refactor: WIP removing tsbd

cmd/influx/debug.go

@@ -8,7 +8,7 @@ import (
 
 	"github.com/influxdata/influxdb/v2"
 	"github.com/influxdata/influxdb/v2/internal/fs"
-	"github.com/influxdata/influxdb/v2/tsdb/tsm1"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/engine/tsm1"
 	"github.com/spf13/cobra"
 )
 

cmd/influx_inspect/buildtsi/buildtsi.go

@@ -12,10 +12,9 @@ import (
 	"github.com/influxdata/influxdb/v2/pkg/fs"
 	"github.com/influxdata/influxdb/v2/storage/wal"
 	"github.com/influxdata/influxdb/v2/toml"
-	"github.com/influxdata/influxdb/v2/tsdb"
-	"github.com/influxdata/influxdb/v2/tsdb/seriesfile"
-	"github.com/influxdata/influxdb/v2/tsdb/tsi1"
-	"github.com/influxdata/influxdb/v2/tsdb/tsm1"
+	"github.com/influxdata/influxdb/v2/v1/tsdb"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/index/tsi1"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/engine/tsm1"
 	"go.uber.org/zap"
 )
 

cmd/influx_inspect/verify/seriesfile/verify.go

@@ -9,7 +9,7 @@ import (
 	"sort"
 	"sync"
 
-	"github.com/influxdata/influxdb/v2/tsdb"
+	"github.com/influxdata/influxdb/v2/v1/tsdb"
 	"go.uber.org/zap"
 )
 
@@ -125,7 +125,7 @@ func (v Verify) VerifyPartition(partitionPath string) (valid bool, err error) {
 		return false, err
 	}
 
-	segments := make([]*SeriesSegment, 0, len(segmentInfos))
+	segments := make([]*tsdb.SeriesSegment, 0, len(segmentInfos))
 	ids := make(map[uint64]IDData)
 
 	// check every segment
@@ -137,7 +137,7 @@ func (v Verify) VerifyPartition(partitionPath string) (valid bool, err error) {
 		}
 
 		segmentPath := filepath.Join(partitionPath, segmentInfo.Name())
-		segmentID, err := ParseSeriesSegmentFilename(segmentInfo.Name())
+		segmentID, err := tsdb.ParseSeriesSegmentFilename(segmentInfo.Name())
 		if err != nil {
 			continue
 		}
@@ -150,7 +150,7 @@ func (v Verify) VerifyPartition(partitionPath string) (valid bool, err error) {
 
 		// open the segment for verifying the index. we want it to be open outside
 		// the for loop as well, so the defer is ok.
-		segment := NewSeriesSegment(segmentID, segmentPath)
+		segment := tsdb.NewSeriesSegment(segmentID, segmentPath)
 		if err := segment.Open(); err != nil {
 			return false, err
 		}
@@ -186,11 +186,11 @@ func (v Verify) VerifySegment(segmentPath string, ids map[uint64]IDData) (valid
 	v.Logger.Info("Verifying segment")
 
 	// Open up the segment and grab it's data.
-	segmentID, err := ParseSeriesSegmentFilename(segmentName)
+	segmentID, err := tsdb.ParseSeriesSegmentFilename(segmentName)
 	if err != nil {
 		return false, err
 	}
-	segment := NewSeriesSegment(segmentID, segmentPath)
+	segment := tsdb.NewSeriesSegment(segmentID, segmentPath)
 	if err := segment.Open(); err != nil {
 		v.Logger.Error("Error opening segment", zap.Error(err))
 		return false, nil
@@ -207,7 +207,7 @@ func (v Verify) VerifySegment(segmentPath string, ids map[uint64]IDData) (valid
 	}()
 
 	// Skip the header: it has already been verified by the Open call.
-	if err := buf.advance(SeriesSegmentHeaderSize); err != nil {
+	if err := buf.advance(tsdb.SeriesSegmentHeaderSize); err != nil {
 		v.Logger.Error("Unable to advance buffer",
 			zap.Int64("offset", buf.offset),
 			zap.Error(err))
@@ -224,39 +224,39 @@ entries:
 			return false, nil
 		}
 
-		flag, id, key, sz := ReadSeriesEntry(buf.data)
+		flag, id, key, sz := tsdb.ReadSeriesEntry(buf.data)
 
