123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268 |
- package filer2
- import (
- "fmt"
- "hash/fnv"
- "math"
- "sort"
- "sync"
- "github.com/chrislusf/seaweedfs/weed/pb/filer_pb"
- )
- func TotalSize(chunks []*filer_pb.FileChunk) (size uint64) {
- for _, c := range chunks {
- t := uint64(c.Offset + int64(c.Size))
- if size < t {
- size = t
- }
- }
- return
- }
- func FileSize(entry *filer_pb.Entry) (size uint64) {
- return maxUint64(TotalSize(entry.Chunks), entry.Attributes.FileSize)
- }
- func ETag(entry *filer_pb.Entry) (etag string) {
- if entry.Attributes == nil || entry.Attributes.Md5 == nil {
- return ETagChunks(entry.Chunks)
- }
- return fmt.Sprintf("%x", entry.Attributes.Md5)
- }
- func ETagEntry(entry *Entry) (etag string) {
- if entry.Attr.Md5 == nil {
- return ETagChunks(entry.Chunks)
- }
- return fmt.Sprintf("%x", entry.Attr.Md5)
- }
- func ETagChunks(chunks []*filer_pb.FileChunk) (etag string) {
- if len(chunks) == 1 {
- return chunks[0].ETag
- }
- h := fnv.New32a()
- for _, c := range chunks {
- h.Write([]byte(c.ETag))
- }
- return fmt.Sprintf("%x", h.Sum32())
- }
- func CompactFileChunks(lookupFileIdFn LookupFileIdFunctionType, chunks []*filer_pb.FileChunk) (compacted, garbage []*filer_pb.FileChunk) {
- visibles, _ := NonOverlappingVisibleIntervals(lookupFileIdFn, chunks)
- fileIds := make(map[string]bool)
- for _, interval := range visibles {
- fileIds[interval.fileId] = true
- }
- for _, chunk := range chunks {
- if _, found := fileIds[chunk.GetFileIdString()]; found {
- compacted = append(compacted, chunk)
- } else {
- garbage = append(garbage, chunk)
- }
- }
- return
- }
- func MinusChunks(lookupFileIdFn LookupFileIdFunctionType, as, bs []*filer_pb.FileChunk) (delta []*filer_pb.FileChunk, err error) {
- aData, aMeta, aErr := ResolveChunkManifest(lookupFileIdFn, as)
- if aErr != nil {
- return nil, aErr
- }
- bData, bMeta, bErr := ResolveChunkManifest(lookupFileIdFn, bs)
- if bErr != nil {
- return nil, bErr
- }
- delta = append(delta, DoMinusChunks(aData, bData)...)
- delta = append(delta, DoMinusChunks(aMeta, bMeta)...)
- return
- }
- func DoMinusChunks(as, bs []*filer_pb.FileChunk) (delta []*filer_pb.FileChunk) {
- fileIds := make(map[string]bool)
- for _, interval := range bs {
- fileIds[interval.GetFileIdString()] = true
- }
- for _, chunk := range as {
- if _, found := fileIds[chunk.GetFileIdString()]; !found {
- delta = append(delta, chunk)
- }
- }
- return
- }
- type ChunkView struct {
- FileId string
- Offset int64
- Size uint64
- LogicOffset int64
- ChunkSize uint64
- CipherKey []byte
- IsGzipped bool
- }
- func (cv *ChunkView) IsFullChunk() bool {
- return cv.Size == cv.ChunkSize
- }
- func ViewFromChunks(lookupFileIdFn LookupFileIdFunctionType, chunks []*filer_pb.FileChunk, offset int64, size int64) (views []*ChunkView) {
- visibles, _ := NonOverlappingVisibleIntervals(lookupFileIdFn, chunks)
- return ViewFromVisibleIntervals(visibles, offset, size)
- }
- func ViewFromVisibleIntervals(visibles []VisibleInterval, offset int64, size int64) (views []*ChunkView) {
- stop := offset + size
- if size == math.MaxInt64 {
- stop = math.MaxInt64
- }
- if stop < offset {
- stop = math.MaxInt64
- }
- for _, chunk := range visibles {
- if chunk.start <= offset && offset < chunk.stop && offset < stop {
- views = append(views, &ChunkView{
- FileId: chunk.fileId,
