Add RebuildIndex() func

by skipping store blocks

README.md

@@ -4,14 +4,15 @@ Simple key-value store for single-user applications.
 
 ## Pros
 
-* Simple file structure
+* Simple two file structure (data file and index file)
-* Internal compression
+* Internal Zstandard compression by [klauspost/compress/zstd](https://github.com/klauspost/compress/tree/master/zstd)
 * Threadsafe operations through `sync.RWMutex`
 
 ## Cons
 
 * Index stored in memory (`map[key hash (28 bytes)]file offset (int64)`)
-* Need to read the whole file on store open to create file index
+* No transaction system
+* Index file is fully rewrited on every store commit
 * No way to recover disk space from deleted records
 * Write/Delete operations block Read and each other operations
 
@@ -20,7 +21,7 @@ Simple key-value store for single-user applications.
 Create or open existing file:
 
 ```go
-db, err := Open("path to file")
+db, err := zkv.Open("path to file")
 ```
 
 Data operations:
@@ -31,12 +32,41 @@ err = db.Set(key, value) // key and value can be any of type
 
 // Read data
 var value ValueType
-err = db.Get(key)
+err = db.Get(key, &value)
 
 // Delete data
 err = db.Delete(key)
 ```
 
+Other methods:
+
+```go
+// Flush data to disk
+err = db.Flush()
+
+// Backup data to another file
+err = db.Backup("new/file/path")
+```
+
+## Store options
+
+```go
+type Options struct {
+	// Maximum number of concurrent reads
+	MaxParallelReads int
+
+	// Compression level
+	CompressionLevel zstd.EncoderLevel
+
+	// Memory write buffer size in bytes
+	MemoryBufferSize int
+
+	// Disk write buffer size in bytes
+	DiskBufferSize int
+}
+
+```
+
 ## File structure
 
 Record is `encoding/gob` structure:
@@ -53,3 +83,26 @@ File is log stuctured list of commands:
 | -------| ------------------------ | -------- |
 | Length | Record body bytes length | int64    |
 | Body   | Gob-encoded record       | variable |
+
+Index file is simple gob-encoded map:
+
+```go
+map[string]struct {
+	BlockOffset  int64
+	RecordOffset int64
+}
+```
+
+where map key is data key hash and value - data offset in data file.
+
+## Resource consumption
+
+Store requirements:
+
+* around 300 Mb of RAM per 1 million of keys
+* around 34 Mb of disk space for index file per 1 million of keys
+
+## TODO
+
+- [ ] Add recovery previous state of store file on write error
+- [ ] Add method for index rebuild

defaults.go

@@ -0,0 +1,17 @@
+package zkv
+
+import (
+	"runtime"
+
+	"github.com/klauspost/compress/zstd"
+)
+
+var defaultOptions = Options{
+	MaxParallelReads: runtime.NumCPU(),
+	CompressionLevel: zstd.SpeedDefault,
+	MemoryBufferSize: 4 * 1024 * 1024,
+	DiskBufferSize:   1 * 1024 * 1024,
+	useIndexFile:     true,
+}
+
+const indexFileExt = ".idx"

go.mod

@@ -3,8 +3,8 @@ module github.com/nxshock/zkv
 go 1.19
 
 require (
-	github.com/klauspost/compress v1.15.12
+	github.com/klauspost/compress v1.16.4
-	github.com/stretchr/testify v1.8.1
+	github.com/stretchr/testify v1.8.2
 )
 
 require (

go.sum

@@ -1,8 +1,8 @@
 github.com/davecgh/go-spew v1.1.0/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
 github.com/davecgh/go-spew v1.1.1 h1:vj9j/u1bqnvCEfJOwUhtlOARqs3+rkHYY13jYWTU97c=
 github.com/davecgh/go-spew v1.1.1/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
-github.com/klauspost/compress v1.15.12 h1:YClS/PImqYbn+UILDnqxQCZ3RehC9N318SU3kElDUEM=
+github.com/klauspost/compress v1.16.4 h1:91KN02FnsOYhuunwU4ssRe8lc2JosWmizWa91B5v1PU=
-github.com/klauspost/compress v1.15.12/go.mod h1:QPwzmACJjUTFsnSHH934V6woptycfrDDJnH7hvFVbGM=
+github.com/klauspost/compress v1.16.4/go.mod h1:ntbaceVETuRiXiv4DpjP66DpAtAGkEQskQzEyD//IeE=
 github.com/pmezard/go-difflib v1.0.0 h1:4DBwDE0NGyQoBHbLQYPwSUPoCMWR5BEzIk/f1lZbAQM=
 github.com/pmezard/go-difflib v1.0.0/go.mod h1:iKH77koFhYxTK1pcRnkKkqfTogsbg7gZNVY4sRDYZ/4=
 github.com/stretchr/objx v0.1.0/go.mod h1:HFkY916IF+rwdDfMAkV7OtwuqBVzrE8GR6GFx+wExME=
@@ -10,8 +10,8 @@ github.com/stretchr/objx v0.4.0/go.mod h1:YvHI0jy2hoMjB+UWwv71VJQ9isScKT/TqJzVSS
 github.com/stretchr/objx v0.5.0/go.mod h1:Yh+to48EsGEfYuaHDzXPcE3xhTkx73EhmCGUpEOglKo=
 github.com/stretchr/testify v1.7.1/go.mod h1:6Fq8oRcR53rry900zMqJjRRixrwX3KX962/h/Wwjteg=
 github.com/stretchr/testify v1.8.0/go.mod h1:yNjHg4UonilssWZ8iaSj1OCr/vHnekPRkoO+kdMU+MU=
-github.com/stretchr/testify v1.8.1 h1:w7B6lhMri9wdJUVmEZPGGhZzrYTPvgJArz7wNPgYKsk=
+github.com/stretchr/testify v1.8.2 h1:+h33VjcLVPDHtOdpUCuF+7gSuG3yGIftsP1YvFihtJ8=
-github.com/stretchr/testify v1.8.1/go.mod h1:w2LPCIKwWwSfY2zedu0+kehJoqGctiVI29o6fzry7u4=
+github.com/stretchr/testify v1.8.2/go.mod h1:w2LPCIKwWwSfY2zedu0+kehJoqGctiVI29o6fzry7u4=
 gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405 h1:yhCVgyC4o1eVCa2tZl7eS0r+SDo693bJlVdllGtEeKM=
 gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405/go.mod h1:Co6ibVJAznAaIkqp8huTwlJQCZ016jof/cbN4VW5Yz0=
 gopkg.in/yaml.v3 v3.0.0-20200313102051-9f266ea9e77c/go.mod h1:K4uyk7z7BCEPqu6E+C64Yfv1cQ7kz7rIZviUmN+EgEM=

options.go

@@ -0,0 +1,40 @@
+package zkv
+
+import "github.com/klauspost/compress/zstd"
+
+type Options struct {
+	// Maximum number of concurrent reads
+	MaxParallelReads int
+
+	// Compression level
+	CompressionLevel zstd.EncoderLevel
+
+	// Memory write buffer size in bytes
+	MemoryBufferSize int
+
+	// Disk write buffer size in bytes
+	DiskBufferSize int
+
+	// Use index file
+	useIndexFile bool
+}
+
+func (o *Options) setDefaults() {
+	o.useIndexFile = true // TODO: implement database search without index
+
+	if o.MaxParallelReads == 0 {
+		o.MaxParallelReads = defaultOptions.MaxParallelReads
+	}
+
+	if o.CompressionLevel == 0 {
+		o.CompressionLevel = defaultOptions.CompressionLevel
+	}
+
+	if o.MemoryBufferSize == 0 {
+		o.MemoryBufferSize = defaultOptions.MemoryBufferSize
+	}
+
+	if o.DiskBufferSize == 0 {
+		o.DiskBufferSize = defaultOptions.DiskBufferSize
+	}
+}

record.go

@@ -2,6 +2,7 @@ package zkv
 
 import (
 	"bytes"
+	"crypto/sha256"
 	"encoding/binary"
 	"encoding/gob"
 	"io"
@@ -20,8 +21,17 @@ type Record struct {
 	ValueBytes []byte
 }
 
+func newRecordBytes(recordType RecordType, keyHash [sha256.Size224]byte, valueBytes []byte) (*Record, error) {
+	record := &Record{
+		Type:       recordType,
+		KeyHash:    keyHash,
+		ValueBytes: valueBytes}
+
+	return record, nil
+}
+
 func newRecord(recordType RecordType, key, value interface{}) (*Record, error) {
-	keyBytes, err := encode(key)
+	keyHash, err := hashInterface(key)
 	if err != nil {
 		return nil, err
 	}
@@ -31,12 +41,7 @@ func newRecord(recordType RecordType, key, value interface{}) (*Record, error) {
 		return nil, err
 	}
 
-	record := &Record{
+	return newRecordBytes(recordType, keyHash, valueBytes)
-		Type:       recordType,
-		KeyHash:    hashBytes(keyBytes),
-		ValueBytes: valueBytes}
-
-	return record, nil
 }
 
 func (r *Record) Marshal() ([]byte, error) {

record_test.go

@@ -0,0 +1,35 @@
+package zkv
+
+import (
+	"bytes"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+)
+
+func TestRecord(t *testing.T) {
+	buf := new(bytes.Buffer)
+
+	var records []Record
+
+	for i := 0; i < 10; i++ {
+		record, err := newRecord(RecordTypeSet, i, i)
+		assert.NoError(t, err)
+
+		records = append(records, *record)
+
+		b, err := record.Marshal()
+		assert.NoError(t, err)
+
+		_, err = buf.Write(b)
+		assert.NoError(t, err)
+	}
+
+	for i := 0; i < 10; i++ {
+		_, record, err := readRecord(buf)
+		assert.NoError(t, err)
+
+		assert.Equal(t, record.KeyHash, records[i].KeyHash)
+		assert.Equal(t, record.ValueBytes, records[i].ValueBytes)
+	}
+}

utils.go

@@ -4,7 +4,9 @@ import (
 	"bytes"
 	"crypto/sha256"
 	"encoding/gob"
+	"errors"
 	"io"
+	"os"
 )
 
 func encode(value interface{}) ([]byte, error) {
@@ -40,3 +42,13 @@ func skip(r io.Reader, count int64) (err error) {
 
 	return err
 }
+
+func isFileExists(filePath string) (bool, error) {
+	if _, err := os.Stat(filePath); err == nil {
+		return true, nil
+	} else if errors.Is(err, os.ErrNotExist) {
+		return false, nil
+	} else {
+		return false, err
+	}
+}

zkv.go

@@ -1,8 +1,11 @@
 package zkv
 
 import (
+	"bufio"
 	"bytes"
+	"crypto/sha256"
 	"encoding/base64"
+	"encoding/gob"
 	"fmt"
 	"io"
 	"os"
@@ -11,156 +14,88 @@ import (
 	"github.com/klauspost/compress/zstd"
 )
 
-type Database struct {
+type Offsets struct {
-	dataOffset map[string]int64
+	BlockOffset  int64
-	file       *os.File
+	RecordOffset int64
-	compressor *zstd.Encoder
+}
-	filePath   string
+
-	offset     int64
+type Store struct {
+	dataOffset map[string]Offsets
+
+	filePath string
+
+	buffer           *bytes.Buffer
+	bufferDataOffset map[string]int64
+
+	options Options
+
+	readOrderChan chan struct{}
 
 	mu sync.RWMutex
 }
 
-func (db *Database) Close() error {
+func OpenWithOptions(filePath string, options Options) (*Store, error) {
-	db.mu.Lock()
+	options.setDefaults()
-	defer db.mu.Unlock()
 
-	err := db.compressor.Close()
+	store := &Store{
-	if err != nil {
+		dataOffset:       make(map[string]Offsets),
-		return err
+		bufferDataOffset: make(map[string]int64),
+		buffer:           new(bytes.Buffer),
+		filePath:         filePath,
+		options:          options,
+		readOrderChan:    make(chan struct{}, int(options.MaxParallelReads))}
+
+	if options.useIndexFile {
+		idxFile, err := os.Open(filePath + indexFileExt)
+		if err == nil {
+			err = gob.NewDecoder(idxFile).Decode(&store.dataOffset)
+			if err != nil {
+				return nil, err
+			}
+
+			return store, nil
+		}
 	}
 
-	return db.file.Close()