 		// Check the flag is valid and for id monotonicity.
 		hasKey := true
 		switch flag {
-		case SeriesEntryInsertFlag:
-			if !firstID && prevID > id.RawID() {
+		case tsdb.SeriesEntryInsertFlag:
+			if !firstID && prevID > id {
 				v.Logger.Error("ID is not monotonically increasing",
 					zap.Uint64("prev_id", prevID),
-					zap.Uint64("id", id.RawID()),
+					zap.Uint64("id", id),
 					zap.Int64("offset", buf.offset))
 				return false, nil
 			}
 
 			firstID = false
-			prevID = id.RawID()
+			prevID = id
 
 			if ids != nil {
 				keyCopy := make([]byte, len(key))
 				copy(keyCopy, key)
 
-				ids[id.RawID()] = IDData{
-					Offset: JoinSeriesOffset(segment.ID(), uint32(buf.offset)),
+				ids[id] = IDData{
+					Offset: tsdb.JoinSeriesOffset(segment.ID(), uint32(buf.offset)),
 					Key:    keyCopy,
 				}
 			}
 
-		case SeriesEntryTombstoneFlag:
+		case tsdb.SeriesEntryTombstoneFlag:
 			hasKey = false
 			if ids != nil {
-				data := ids[id.RawID()]
+				data := ids[id]
 				data.Deleted = true
-				ids[id.RawID()] = data
+				ids[id] = data
 			}
 
 		case 0: // if zero, there are no more entries
@@ -288,7 +288,7 @@ entries:
 							zap.String("recovered", fmt.Sprint(rec)))
 					}
 				}()
-				ParseSeriesKey(key)
+				tsdb.ParseSeriesKey(key)
 				parsed = true
 			}()
 			if !parsed {
@@ -311,7 +311,7 @@ entries:
 // VerifyIndex performs verification on an index in a series file. The error is only returned
 // if there was some fatal problem with operating, not if there was a problem with the partition.
 // The ids map must be built from verifying the passed in segments.
-func (v Verify) VerifyIndex(indexPath string, segments []*SeriesSegment,
+func (v Verify) VerifyIndex(indexPath string, segments []*tsdb.SeriesSegment,
 	ids map[uint64]IDData) (valid bool, err error) {
 	v.Logger.Info("Verifying index")
 
@@ -322,7 +322,7 @@ func (v Verify) VerifyIndex(indexPath string, segments []*SeriesSegment,
 		}
 	}()
 
-	index := NewSeriesIndex(indexPath)
+	index := tsdb.NewSeriesIndex(indexPath)
 	if err := index.Open(); err != nil {
 		v.Logger.Error("Error opening index", zap.Error(err))
 		return false, nil
@@ -353,7 +353,7 @@ func (v Verify) VerifyIndex(indexPath string, segments []*SeriesSegment,
 
 		IDData := ids[id]
 
-		if gotDeleted := index.IsDeleted(tsdb.NewSeriesID(id)); gotDeleted != IDData.Deleted {
+		if gotDeleted := index.IsDeleted(id); gotDeleted != IDData.Deleted {
 			v.Logger.Error("Index inconsistency",
 				zap.Uint64("id", id),
 				zap.Bool("got_deleted", gotDeleted),
@@ -367,7 +367,7 @@ func (v Verify) VerifyIndex(indexPath string, segments []*SeriesSegment,
 		}
 
 		// otherwise, check both that the offset is right and that we get the right id for the key
-		if gotOffset := index.FindOffsetByID(tsdb.NewSeriesID(id)); gotOffset != IDData.Offset {
+		if gotOffset := index.FindOffsetByID(id); gotOffset != IDData.Offset {
 			v.Logger.Error("Index inconsistency",
 				zap.Uint64("id", id),
 				zap.Int64("got_offset", gotOffset),
@@ -375,10 +375,10 @@ func (v Verify) VerifyIndex(indexPath string, segments []*SeriesSegment,
 			return false, nil
 		}
 
-		if gotID := index.FindIDBySeriesKey(segments, IDData.Key); gotID != tsdb.NewSeriesIDTyped(id) {
+		if gotID := index.FindIDBySeriesKey(segments, IDData.Key); gotID != id {
 			v.Logger.Error("Index inconsistency",
 				zap.Uint64("id", id),
-				zap.Uint64("got_id", gotID.RawID()),
+				zap.Uint64("got_id", gotID),
 				zap.Uint64("expected_id", id))
 			return false, nil
 		}

cmd/influx_inspect/verify/seriesfile/verify_test.go

@@ -1,7 +1,6 @@
 package seriesfile_test
 
 import (
-	"context"
 	"fmt"
 	"io"
 	"io/ioutil"
@@ -10,9 +9,9 @@ import (
 	"testing"
 	"time"
 
