {
 	metaData := c.chunkMetaBuf.Bytes()
 	c.final = append(c.final, c.chunkMetaBuf.Bytes()...)
 	// write the compressed data to the final data
-	compressedData := snappy.Encode(nil, c.chunkBuf.Bytes())
-	c.final = append(c.final, compressedData...)
+	c.compressed = snappy.Encode(c.compressed[:cap(c.compressed)], c.chunkBuf.Bytes())
+	c.final = append(c.final, c.compressed...)
 
-	c.chunkLens[c.currChunk] = uint64(len(compressedData) + len(metaData))
+	c.chunkLens[c.currChunk] = uint64(len(c.compressed) + len(metaData))
 	return nil
 }
 
diff --git a/index/scorch/segment/zap/docvalues.go b/index/scorch/segment/zap/docvalues.go
index 84427190..dcd2cb05 100644
--- a/index/scorch/segment/zap/docvalues.go
+++ b/index/scorch/segment/zap/docvalues.go
@@ -42,6 +42,7 @@ type docValueIterator struct {
 	dvDataLoc      uint64
 	curChunkHeader []MetaData
 	curChunkData   []byte // compressed data cache
+	uncompressed   []byte // temp buf for snappy decompression
 }
 
 func (di *docValueIterator) size() int {
@@ -135,10 +136,11 @@ func (di *docValueIterator) visitDocValues(docNum uint64,
 		return nil
 	}
 	// uncompress the already loaded data
-	uncompressed, err := snappy.Decode(nil, di.curChunkData)
+	uncompressed, err := snappy.Decode(di.uncompressed[:cap(di.uncompressed)], di.curChunkData)
 	if err != nil {
 		return err
 	}
+	di.uncompressed = uncompressed
 
 	// pick the terms for the given docNum
 	uncompressed = uncompressed[start:end]
diff --git a/index/scorch/segment/zap/merge.go b/index/scorch/segment/zap/merge.go
index 6faca646..167ebfa2 100644
--- a/index/scorch/segment/zap/merge.go
+++ b/index/scorch/segment/zap/merge.go
@@ -604,7 +604,6 @@ func mergeStoredAndRemap(segments []*SegmentBase, drops []*roaring.Bitmap,
 			curr = 0
 			metaBuf.Reset()
 			data = data[:0]
-			compressed = compressed[:0]
 
 			// collect all the data
 			for i := 0; i < len(fieldsInv); i++ {
@@ -641,7 +640,7 @@ func mergeStoredAndRemap(segments []*SegmentBase, drops []*roaring.Bitmap,
 			metaEncoder.Close()
 			metaBytes := metaBuf.Bytes()
 
-			compressed = snappy.Encode(compressed, data)
+			compressed = snappy.Encode(compressed[:cap(compressed)], data)
 
 			// record where we're about to start writing
 			docNumOffsets[newDocNum] = uint64(w.Count())
diff --git a/index/scorch/segment/zap/new.go b/index/scorch/segment/zap/new.go
index 5837436f..83a0cbce 100644
--- a/index/scorch/segment/zap/new.go
+++ b/index/scorch/segment/zap/new.go
@@ -517,7 +517,6 @@ func (s *interim) writeStoredFields() (
 
 		s.metaBuf.Reset()
 		data = data[:0]
-		compressed = compressed[:0]
 
 		for fieldID := range s.FieldsInv {
 			isf, exists := docStoredFields[uint16(fieldID)]
@@ -534,7 +533,7 @@ func (s *interim) writeStoredFields() (
 		metaEncoder.Close()
 		metaBytes := s.metaBuf.Bytes()
 
-		compressed = snappy.Encode(compressed, data)
+		compressed = snappy.Encode(compressed[:cap(compressed)], data)
 
 		docStoredOffsets[docNum] = uint64(s.w.Count())