+	exists, err := isFileExists(filePath)
-}
-
-func (db *Database) Set(key, value interface{}) error {
-	db.mu.Lock()
-	defer db.mu.Unlock()
-
-	record, err := newRecord(RecordTypeSet, key, value)
 	if err != nil {
-		return err
+		return nil, err
 	}
 
-	b, err := record.Marshal()
-	if err != nil {
-		return err
-	}
-
-	db.dataOffset[string(record.KeyHash[:])] = db.offset // TODO: удалить хеш и откатить запись в случае ошибки
-
-	_, err = db.compressor.Write(b)
-	if err != nil {
-		return err
-	}
-
-	db.offset += int64(len(b)) // TODO: удалить хеш и откатить запись в случае ошибки
-
-	return nil
-}
-
-func (db *Database) Get(key, value interface{}) error {
-	db.mu.RLock()
-	defer db.mu.RUnlock()
-
-	hashToFind, err := hashInterface(key)
-	if err != nil {
-		return err
-	}
-
-	offset, exists := db.dataOffset[string(hashToFind[:])]
 	if !exists {
-		return ErrNotExists
+		return store, nil
 	}
 
-	readF, err := os.Open(db.filePath)
+	err = store.rebuildIndex()
 	if err != nil {
-		return err
+		return nil, err
-	}
-	defer readF.Close()
-
-	decompressor, err := zstd.NewReader(readF)
-	if err != nil {
-		return err
-	}
-	defer decompressor.Close()
-
-	err = skip(decompressor, offset)
-	if err != nil {
-		return err
 	}
 
-	_, record, err := readRecord(decompressor)
+	return store, nil
-	if err != nil {
-		return err
-	}
-
-	if !bytes.Equal(record.KeyHash[:], hashToFind[:]) {
-		return fmt.Errorf("wrong hash on this offset: expected %s, got %s", base64.StdEncoding.EncodeToString(hashToFind[:]), base64.StdEncoding.EncodeToString(record.KeyHash[:])) // TODO: заменить на константную ошибку
-	}
-
-	return decode(record.ValueBytes, value)
 }
 
-func Open(filePath string) (*Database, error) {
+func Open(filePath string) (*Store, error) {
-	f, err := os.OpenFile(filePath, os.O_APPEND|os.O_CREATE|os.O_WRONLY, 0644)
+	options := defaultOptions
-	if err != nil {
+	return OpenWithOptions(filePath, options)
-		return nil, fmt.Errorf("ошибка при открытии файла для записи: %v", err)
-	}
-
-	compressor, err := zstd.NewWriter(f)
-	if err != nil {
-		return nil, fmt.Errorf("ошибка при инициализации компрессора: %v", err)
-	}
-
-	database := &Database{
-		dataOffset: make(map[string]int64),
-		offset:     0,
-		file:       f,
-		compressor: compressor,
-		filePath:   filePath}
-
-	// restore file data
-	readF, err := os.Open(filePath)
-	if err != nil {
-		f.Close()
-		return nil, fmt.Errorf("ошибка при открытии файла для чтения: %v", err)
-	}
-	defer readF.Close()
-
-	decompressor, err := zstd.NewReader(readF)
-	if err != nil {
-		return nil, fmt.Errorf("ошибка при инициализации декомпрессора: %v", err)
-	}
-	defer decompressor.Close()
-
-	offset := int64(0)
-	for {
-		n, record, err := readRecord(decompressor)
-		if err == io.EOF {
-			break
-		}
-		if err != nil {
-			f.Close()
-			return nil, fmt.Errorf("ошибка при чтении записи из файла: %v", err)
-		}
-
-		switch record.Type {
-		case RecordTypeSet:
-			database.dataOffset[string(record.KeyHash[:])] = offset
-		case RecordTypeDelete:
-			delete(database.dataOffset, string(record.KeyHash[:]))
-		}
-
-		offset += n
-	}
-
-	return database, nil
 }
 
-func (db *Database) Delete(key interface{}) error {
+func (s *Store) Set(key, value interface{}) error {
-	db.mu.Lock()
+	s.mu.Lock()
-	defer db.mu.Unlock()
+	defer s.mu.Unlock()
+
+	return s.set(key, value)
+}
+
+func (s *Store) Get(key, value interface{}) error {
+	s.mu.RLock()
+	defer s.mu.RUnlock()
+
+	return s.get(key, value)
+}
+
+func (s *Store) Delete(key interface{}) error {
+	s.mu.Lock()
+	defer s.mu.Unlock()
 
 	keyHash, err := hashInterface(key)
 	if err != nil {
@@ -177,14 +112,397 @@ func (db *Database) Delete(key interface{}) error {
 		return err
 	}
 
-	delete(db.dataOffset, string(record.KeyHash[:]))
+	delete(s.dataOffset, string(record.KeyHash[:]))
+	delete(s.bufferDataOffset, string(record.KeyHash[:]))
 
-	_, err = db.compressor.Write(b)
+	_, err = s.buffer.Write(b)
 	if err != nil {
 		return err
 	}
 
-	db.offset += int64(len(b))
+	if s.buffer.Len() > s.options.MemoryBufferSize {
+		err = s.flush()
+
+		if err != nil {
+			return err
+		}
+	}
 
 	return nil
 }
+
+func (s *Store) Flush() error {
+	s.mu.Lock()
+	defer s.mu.Unlock()
+
+	return s.flush()
+}
+
+func (s *Store) BackupWithOptions(filePath string, newFileOptions Options) error {
+	s.mu.Lock()
+	defer s.mu.Unlock()
+
+	err := s.flush()
+	if err != nil {
+		return err
+	}
+
+	newStore, err := OpenWithOptions(filePath, newFileOptions)
+	if err != nil {
+		return err
+	}
+
+	for keyHashStr := range s.dataOffset {
+		var keyHash [sha256.Size224]byte
+		copy(keyHash[:], keyHashStr)
+
+		valueBytes, err := s.getGobBytes(keyHash)
+		if err != nil {
+			newStore.Close()
