merge intervals

weed/filer2/filechunks.go

@@ -1,10 +1,13 @@
 package filer2
 
-import "github.com/chrislusf/seaweedfs/weed/pb/filer_pb"
+import (
+	"github.com/chrislusf/seaweedfs/weed/pb/filer_pb"
+	"sort"
+	"log"
+	"math"
+)
 
-type Chunks []*filer_pb.FileChunk
-
-func (chunks Chunks) TotalSize() (size uint64) {
+func TotalSize(chunks []*filer_pb.FileChunk) (size uint64) {
 	for _, c := range chunks {
 		t := uint64(c.Offset + int64(c.Size))
 		if size < t {
@@ -14,12 +17,345 @@ func (chunks Chunks) TotalSize() (size uint64) {
 	return
 }
 
-func (chunks Chunks) Len() int {
-	return len(chunks)
+func CompactFileChunks(chunks []*filer_pb.FileChunk) (compacted, garbage []*filer_pb.FileChunk) {
+	return
 }
-func (chunks Chunks) Swap(i, j int) {
-	chunks[i], chunks[j] = chunks[j], chunks[i]
+
+func mergeToVisibleIntervals(visibles []*visibleInterval, chunk *filer_pb.FileChunk) (merged []*visibleInterval) {
+	if len(visibles) == 0 {
+		return []*visibleInterval{newVisibleInterval(chunk.Offset, chunk.Offset+int64(chunk.Size), chunk.FileId, chunk.Mtime)}
+	}
+
+	log.Printf("merge chunk %+v => %d", chunk, len(visibles))
+	for _, v := range visibles {
+		log.Printf("=> %+v", v)
+	}
+
+	var nonOverlappingStop int
+
+	// find merge candidates
+	var mergeCandidates []int
+	for t := len(visibles) - 1; t >= 0; t-- {
+		if visibles[t].stop > chunk.Offset {
+			mergeCandidates = append(mergeCandidates, t)
+		} else {
+			nonOverlappingStop = t
+			break
+		}
+	}
+	log.Printf("merged candidates: %+v, starting from %d", mergeCandidates, nonOverlappingStop)
+
+	if len(mergeCandidates) == 0 {
+		merged = append(visibles, newVisibleInterval(
+			chunk.Offset,
+			chunk.Offset+int64(chunk.Size),
+			chunk.FileId,
+			chunk.Mtime,
+		))
+		return
+	}
+
+	// reverse merge candidates
+	i, j := 0, len(mergeCandidates)-1
+	for i < j {
+		mergeCandidates[i], mergeCandidates[j] = mergeCandidates[j], mergeCandidates[i]
+		i++
+		j--
+	}
+	log.Printf("reversed merged candidates: %+v", mergeCandidates)
+
+	// add chunk into a possibly connected intervals
+	var overlappingIntevals []*visibleInterval
+	for i = 0; i < len(mergeCandidates); i++ {
+		interval := visibles[mergeCandidates[i]]
+		if interval.modifiedTime >= chunk.Mtime {
+			log.Printf("overlappingIntevals add existing interval: [%d,%d)", interval.start, interval.stop)
+			overlappingIntevals = append(overlappingIntevals, interval)
+		} else {
+			start := max(interval.start, chunk.Offset)
+			stop := min(interval.stop, chunk.Offset+int64(chunk.Size))
+			if interval.start <= chunk.Offset {
+				if interval.start < start {
+					log.Printf("overlappingIntevals add 1: [%d,%d)", interval.start, start)
+					overlappingIntevals = append(overlappingIntevals, newVisibleInterval(
+						interval.start,
+						start,
+						interval.fileId,
+						interval.modifiedTime,
+					))
+				}
+				log.Printf("overlappingIntevals add 2: [%d,%d)", start, stop)
+				overlappingIntevals = append(overlappingIntevals, newVisibleInterval(
+					start,
+					stop,
+					chunk.FileId,
+					chunk.Mtime,
+				))
+				if interval.stop < stop {
+					log.Printf("overlappingIntevals add 3: [%d,%d)", interval.stop, stop)
+					overlappingIntevals = append(overlappingIntevals, newVisibleInterval(
+						interval.stop,
+						stop,
+						interval.fileId,
+						interval.modifiedTime,
+					))
+				}
+			}
+		}
+	}
+	logPrintf("overlappingIntevals", overlappingIntevals)
+
+	// merge connected intervals
+	merged = visibles[:nonOverlappingStop]