-	"github.com/influxdata/influxdb/v2/models"
-	"github.com/influxdata/influxdb/v2/tsdb"
-	"github.com/influxdata/influxdb/v2/tsdb/seriesfile"
+	"github.com/influxdata/influxdb/v2/cmd/influx_inspect/verify/seriesfile"
+	"github.com/influxdata/influxdb/v2/v1/models"
+	"github.com/influxdata/influxdb/v2/v1/tsdb"
 	"go.uber.org/zap"
 )
 
@@ -79,8 +78,8 @@ func NewTest(t *testing.T) *Test {
 
 	// create a series file in the directory
 	err = func() error {
-		seriesFile := seriesfile.NewSeriesFile(dir)
-		if err := seriesFile.Open(context.Background()); err != nil {
+		seriesFile := tsdb.NewSeriesFile(dir)
+		if err := seriesFile.Open(); err != nil {
 			return err
 		}
 		defer seriesFile.Close()
@@ -88,7 +87,7 @@ func NewTest(t *testing.T) *Test {
 
 		const (
 			compactionThreshold = 100
-			numSeries           = 2 * seriesfile.SeriesFilePartitionN * compactionThreshold
+			numSeries           = 2 * tsdb.SeriesFilePartitionN * compactionThreshold
 		)
 
 		for _, partition := range seriesFile.Partitions() {
@@ -103,17 +102,13 @@ func NewTest(t *testing.T) *Test {
 			tagsSlice = append(tagsSlice, nil)
 		}
 
-		keys := seriesfile.GenerateSeriesKeys(names, tagsSlice)
-		//keyPartitionIDs := seriesFile.SeriesKeysPartitionIDs(keys)
-		ids := make([]uint64, len(keys))
-
-		//ids, err := seriesFile.CreateSeriesListIfNotExists(names, tagsSlice)
+		ids, err := seriesFile.CreateSeriesListIfNotExists(names, tagsSlice)
 		if err != nil {
 			return err
 		}
 
 		// delete one series
-		if err := seriesFile.DeleteSeriesIDs([]tsdb.SeriesID{tsdb.NewSeriesID(ids[0])}); err != nil {
+		if err := seriesFile.DeleteSeriesID(ids[0]); err != nil {
 			return err
 		}
 

cmd/influx_inspect/verify/tombstone/verify.go

@@ -0,0 +1,142 @@
+// Package tombstone verifies integrity of tombstones.
+package tombstone
+
+import (
+	"errors"
+	"flag"
+	"fmt"
+	"io"
+	"os"
+	"path/filepath"
+	"time"
+
+	"github.com/influxdata/influxdb/v2/v1/tsdb/engine/tsm1"
+)
+
+// Command represents the program execution for "influx_inspect verify-tombstone".
+type Command struct {
+	Stderr io.Writer
+	Stdout io.Writer
+}
+
+// NewCommand returns a new instance of Command.
+func NewCommand() *Command {
+	return &Command{
+		Stderr: os.Stderr,
+		Stdout: os.Stdout,
+	}
+}
+
+// Run executes the command.
+func (cmd *Command) Run(args ...string) error {
+	runner := verifier{w: cmd.Stdout}
+	fs := flag.NewFlagSet("verify-tombstone", flag.ExitOnError)
+	fs.StringVar(&runner.path, "path", os.Getenv("HOME")+"/.influxdb", "path to find tombstone files")
+	v := fs.Bool("v", false, "verbose: emit periodic progress")
+	vv := fs.Bool("vv", false, "very verbose: emit every tombstone entry key and time range")
+	vvv := fs.Bool("vvv", false, "very very verbose: emit every tombstone entry key and RFC3339Nano time range")
+
+	fs.SetOutput(cmd.Stdout)
+
+	if err := fs.Parse(args); err != nil {
+		return err
+	}
+
+	if *v {
+		runner.verbosity = verbose
+	}
+	if *vv {
+		runner.verbosity = veryVerbose
+	}
+	if *vvv {
+		runner.verbosity = veryVeryVerbose
+	}
+
+	return runner.Run()
+}
+
+const (
+	quiet = iota
+	verbose
+	veryVerbose
+	veryVeryVerbose
+)
+
+type verifier struct {
+	path      string
+	verbosity int
+
+	w     io.Writer
+	files []string
+	f     string
+}
+
+func (v *verifier) loadFiles() error {
+	return filepath.Walk(v.path, func(path string, f os.FileInfo, err error) error {
+		if err != nil {
+			return err
+		}
+		if filepath.Ext(path) == "."+tsm1.TombstoneFileExtension {
+			v.files = append(v.files, path)
+		}
+		return nil
+	})
+}
+
+func (v *verifier) Next() bool {
+	if len(v.files) == 0 {
+		return false
+	}
+
+	v.f, v.files = v.files[0], v.files[1:]
+	return true
+}
+
+func (v *verifier) Run() error {
+	if err := v.loadFiles(); err != nil {
+		return err
+	}
+
+	var failed bool
+	start := time.Now()
+	for v.Next() {
+		if v.verbosity > quiet {
+			fmt.Fprintf(v.w, "Verifying: %q\n", v.f)
+		}
+
+		tombstoner := tsm1.NewTombstoner(v.f, nil)
+		if !tombstoner.HasTombstones() {
+			fmt.Fprintf(v.w, "%s has no tombstone entries", v.f)
+			continue
+		}
+
+		var totalEntries int64
+		err := tombstoner.Walk(func(t tsm1.Tombstone) error {
+			totalEntries++
+			if v.verbosity > quiet && totalEntries%(10*1e6) == 0 {
+				fmt.Fprintf(v.w, "Verified %d tombstone entries\n", totalEntries)
+			} else if v.verbosity > verbose {
+				var min interface{} = t.Min
+				var max interface{} = t.Max
+				if v.verbosity > veryVerbose {
+					min = time.Unix(0, t.Min)
+					max = time.Unix(0, t.Max)
+				}
+				fmt.Printf("key: %q, min: %v, max: %v\n", t.Key, min, max)
+			}
+			return nil
+		})
+		if err != nil {
+			fmt.Fprintf(v.w, "%q failed to walk tombstone entries: %v. Last okay entry: %d\n", v.f, err, totalEntries)
+			failed = true
+			continue
+		}
+
+		fmt.Fprintf(v.w, "Completed verification for %q in %v.\nVerified %d entries\n\n", v.f, time.Since(start), totalEntries)
+	}
+
+	if failed {
+		return errors.New("failed tombstone verification")
+	}
+	return nil
+}