+			return err
+		}
+		err = newStore.setBytes(keyHash, valueBytes)
+		if err != nil {
+			newStore.Close()
+			return err
+		}
+	}
+
+	return newStore.Close()
+}
+
+func (s *Store) Backup(filePath string) error {
+	return s.BackupWithOptions(filePath, defaultOptions)
+}
+
+func (s *Store) Close() error {
+	s.mu.Lock()
+	defer s.mu.Unlock()
+
+	err := s.flush()
+	if err != nil {
+		return err
+	}
+
+	return nil
+}
+
+func (s *Store) setBytes(keyHash [sha256.Size224]byte, valueBytes []byte) error {
+	record, err := newRecordBytes(RecordTypeSet, keyHash, valueBytes)
+	if err != nil {
+		return err
+	}
+
+	b, err := record.Marshal()
+	if err != nil {
+		return err
+	}
+
+	s.bufferDataOffset[string(record.KeyHash[:])] = int64(s.buffer.Len())
+
+	_, err = s.buffer.Write(b)
+	if err != nil {
+		return err
+	}
+
+	if s.buffer.Len() > s.options.MemoryBufferSize {
+		err = s.flush()
+
+		if err != nil {
+			return err
+		}
+	}
+
+	return nil
+}
+
+func (s *Store) set(key, value interface{}) error {
+	record, err := newRecord(RecordTypeSet, key, value)
+	if err != nil {
+		return err
+	}
+
+	b, err := record.Marshal()
+	if err != nil {
+		return err
+	}
+
+	s.bufferDataOffset[string(record.KeyHash[:])] = int64(s.buffer.Len())
+
+	_, err = s.buffer.Write(b)
+	if err != nil {
+		return err
+	}
+
+	if s.buffer.Len() > s.options.MemoryBufferSize {
+		err = s.flush()
+
+		if err != nil {
+			return err
+		}
+	}
+
+	return nil
+}
+
+func (s *Store) getGobBytes(keyHash [sha256.Size224]byte) ([]byte, error) {
+	s.readOrderChan <- struct{}{}
+	defer func() { <-s.readOrderChan }()
+
+	offset, exists := s.bufferDataOffset[string(keyHash[:])]
+	if exists {
+		reader := bytes.NewReader(s.buffer.Bytes())
+
+		err := skip(reader, offset)
+		if err != nil {
+			return nil, err
+		}
+
+		_, record, err := readRecord(reader)
+		if err != nil {
+			return nil, err
+		}
+
+		return record.ValueBytes, nil
+	}
+
+	offsets, exists := s.dataOffset[string(keyHash[:])]
+	if !exists {
+		return nil, ErrNotExists
+	}
+
+	readF, err := os.Open(s.filePath)
+	if err != nil {
+		return nil, err
+	}
+	defer readF.Close()
+
+	_, err = readF.Seek(offsets.BlockOffset, io.SeekStart)
+	if err != nil {
+		return nil, err
+	}
+
+	decompressor, err := zstd.NewReader(readF)
+	if err != nil {
+		return nil, err
+	}
+	defer decompressor.Close()
+
+	err = skip(decompressor, offsets.RecordOffset)
+	if err != nil {
+		return nil, err
+	}
+
+	_, record, err := readRecord(decompressor)
+	if err != nil {
+		return nil, err
+	}
+
+	if !bytes.Equal(record.KeyHash[:], keyHash[:]) {
+		expectedHashStr := base64.StdEncoding.EncodeToString(keyHash[:])
+		gotHashStr := base64.StdEncoding.EncodeToString(record.KeyHash[:])
+		return nil, fmt.Errorf("wrong hash of offset %d: expected %s, got %s", offset, expectedHashStr, gotHashStr)
+	}
+
+	return record.ValueBytes, nil
+}
+
+func (s *Store) get(key, value interface{}) error {
+	s.readOrderChan <- struct{}{}
+	defer func() { <-s.readOrderChan }()
+
+	hashToFind, err := hashInterface(key)
+	if err != nil {
+		return err
+	}
+
+	b, err := s.getGobBytes(hashToFind)
+	if err != nil {
+		return err
+	}
+
+	return decode(b, value)
+}
+
+func (s *Store) flush() error {
+	l := int64(s.buffer.Len())
+
+	f, err := os.OpenFile(s.filePath, os.O_APPEND|os.O_CREATE|os.O_WRONLY, 0644)
+	if err != nil {
+		return fmt.Errorf("open store file: %v", err)
+	}
+	stat, err := f.Stat()
+	if err != nil {
+		f.Close()
+		return fmt.Errorf("stat store file: %v", err)
+	}
+
+	diskWriteBuffer := bufio.NewWriterSize(f, s.options.DiskBufferSize)
+
+	encoder, err := zstd.NewWriter(diskWriteBuffer, zstd.WithEncoderLevel(s.options.CompressionLevel))
+	if err != nil {
+		f.Close()
+		return fmt.Errorf("init encoder: %v", err)
+	}
+
+	_, err = s.buffer.WriteTo(encoder)
+	if err != nil {
+		return err
+	}
+
+	for key, val := range s.bufferDataOffset {
+		s.dataOffset[key] = Offsets{BlockOffset: stat.Size(), RecordOffset: val}
+	}
+
+	s.bufferDataOffset = make(map[string]int64)
+
+	err = encoder.Close()
+	if err != nil {
+		// TODO: truncate file to previous state
+		return err
+	}
+
+	err = diskWriteBuffer.Flush()
+	if err != nil {
+		// TODO: truncate file to previous state
+		return err
+	}
+
+	err = f.Close()
+	if err != nil {
+		return err
+	}
+
+	// Update index file only on data update
+	if s.options.useIndexFile && l > 0 {
+		err = s.saveIndex()
+		if err != nil {
+			return err
+		}
+	}
+
+	return nil
+}
+
+func readBlock(r *bufio.Reader) (line []byte, n int, err error) {
+	delim := []byte{0x28, 0xb5, 0x2f, 0xfd}
+
+	line = make([]byte, len(delim))
+	copy(line, delim)
+
+	for {
+		s, err := r.ReadBytes(delim[len(delim)-1])
+		line = append(line, []byte(s)...)