+	var lastInterval *visibleInterval
+	var prevIntervalIndex int
+	for i, interval := range overlappingIntevals {
+		if i == 0 {
+			prevIntervalIndex = i
+			continue
+		}
+		if overlappingIntevals[prevIntervalIndex].fileId != interval.fileId {
+			merged = append(merged, newVisibleInterval(
+				overlappingIntevals[prevIntervalIndex].start,
+				interval.start,
+				overlappingIntevals[prevIntervalIndex].fileId,
+				overlappingIntevals[prevIntervalIndex].modifiedTime,
+			))
+			prevIntervalIndex = i
+		}
+	}
+
+	if lastInterval != nil {
+		merged = append(merged, newVisibleInterval(
+			overlappingIntevals[prevIntervalIndex].start,
+			lastInterval.start,
+			overlappingIntevals[prevIntervalIndex].fileId,
+			overlappingIntevals[prevIntervalIndex].modifiedTime,
+		))
+	}
+
+	logPrintf("merged", merged)
+
+	return
 }
-func (chunks Chunks) Less(i, j int) bool {
-	return chunks[i].Offset < chunks[j].Offset
+
+func logPrintf(name string, visibles []*visibleInterval) {
+	log.Printf("%s len %d", name, len(visibles))
+	for _, v := range visibles {
+		log.Printf("%s:  => %+v", name, v)
+	}
+}
+
+func nonOverlappingVisibleIntervals(chunks []*filer_pb.FileChunk) (visibles []*visibleInterval) {
+
+	sort.Slice(chunks, func(i, j int) bool {
+		if chunks[i].Offset < chunks[j].Offset {
+			return true
+		}
+		if chunks[i].Offset == chunks[j].Offset {
+			return chunks[i].Mtime < chunks[j].Mtime
+		}
+		return false
+	})
+
+	if len(chunks) == 0 {
+		return
+	}
+
+	var parallelIntervals, intervals []*visibleInterval
+	var minStopInterval, upToDateInterval *visibleInterval
+	watermarkStart := chunks[0].Offset
+	for _, chunk := range chunks {
+		log.Printf("checking chunk: [%d,%d)", chunk.Offset, chunk.Offset+int64(chunk.Size))
+		logPrintf("parallelIntervals", parallelIntervals)
+		for len(parallelIntervals) > 0 && watermarkStart < chunk.Offset {
+			logPrintf("parallelIntervals loop 1", parallelIntervals)
+			logPrintf("parallelIntervals loop 1 intervals", intervals)
+			minStopInterval, upToDateInterval = findMinStopInterval(parallelIntervals)
+			nextStop := min(minStopInterval.stop, chunk.Offset)
+			intervals = append(intervals, newVisibleInterval(
+				max(watermarkStart, minStopInterval.start),
+				nextStop,
+				upToDateInterval.fileId,
+				upToDateInterval.modifiedTime,
+			))
+			watermarkStart = nextStop
+			logPrintf("parallelIntervals loop intervals =>", intervals)
+
+			// remove processed intervals, possibly multiple
+			var remaining []*visibleInterval
+			for _, interval := range parallelIntervals {
+				if interval.stop != watermarkStart {
+					remaining = append(remaining, newVisibleInterval(
+						interval.start,
+						interval.stop,
+						interval.fileId,
+						interval.modifiedTime,
+					))
+				}
+			}
+			parallelIntervals = remaining
+			logPrintf("parallelIntervals loop 2", parallelIntervals)
+			logPrintf("parallelIntervals loop 2 intervals", intervals)
+		}
+		parallelIntervals = append(parallelIntervals, newVisibleInterval(
+			chunk.Offset,
+			chunk.Offset+int64(chunk.Size),
+			chunk.FileId,
+			chunk.Mtime,
+		))
+	}
+
+	logPrintf("parallelIntervals loop 3", parallelIntervals)
+	logPrintf("parallelIntervals loop 3 intervals", intervals)
+	for len(parallelIntervals) > 0 {
+		minStopInterval, upToDateInterval = findMinStopInterval(parallelIntervals)
+		intervals = append(intervals, newVisibleInterval(
+			max(watermarkStart, minStopInterval.start),
+			minStopInterval.stop,