- Offset: offset - chunk.start, // offset is the data starting location in this file id
- Size: uint64(min(chunk.stop, stop) - offset),
- LogicOffset: offset,
- ChunkSize: chunk.chunkSize,
- CipherKey: chunk.cipherKey,
- IsGzipped: chunk.isGzipped,
- })
- offset = min(chunk.stop, stop)
- }
- }
- return views
- }
- func logPrintf(name string, visibles []VisibleInterval) {
- /*
- log.Printf("%s len %d", name, len(visibles))
- for _, v := range visibles {
- log.Printf("%s: => %+v", name, v)
- }
- */
- }
- var bufPool = sync.Pool{
- New: func() interface{} {
- return new(VisibleInterval)
- },
- }
- func MergeIntoVisibles(visibles, newVisibles []VisibleInterval, chunk *filer_pb.FileChunk) []VisibleInterval {
- newV := newVisibleInterval(chunk.Offset, chunk.Offset+int64(chunk.Size), chunk.GetFileIdString(), chunk.Mtime, chunk.Size, chunk.CipherKey, chunk.IsCompressed)
- length := len(visibles)
- if length == 0 {
- return append(visibles, newV)
- }
- last := visibles[length-1]
- if last.stop <= chunk.Offset {
- return append(visibles, newV)
- }
- logPrintf(" before", visibles)
- for _, v := range visibles {
- if v.start < chunk.Offset && chunk.Offset < v.stop {
- newVisibles = append(newVisibles, newVisibleInterval(v.start, chunk.Offset, v.fileId, v.modifiedTime, chunk.Size, v.cipherKey, v.isGzipped))
- }
- chunkStop := chunk.Offset + int64(chunk.Size)
- if v.start < chunkStop && chunkStop < v.stop {
- newVisibles = append(newVisibles, newVisibleInterval(chunkStop, v.stop, v.fileId, v.modifiedTime, chunk.Size, v.cipherKey, v.isGzipped))
- }
- if chunkStop <= v.start || v.stop <= chunk.Offset {
- newVisibles = append(newVisibles, v)
- }
- }
- newVisibles = append(newVisibles, newV)
- logPrintf(" append", newVisibles)
- for i := len(newVisibles) - 1; i >= 0; i-- {
- if i > 0 && newV.start < newVisibles[i-1].start {
- newVisibles[i] = newVisibles[i-1]
- } else {
- newVisibles[i] = newV
- break
- }
- }
- logPrintf(" sorted", newVisibles)
- return newVisibles
- }
- // NonOverlappingVisibleIntervals translates the file chunk into VisibleInterval in memory
- // If the file chunk content is a chunk manifest
- func NonOverlappingVisibleIntervals(lookupFileIdFn LookupFileIdFunctionType, chunks []*filer_pb.FileChunk) (visibles []VisibleInterval, err error) {
- chunks, _, err = ResolveChunkManifest(lookupFileIdFn, chunks)
- sort.Slice(chunks, func(i, j int) bool {
- return chunks[i].Mtime < chunks[j].Mtime
- })
- var newVisibles []VisibleInterval
- for _, chunk := range chunks {
- newVisibles = MergeIntoVisibles(visibles, newVisibles, chunk)
- t := visibles[:0]
- visibles = newVisibles
- newVisibles = t
- logPrintf("add", visibles)
- }
- return
- }
- // find non-overlapping visible intervals
- // visible interval map to one file chunk
- type VisibleInterval struct {
- start int64
- stop int64
- modifiedTime int64
- fileId string
- chunkSize uint64
- cipherKey []byte
- isGzipped bool
- }
- func newVisibleInterval(start, stop int64, fileId string, modifiedTime int64, chunkSize uint64, cipherKey []byte, isGzipped bool) VisibleInterval {
- return VisibleInterval{
- start: start,
- stop: stop,
- fileId: fileId,
- modifiedTime: modifiedTime,
- chunkSize: chunkSize,
- cipherKey: cipherKey,
- isGzipped: isGzipped,
- }
- }
- func min(x, y int64) int64 {
- if x <= y {
- return x
- }
- return y
- }
|