+		if err != nil {
+			if bytes.Equal(line, delim) { // contains only magic number
+				return []byte{}, 0, err
+			} else {
+				return line, len(s), err
+			}
+		}
+
+		if bytes.Equal(line, append(delim, delim...)) { // first block
+			line = make([]byte, len(delim))
+			copy(line, delim)
+			continue
+		}
+
+		if bytes.HasSuffix(line, delim) {
+			return line[:len(line)-len(delim)], len(s), nil
+		}
+	}
+}
+
+// RebuildIndex renews index from store file
+func (s *Store) RebuildIndex() error {
+	s.mu.Lock()
+	defer s.mu.Unlock()
+
+	err := s.rebuildIndex()
+	if err != nil {
+		return err
+	}
+
+	if s.options.useIndexFile {
+		return s.saveIndex()
+	}
+
+	return nil
+}
+
+func (s *Store) rebuildIndex() error {
+	f, err := os.Open(s.filePath)
+	if err != nil {
+		return err
+	}
+	defer f.Close()
+
+	r := bufio.NewReader(f)
+
+	var blockOffset int64
+
+	s.dataOffset = make(map[string]Offsets)
+
+	for {
+		l, n, err := readBlock(r)
+		if err != nil {
+			if err != io.EOF {
+				return err
+			} else if err == io.EOF && len(l) == 0 {
+				break
+			}
+		}
+
+		dec, err := zstd.NewReader(bytes.NewReader(l))
+
+		var recordOffset int64
+		for {
+			n, record, err := readRecord(dec)
+			if err != nil {
+				if err == io.EOF {
+					break
+				} else {
+					return err
+				}
+			}
+
+			switch record.Type {
+			case RecordTypeSet:
+				s.dataOffset[string(record.KeyHash[:])] = Offsets{BlockOffset: blockOffset, RecordOffset: recordOffset}
+			case RecordTypeDelete:
+				delete(s.dataOffset, string(record.KeyHash[:]))
+			}
+			recordOffset += n
+		}
+
+		blockOffset += int64(n)
+	}
+
+	idxBuf := new(bytes.Buffer)
+
+	err = gob.NewEncoder(idxBuf).Encode(s.dataOffset)
+	if err != nil {
+		return err
+	}
+
+	err = os.WriteFile(s.filePath+indexFileExt, idxBuf.Bytes(), 0644)
+	if err != nil {
+		return err
+	}
+
+	return nil
+}
+
+func (s *Store) saveIndex() error {
+	f, err := os.OpenFile(s.filePath+indexFileExt, os.O_WRONLY|os.O_CREATE|os.O_TRUNC, 0644)
+	if err != nil {
+		return err
+	}
+
+	err = gob.NewEncoder(f).Encode(s.dataOffset)
+	if err != nil {
+		return err
+	}
+
+	return f.Close()
+}

zkv_test.go

@@ -1,6 +1,8 @@
 package zkv
 
 import (
+	"bufio"
+	"io"
 	"os"
 	"testing"
 
@@ -11,6 +13,7 @@ func TestReadWriteBasic(t *testing.T) {
 	const filePath = "TestReadWriteBasic.zkv"
 	const recordCount = 100
 	defer os.Remove(filePath)
+	defer os.Remove(filePath + indexFileExt)
 
 	db, err := Open(filePath)
 	assert.NoError(t, err)
@@ -20,7 +23,16 @@ func TestReadWriteBasic(t *testing.T) {
 		assert.NoError(t, err)
 	}
 
-	assert.Len(t, db.dataOffset, recordCount)
+	assert.Len(t, db.dataOffset, 0)
+	assert.Len(t, db.bufferDataOffset, recordCount)
+
+	for i := 1; i <= recordCount; i++ {
+		var gotValue int
+
+		err = db.Get(i, &gotValue)
+		assert.NoError(t, err)
+		assert.Equal(t, i, gotValue)
+	}
 
 	err = db.Close()
 	assert.NoError(t, err)
@@ -47,6 +59,7 @@ func TestSmallWrites(t *testing.T) {
 	const filePath = "TestSmallWrites.zkv"
 	const recordCount = 100
 	defer os.Remove(filePath)
+	defer os.Remove(filePath + indexFileExt)
 
 	for i := 1; i <= recordCount; i++ {
 		db, err := Open(filePath)
@@ -82,6 +95,7 @@ func TestDeleteBasic(t *testing.T) {
 	const filePath = "TestDeleteBasic.zkv"
 	const recordCount = 100
 	defer os.Remove(filePath)
+	defer os.Remove(filePath + indexFileExt)
 
 	db, err := Open(filePath)
 	assert.NoError(t, err)
@@ -91,12 +105,15 @@ func TestDeleteBasic(t *testing.T) {