+			upToDateInterval.fileId,
+			upToDateInterval.modifiedTime,
+		))
+		watermarkStart = minStopInterval.stop
+
+		// remove processed intervals, possibly multiple
+		var remaining []*visibleInterval
+		for _, interval := range parallelIntervals {
+			if interval.stop != watermarkStart {
+				remaining = append(remaining, newVisibleInterval(
+					interval.start,
+					interval.stop,
+					interval.fileId,
+					interval.modifiedTime,
+				))
+			}
+		}
+		parallelIntervals = remaining
+	}
+	logPrintf("parallelIntervals loop 4", parallelIntervals)
+	logPrintf("intervals", intervals)
+
+	// merge connected intervals, now the intervals are non-intersecting
+	var lastInterval *visibleInterval
+	var prevIntervalIndex int
+	for i, interval := range intervals {
+		if i == 0 {
+			prevIntervalIndex = i
+			continue
+		}
+		if intervals[i-1].fileId != interval.fileId ||
+			intervals[i-1].stop < intervals[i].start {
+			visibles = append(visibles, newVisibleInterval(
+				intervals[prevIntervalIndex].start,
+				intervals[i-1].stop,
+				intervals[prevIntervalIndex].fileId,
+				intervals[prevIntervalIndex].modifiedTime,
+			))
+			prevIntervalIndex = i
+		}
+		lastInterval = intervals[i]
+		logPrintf("intervals loop 1 visibles", visibles)
+	}
+
+	if lastInterval != nil {
+		visibles = append(visibles, newVisibleInterval(
+			intervals[prevIntervalIndex].start,
+			lastInterval.stop,
+			intervals[prevIntervalIndex].fileId,
+			intervals[prevIntervalIndex].modifiedTime,
+		))
+	}
+
+	logPrintf("visibles", visibles)
+
+	return
+}
+
+func findMinStopInterval(intervals []*visibleInterval) (minStopInterval, upToDateInterval *visibleInterval) {
+	var latestMtime int64
+	latestIntervalIndex := 0
+	minStop := int64(math.MaxInt64)
+	minIntervalIndex := 0
+	for i, interval := range intervals {
+		if minStop > interval.stop {
+			minIntervalIndex = i
+			minStop = interval.stop
+		}
+		if latestMtime < interval.modifiedTime {
+			latestMtime = interval.modifiedTime
+			latestIntervalIndex = i
+		}
+	}
+	minStopInterval = intervals[minIntervalIndex]
+	upToDateInterval = intervals[latestIntervalIndex]
+	return
+}
+
+func nonOverlappingVisibleIntervals0(chunks []*filer_pb.FileChunk) (visibles []*visibleInterval) {
+
+	sort.Slice(chunks, func(i, j int) bool {
+		if chunks[i].Offset < chunks[j].Offset {
+			return true
+		}
+		if chunks[i].Offset == chunks[j].Offset {
+			return chunks[i].Mtime < chunks[j].Mtime
+		}
+		return false
+	})
+
+	for _, c := range chunks {
+		visibles = mergeToVisibleIntervals(visibles, c)
+	}
+
+	return
+}
+
+// find non-overlapping visible intervals
+// visible interval map to one file chunk
+
+type visibleInterval struct {
+	start        int64
+	stop         int64
+	modifiedTime int64
+	fileId       string
+}
+
+func newVisibleInterval(start, stop int64, fileId string, modifiedTime int64) *visibleInterval {
+	return &visibleInterval{start: start, stop: stop, fileId: fileId, modifiedTime: modifiedTime}
+}
+
+type stackOfChunkIds struct {
+	ids []int
+}
+
+func (s *stackOfChunkIds) isEmpty() bool {
+	return len(s.ids) == 0
+}
+
+func (s *stackOfChunkIds) pop() int {
+	t := s.ids[len(s.ids)-1]
+	s.ids = s.ids[:len(s.ids)-1]
+	return t
+}
+
+func (s *stackOfChunkIds) push(x int) {
+	s.ids = append(s.ids, x)
+}
+
+func (s *stackOfChunkIds) peek() int {
+	return s.ids[len(s.ids)-1]
+}
+
+func min(x, y int64) int64 {
+	if x <= y {
+		return x
+	}
+	return y
+}
+
+func max(x, y int64) int64 {
+	if x > y {
+		return x
+	}
+	return y
 }

