Revert "Fix(builder): Too many small tables when compression is enabl…

…ed (#1549)"

This reverts commit 5d1bab4.
We'reverting this commit because it seems to cause a strange issue while
writing data. Running `go run -tags main.go benchmark write --sorted
--compression=true --block-cache-mb=100` creates a directory which does
not have all the keys. I don't know why this would fix the issue but the
test works fine after reverting this commit.

levels.go

@@ -585,7 +585,7 @@ nextTable:
 			}
 
 			if !y.SameKey(it.Key(), lastKey) {
-				if builder.ReachedCapacity(uint64(float64(s.kv.opt.MaxTableSize) * 0.9)) {
+				if builder.ReachedCapacity(s.kv.opt.MaxTableSize) {
 					// Only break if we are on a different key, and have reached capacity. We want
 					// to ensure that all versions of the key are stored in the same sstable, and
 					// not divided across multiple tables at the same level.

stream_writer.go

@@ -320,7 +320,7 @@ func (w *sortedWriter) Add(key []byte, vs y.ValueStruct) error {
 
 	sameKey := y.SameKey(key, w.lastKey)
 	// Same keys should go into the same SSTable.
-	if !sameKey && w.builder.ReachedCapacity(uint64(float64(w.db.opt.MaxTableSize)*0.9)) {
+	if !sameKey && w.builder.ReachedCapacity(w.db.opt.MaxTableSize) {
 		if err := w.send(false); err != nil {
 			return err
 		}

table/builder.go

@@ -21,7 +21,6 @@ import (
 	"math"
 	"runtime"
 	"sync"
-	"sync/atomic"
 	"unsafe"
 
 	"github.com/dgraph-io/badger/v2/fb"
@@ -77,10 +76,9 @@ type bblock struct {
 // Builder is used in building a table.
 type Builder struct {
 	// Typically tens or hundreds of meg. This is for one single file.
-	buf        []byte
-	sz         uint32
-	bufLock    sync.Mutex // This lock guards the buf. We acquire lock when we resize the buf.
-	actualSize uint32     // Used to store the sum of sizes of blocks after compression/encryption.
+	buf     []byte
+	sz      uint32
+	bufLock sync.Mutex // This lock guards the buf. We acquire lock when we resize the buf.
 
 	baseKey    []byte // Base key for the current block.
 	baseOffset uint32 // Offset for the current block.
@@ -158,9 +156,6 @@ func (b *Builder) handleBlock() {
 		copy(b.buf[item.start:], blockBuf)
 		b.bufLock.Unlock()
 
-		// Add the actual size of current block.
-		atomic.AddUint32(&b.actualSize, uint32(len(blockBuf)))
-
 		// Fix the boundary of the block.
 		item.end = item.start + uint32(len(blockBuf))
 
@@ -290,7 +285,6 @@ func (b *Builder) finishBlock() {
 	// If compression/encryption is disabled, no need to send the block to the blockChan.
 	// There's nothing to be done.
 	if b.blockChan == nil {
-		atomic.StoreUint32(&b.actualSize, b.sz)
 		b.addBlockToIndex()
 		return
 	}
@@ -371,8 +365,8 @@ func (b *Builder) Add(key []byte, value y.ValueStruct, valueLen uint32) {
 // at the end. The diff can vary.
 
 // ReachedCapacity returns true if we... roughly (?) reached capacity?
-func (b *Builder) ReachedCapacity(capacity uint64) bool {
-	blocksSize := atomic.LoadUint32(&b.actualSize) + // actual length of current buffer
+func (b *Builder) ReachedCapacity(cap int64) bool {
+	blocksSize := b.sz + // length of current buffer
 		uint32(len(b.entryOffsets)*4) + // all entry offsets size
 		4 + // count of all entry offsets
 		8 + // checksum bytes
@@ -382,7 +376,7 @@ func (b *Builder) ReachedCapacity(capacity uint64) bool {
 		4 + // Index length
 		uint32(b.offsets.Len())
 
-	return uint64(estimateSz) > capacity
+	return int64(estimateSz) > cap
 }
 
 // Finish finishes the table by appending the index.