 		assert.NoError(t, err)
 	}
 
-	assert.Len(t, db.dataOffset, recordCount)
+	assert.Len(t, db.dataOffset, 0)
+	assert.Len(t, db.bufferDataOffset, recordCount)
 
 	err = db.Delete(50)
 	assert.NoError(t, err)
 
-	assert.Len(t, db.dataOffset, recordCount-1)
+	assert.Len(t, db.dataOffset, 0)
+	assert.Len(t, db.bufferDataOffset, recordCount-1)
+
 	var value int
 	err = db.Get(50, &value)
 	assert.Equal(t, 0, value)
@@ -110,6 +127,8 @@ func TestDeleteBasic(t *testing.T) {
 	assert.NoError(t, err)
 
 	assert.Len(t, db.dataOffset, recordCount-1)
+	assert.Len(t, db.bufferDataOffset, 0)
+
 	value = 0
 	err = db.Get(50, &value)
 	assert.Equal(t, 0, value)
@@ -118,3 +137,276 @@ func TestDeleteBasic(t *testing.T) {
 	err = db.Close()
 	assert.NoError(t, err)
 }
+
+func TestBufferBasic(t *testing.T) {
+	const filePath = "TestBuffer.zkv"
+	defer os.Remove(filePath)
+	defer os.Remove(filePath + indexFileExt)
+
+	db, err := OpenWithOptions(filePath, Options{MemoryBufferSize: 100})
+	assert.NoError(t, err)
+
+	err = db.Set(1, make([]byte, 100))
+	assert.NoError(t, err)
+
+	assert.NotEqual(t, 0, db.dataOffset)
+	assert.Len(t, db.bufferDataOffset, 0)
+	assert.Equal(t, 0, db.buffer.Len())
+
+	var gotValue []byte
+	err = db.Get(1, &gotValue)
+	assert.NoError(t, err)
+
+	assert.Equal(t, make([]byte, 100), gotValue)
+
+	err = db.Close()
+	assert.NoError(t, err)
+}
+
+func TestBufferRead(t *testing.T) {
+	const filePath = "TestBufferRead.zkv"
+	const recordCount = 2
+	defer os.Remove(filePath)
+	defer os.Remove(filePath + indexFileExt)
+
+	db, err := OpenWithOptions(filePath, Options{MemoryBufferSize: 100})
+	assert.NoError(t, err)
+
+	for i := 1; i <= recordCount; i++ {
+		err = db.Set(i, i)
+		assert.NoError(t, err)
+	}
+
+	for i := 1; i <= recordCount; i++ {
+		var gotValue int
+
+		err = db.Get(i, &gotValue)
+		assert.NoError(t, err)
+		assert.Equal(t, i, gotValue)
+	}
+
+	for i := 1; i <= recordCount; i++ {
+		var gotValue int
+
+		err = db.Get(i, &gotValue)
+		assert.NoError(t, err)
+		assert.Equal(t, i, gotValue)
+	}
+
+	err = db.Close()
+	assert.NoError(t, err)
+
+	// try to read
+	db, err = Open(filePath)
+	assert.NoError(t, err)
+
+	assert.Len(t, db.dataOffset, recordCount)
+
+	for i := 1; i <= recordCount; i++ {
+		var gotValue int
+
+		err = db.Get(i, &gotValue)
+		assert.NoError(t, err)
+		assert.Equal(t, i, gotValue)
+	}
+
+	err = db.Close()
+	assert.NoError(t, err)
+
+}
+
+func TestBackupBasic(t *testing.T) {
+	const filePath = "TestBackupBasic.zkv"
+	const newFilePath = "TestBackupBasic2.zkv"
+	const recordCount = 100
+	defer os.Remove(filePath)
+	defer os.Remove(filePath + indexFileExt)
+	defer os.Remove(newFilePath)
+	defer os.Remove(newFilePath + indexFileExt)
+
+	db, err := Open(filePath)
+	assert.NoError(t, err)
+
+	for i := 1; i <= recordCount; i++ {
+		err = db.Set(i, i)
+		assert.NoError(t, err)
+	}
+
+	err = db.Backup(newFilePath)
+	assert.NoError(t, err)
+
+	err = db.Close()
+	assert.NoError(t, err)
+
+	db, err = Open(newFilePath)
+	assert.NoError(t, err)
+
+	assert.Len(t, db.dataOffset, recordCount)
+
+	for i := 1; i <= recordCount; i++ {
+		var gotValue int
+
+		err = db.Get(i, &gotValue)
+		assert.NoError(t, err)
+		assert.Equal(t, i, gotValue)
+	}
+
+	err = db.Close()
+	assert.NoError(t, err)
+
+}
+
+func TestBackupWithDeletedRecords(t *testing.T) {
+	const filePath = "TestBackupWithDeletedRecords.zkv"
+	const newFilePath = "TestBackupWithDeletedRecords2.zkv"
+	const recordCount = 100
+	defer os.Remove(filePath)