weed/filer2/filechunks_test.go

@@ -0,0 +1,112 @@
+package filer2
+
+import (
+	"testing"
+	"github.com/chrislusf/seaweedfs/weed/pb/filer_pb"
+	"log"
+)
+
+func TestIntervalMerging(t *testing.T) {
+
+	testcases := []struct {
+		Chunks   []*filer_pb.FileChunk
+		Expected []*visibleInterval
+	}{
+		// case 0: normal
+		{
+			Chunks: []*filer_pb.FileChunk{
+				{Offset: 0, Size: 100, FileId: "abc", Mtime: 123},
+				{Offset: 100, Size: 100, FileId: "asdf", Mtime: 134},
+				{Offset: 200, Size: 100, FileId: "fsad", Mtime: 353},
+			},
+			Expected: []*visibleInterval{
+				{start: 0, stop: 100, fileId: "abc"},
+				{start: 100, stop: 200, fileId: "asdf"},
+				{start: 200, stop: 300, fileId: "fsad"},
+			},
+		},
+		// case 1: updates overwrite full chunks
+		{
+			Chunks: []*filer_pb.FileChunk{
+				{Offset: 0, Size: 100, FileId: "abc", Mtime: 123},
+				{Offset: 0, Size: 200, FileId: "asdf", Mtime: 134},
+			},
+			Expected: []*visibleInterval{
+				{start: 0, stop: 200, fileId: "asdf"},
+			},
+		},
+		// case 2: updates overwrite part of previous chunks
+		{
+			Chunks: []*filer_pb.FileChunk{
+				{Offset: 0, Size: 100, FileId: "abc", Mtime: 123},
+				{Offset: 0, Size: 50, FileId: "asdf", Mtime: 134},
+			},
+			Expected: []*visibleInterval{
+				{start: 0, stop: 50, fileId: "asdf"},
+				{start: 50, stop: 100, fileId: "abc"},
+			},
+		},
+		// case 3: updates overwrite full chunks
+		{
+			Chunks: []*filer_pb.FileChunk{
+				{Offset: 0, Size: 100, FileId: "abc", Mtime: 123},
+				{Offset: 0, Size: 200, FileId: "asdf", Mtime: 134},
+				{Offset: 50, Size: 250, FileId: "xxxx", Mtime: 154},
+			},
+			Expected: []*visibleInterval{
+				{start: 0, stop: 50, fileId: "asdf"},
+				{start: 50, stop: 300, fileId: "xxxx"},
+			},
+		},
+		// case 4: updates far away from prev chunks
+		{
+			Chunks: []*filer_pb.FileChunk{
+				{Offset: 0, Size: 100, FileId: "abc", Mtime: 123},
+				{Offset: 0, Size: 200, FileId: "asdf", Mtime: 134},
+				{Offset: 250, Size: 250, FileId: "xxxx", Mtime: 154},
+			},
+			Expected: []*visibleInterval{
+				{start: 0, stop: 200, fileId: "asdf"},
+				{start: 250, stop: 500, fileId: "xxxx"},
+			},
+		},
+		// case 5: updates overwrite full chunks
+		{
+			Chunks: []*filer_pb.FileChunk{
+				{Offset: 0, Size: 100, FileId: "abc", Mtime: 123},
+				{Offset: 0, Size: 200, FileId: "asdf", Mtime: 184},
+				{Offset: 70, Size: 150, FileId: "abc", Mtime: 143},
+				{Offset: 80, Size: 100, FileId: "xxxx", Mtime: 134},
+			},
+			Expected: []*visibleInterval{
+				{start: 0, stop: 200, fileId: "asdf"},
+				{start: 200, stop: 220, fileId: "abc"},
+			},
+		},
+	}
+
+	for i, testcase := range testcases {
+		log.Printf("++++++++++ merged test case %d ++++++++++++++++++++", i)
+		intervals := nonOverlappingVisibleIntervals(testcase.Chunks)
+		for x, interval := range intervals {
+			log.Printf("test case %d, interval %d, start=%d, stop=%d, fileId=%s",
+				i, x, interval.start, interval.stop, interval.fileId)
+			if interval.start != testcase.Expected[x].start {
+				t.Fatalf("failed on test case %d, interval %d, start %d, expect %d",
+					i, x, interval.start, testcase.Expected[x].start)
+			}
+			if interval.stop != testcase.Expected[x].stop {
+				t.Fatalf("failed on test case %d, interval %d, stop %d, expect %d",
+					i, x, interval.stop, testcase.Expected[x].stop)
+			}
+			if interval.fileId != testcase.Expected[x].fileId {
+				t.Fatalf("failed on test case %d, interval %d, chunkId %s, expect %s",
+					i, x, interval.fileId, testcase.Expected[x].fileId)
+			}
+		}
+		if len(intervals) != len(testcase.Expected) {
+			t.Fatalf("failed to compact test case %d, len %d expected %d", i, len(intervals), len(testcase.Expected))
+		}
+	}
+
+}