+	defer os.Remove(filePath + indexFileExt)
+	defer os.Remove(newFilePath)
+	defer os.Remove(newFilePath + indexFileExt)
+
+	db, err := Open(filePath)
+	assert.NoError(t, err)
+
+	for i := 1; i <= recordCount; i++ {
+		err = db.Set(i, i)
+		assert.NoError(t, err)
+	}
+
+	err = db.Flush()
+	assert.NoError(t, err)
+
+	for i := 1; i <= recordCount; i++ {
+		if i%2 == 1 {
+			continue
+		}
+
+		err = db.Delete(i)
+		assert.NoError(t, err)
+	}
+
+	err = db.Backup(newFilePath)
+	assert.NoError(t, err)
+
+	err = db.Close()
+	assert.NoError(t, err)
+
+	db, err = Open(newFilePath)
+	assert.NoError(t, err)
+
+	assert.Len(t, db.dataOffset, recordCount/2)
+
+	for i := 1; i <= recordCount; i++ {
+		var gotValue int
+
+		err = db.Get(i, &gotValue)
+		if i%2 == 0 {
+			assert.ErrorIs(t, err, ErrNotExists)
+		} else {
+			assert.NoError(t, err)
+			assert.Equal(t, i, gotValue)
+		}
+	}
+
+	err = db.Close()
+	assert.NoError(t, err)
+}
+
+func TestIndexFileBasic(t *testing.T) {
+	const filePath = "TestReadWriteBasic.zkv"
+	const recordCount = 100
+	defer os.Remove(filePath)
+	defer os.Remove(filePath + indexFileExt)
+
+	db, err := Open(filePath)
+	assert.NoError(t, err)
+
+	for i := 1; i <= recordCount; i++ {
+		err = db.Set(i, i)
+		assert.NoError(t, err)
+	}
+
+	assert.Len(t, db.dataOffset, 0)
+	assert.Len(t, db.bufferDataOffset, recordCount)
+
+	for i := 1; i <= recordCount; i++ {
+		var gotValue int
+
+		err = db.Get(i, &gotValue)
+		assert.NoError(t, err)
+		assert.Equal(t, i, gotValue)
+	}
+
+	err = db.Close()
+	assert.NoError(t, err)
+
+	// try to read
+	db, err = Open(filePath)
+	assert.NoError(t, err)
+
+	assert.Len(t, db.dataOffset, recordCount)
+
+	for i := 1; i <= recordCount; i++ {
+		var gotValue int
+
+		err = db.Get(i, &gotValue)
+		assert.NoError(t, err)
+		assert.Equal(t, i, gotValue)
+	}
+
+	err = db.Close()
+	assert.NoError(t, err)
+}
+
+func TestReadBlock(t *testing.T) {
+	file, err := os.Open("testdata/TestReadBlock.zkv")
+	assert.NoError(t, err)
+	defer file.Close()
+
+	r := bufio.NewReader(file)
+
+	line, _, err := readBlock(r)
+	assert.Equal(t, []byte{0x28, 0xb5, 0x2f, 0xfd, 0x04, 0x00, 0x99, 0x04, 0x00, 0x8b, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x39, 0xff, 0x81, 0x03, 0x01, 0x01, 0x06, 0x52, 0x65, 0x63, 0x6f, 0x72, 0x64, 0x01, 0xff, 0x82, 0x00, 0x01, 0x03, 0x01, 0x04, 0x54, 0x79, 0x70, 0x65, 0x01, 0x06, 0x00, 0x01, 0x07, 0x4b, 0x65, 0x79, 0x48, 0x61, 0x73, 0x68, 0x01, 0xff, 0x84, 0x00, 0x01, 0x0a, 0x56, 0x61, 0x6c, 0x75, 0x65, 0x42, 0x79, 0x74, 0x65, 0x73, 0x01, 0x0a, 0x00, 0x00, 0x00, 0x19, 0xff, 0x83, 0x01, 0x01, 0x01, 0x09, 0x5b, 0x32, 0x38, 0x5d, 0x75, 0x69, 0x6e, 0x74, 0x38, 0x01, 0xff, 0x84, 0x00, 0x01, 0x06, 0x01, 0x38, 0x00, 0x00, 0x36, 0xff, 0x82, 0x01, 0x01, 0x01, 0x1c, 0xff, 0x90, 0xff, 0xf4, 0x25, 0x15, 0x70, 0x75, 0x5c, 0xff, 0xf4, 0xff, 0xbc, 0xff, 0xf9, 0xff, 0xde, 0xff, 0x93, 0xff, 0xf8, 0x0d, 0x0e, 0x78, 0x5b, 0xff, 0x81, 0xff, 0x95, 0x6e, 0xff, 0xab, 0x4b, 0xff, 0xe8, 0x37, 0xff, 0x97, 0x68, 0x41, 0x3d, 0x01, 0x04, 0x03, 0x04, 0x00, 0x02, 0x00, 0x25, 0xd5, 0x63, 0x21}, line)
+	line, _, err = readBlock(r)