weed/server/filer_grpc_server.go

@@ -42,13 +42,12 @@ func (fs *FilerServer) ListEntries(ctx context.Context, req *filer_pb.ListEntrie
 	resp := &filer_pb.ListEntriesResponse{}
 	for _, entry := range entries {
 
-		glog.V(0).Infof("%s attr=%v size=%d", entry.Name(), entry.Attr, filer2.Chunks(entry.Chunks).TotalSize())
 		resp.Entries = append(resp.Entries, &filer_pb.Entry{
 			Name:        entry.Name(),
 			IsDirectory: entry.IsDirectory(),
 			Chunks:      entry.Chunks,
 			Attributes: &filer_pb.FuseAttributes{
-				FileSize: filer2.Chunks(entry.Chunks).TotalSize(),
+				FileSize: filer2.TotalSize(entry.Chunks),
 				Mtime:    entry.Mtime.Unix(),
 				Gid:      entry.Gid,
 				Uid:      entry.Uid,
@@ -71,7 +70,7 @@ func (fs *FilerServer) GetFileAttributes(ctx context.Context, req *filer_pb.GetF
 	if !found {
 		attributes.FileSize = 0
 	} else {
-		attributes.FileSize = filer2.Chunks(entry.Chunks).TotalSize()
+		attributes.FileSize = filer2.TotalSize(entry.Chunks)
 		attributes.FileMode = uint32(entry.Mode)
 		attributes.Uid = entry.Uid
 		attributes.Gid = entry.Gid