+	assert.Equal(t, []byte{0x28, 0xb5, 0x2f, 0xfd, 0x04, 0x00, 0x89, 0x04, 0x00, 0x89, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x39, 0xff, 0x81, 0x03, 0x01, 0x01, 0x06, 0x52, 0x65, 0x63, 0x6f, 0x72, 0x64, 0x01, 0xff, 0x82, 0x00, 0x01, 0x03, 0x01, 0x04, 0x54, 0x79, 0x70, 0x65, 0x01, 0x06, 0x00, 0x01, 0x07, 0x4b, 0x65, 0x79, 0x48, 0x61, 0x73, 0x68, 0x01, 0xff, 0x84, 0x00, 0x01, 0x0a, 0x56, 0x61, 0x6c, 0x75, 0x65, 0x42, 0x79, 0x74, 0x65, 0x73, 0x01, 0x0a, 0x00, 0x00, 0x00, 0x19, 0xff, 0x83, 0x01, 0x01, 0x01, 0x09, 0x5b, 0x32, 0x38, 0x5d, 0x75, 0x69, 0x6e, 0x74, 0x38, 0x01, 0xff, 0x84, 0x00, 0x01, 0x06, 0x01, 0x38, 0x00, 0x00, 0x34, 0xff, 0x82, 0x01, 0x01, 0x01, 0x1c, 0xff, 0x84, 0xff, 0x84, 0xff, 0xc1, 0x21, 0x02, 0xff, 0x8b, 0xff, 0xd7, 0x6d, 0xff, 0xd0, 0xff, 0xad, 0x1a, 0x55, 0x14, 0x5c, 0xff, 0xb1, 0x04, 0x37, 0x29, 0x2f, 0x78, 0x18, 0xff, 0xb5, 0xff, 0xe4, 0x56, 0x4e, 0xff, 0x8d, 0x19, 0x46, 0x01, 0x04, 0x03, 0x04, 0x00, 0x04, 0x00, 0x0c, 0x3b, 0xbf, 0x39}, line)
+	line, _, err = readBlock(r)
+	assert.Equal(t, []byte{0x28, 0xb5, 0x2f, 0xfd, 0x04, 0x00, 0x99, 0x04, 0x00, 0x8b, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x39, 0xff, 0x81, 0x03, 0x01, 0x01, 0x06, 0x52, 0x65, 0x63, 0x6f, 0x72, 0x64, 0x01, 0xff, 0x82, 0x00, 0x01, 0x03, 0x01, 0x04, 0x54, 0x79, 0x70, 0x65, 0x01, 0x06, 0x00, 0x01, 0x07, 0x4b, 0x65, 0x79, 0x48, 0x61, 0x73, 0x68, 0x01, 0xff, 0x84, 0x00, 0x01, 0x0a, 0x56, 0x61, 0x6c, 0x75, 0x65, 0x42, 0x79, 0x74, 0x65, 0x73, 0x01, 0x0a, 0x00, 0x00, 0x00, 0x19, 0xff, 0x83, 0x01, 0x01, 0x01, 0x09, 0x5b, 0x32, 0x38, 0x5d, 0x75, 0x69, 0x6e, 0x74, 0x38, 0x01, 0xff, 0x84, 0x00, 0x01, 0x06, 0x01, 0x38, 0x00, 0x00, 0x36, 0xff, 0x82, 0x01, 0x01, 0x01, 0x1c, 0x25, 0x79, 0x3e, 0x46, 0x4e, 0xff, 0xac, 0x06, 0x27, 0xff, 0xb1, 0xff, 0xa3, 0xff, 0xaa, 0xff, 0xe3, 0xff, 0xde, 0x37, 0x71, 0x63, 0x72, 0xff, 0x89, 0x0d, 0xff, 0x85, 0x39, 0xff, 0xb5, 0xff, 0xb9, 0xff, 0x8a, 0xff, 0x9e, 0x60, 0xff, 0xad, 0x17, 0x01, 0x04, 0x03, 0x04, 0x00, 0x06, 0x00, 0x52, 0x08, 0x3e, 0x26}, line)
+	line, _, err = readBlock(r)
+	assert.Equal(t, []byte{0x28, 0xb5, 0x2f, 0xfd, 0x04, 0x00, 0xc9, 0x04, 0x00, 0x91, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x39, 0xff, 0x81, 0x03, 0x01, 0x01, 0x06, 0x52, 0x65, 0x63, 0x6f, 0x72, 0x64, 0x01, 0xff, 0x82, 0x00, 0x01, 0x03, 0x01, 0x04, 0x54, 0x79, 0x70, 0x65, 0x01, 0x06, 0x00, 0x01, 0x07, 0x4b, 0x65, 0x79, 0x48, 0x61, 0x73, 0x68, 0x01, 0xff, 0x84, 0x00, 0x01, 0x0a, 0x56, 0x61, 0x6c, 0x75, 0x65, 0x42, 0x79, 0x74, 0x65, 0x73, 0x01, 0x0a, 0x00, 0x00, 0x00, 0x19, 0xff, 0x83, 0x01, 0x01, 0x01, 0x09, 0x5b, 0x32, 0x38, 0x5d, 0x75, 0x69, 0x6e, 0x74, 0x38, 0x01, 0xff, 0x84, 0x00, 0x01, 0x06, 0x01, 0x38, 0x00, 0x00, 0x3c, 0xff, 0x82, 0x01, 0x01, 0x01, 0x1c, 0xff, 0xbf, 0x25, 0xff, 0xef, 0xff, 0xc8, 0xff, 0x85, 0x2c, 0xff, 0xbf, 0xff, 0xb5, 0xff, 0xad, 0xff, 0xfa, 0xff, 0xaf, 0x1c, 0xff, 0xe7, 0x71, 0xff, 0xfa, 0x36, 0xff, 0x95, 0x1b, 0xff, 0x91, 0xff, 0xab, 0x36, 0xff, 0xcd, 0x7a, 0x33, 0xff, 0xf7, 0xff, 0xec, 0xff, 0xee, 0xff, 0xc1, 0x01, 0x04, 0x03, 0x04, 0x00, 0x08, 0x00, 0xa5, 0x0e, 0x62, 0x53}, line)
+
+	line, _, err = readBlock(r)
+	assert.Equal(t, line, []byte{})
+	assert.Equal(t, io.EOF, err)
+}
+
+func TestRebuildIndex(t *testing.T) {
+	const filePath = "TestRebuiltIndex.zkv"
+	const recordCount = 4
+	defer os.Remove(filePath)
+	defer os.Remove(filePath + indexFileExt)
+
+	for i := 1; i <= recordCount; i++ {
+		db, err := Open(filePath)
+		assert.NoError(t, err)
+
+		err = db.Set(i, i)
+		assert.NoError(t, err)
+
+		err = db.Close()
+		assert.NoError(t, err)
+	}
+
+	db, err := Open(filePath)
+	assert.NoError(t, err)
+
+	err = db.RebuildIndex()
+	assert.NoError(t, err)
+
+	for i := 1; i <= recordCount; i++ {
+		var gotValue int
+
+		err = db.Get(i, &gotValue)
+		assert.NoError(t, err)
+		assert.Equal(t, i, gotValue)
+	}
+}