cmd/influx_inspect/verify/tsm/verify.go

@@ -0,0 +1,232 @@
+// Package tsm verifies integrity of TSM files.
+package tsm
+
+import (
+	"flag"
+	"fmt"
+	"hash/crc32"
+	"io"
+	"os"
+	"path/filepath"
+	"text/tabwriter"
+	"time"
+	"unicode/utf8"
+
+	"github.com/influxdata/influxdb/v2/v1/tsdb/engine/tsm1"
+	"github.com/pkg/errors"
+)
+
+// Command represents the program execution for "influx_inspect verify".
+type Command struct {
+	Stderr io.Writer
+	Stdout io.Writer
+}
+
+// NewCommand returns a new instance of Command.
+func NewCommand() *Command {
+	return &Command{
+		Stderr: os.Stderr,
+		Stdout: os.Stdout,
+	}
+}
+
+// Run executes the command.
+func (cmd *Command) Run(args ...string) error {
+	var path string
+	fs := flag.NewFlagSet("verify", flag.ExitOnError)
+	fs.StringVar(&path, "dir", os.Getenv("HOME")+"/.influxdb", "Root storage path. [$HOME/.influxdb]")
+
+	var checkUTF8 bool
+	fs.BoolVar(&checkUTF8, "check-utf8", false, "Verify series keys are valid UTF-8")
+
+	fs.SetOutput(cmd.Stdout)
+	fs.Usage = cmd.printUsage
+
+	if err := fs.Parse(args); err != nil {
+		return err
+	}
+
+	dataPath := filepath.Join(path, "data")
+	tw := tabwriter.NewWriter(cmd.Stdout, 16, 8, 0, '\t', 0)
+
+	var runner verifier
+	if checkUTF8 {
+		runner = &verifyUTF8{}
+	} else {
+		runner = &verifyChecksums{}
+	}
+	err := runner.Run(tw, dataPath)
+	tw.Flush()
+	return err
+}
+
+// printUsage prints the usage message to STDERR.
+func (cmd *Command) printUsage() {
+	usage := fmt.Sprintf(`Verifies the integrity of TSM files.
+
+Usage: influx_inspect verify [flags]
+
+    -dir <path>
+            The root storage path.
+            Must be changed if you are using a non-default storage directory.
+            Defaults to "%[1]s/.influxdb".
+    -check-utf8 
+            Verify series keys are valid UTF-8.
+            This check skips verification of block checksums.
+ `, os.Getenv("HOME"))
+
+	fmt.Fprintf(cmd.Stdout, usage)
+}
+
+type verifyTSM struct {
+	files []string
+	f     string
+	start time.Time
+	err   error
+}
+
+func (v *verifyTSM) loadFiles(dataPath string) error {
+	err := filepath.Walk(dataPath, func(path string, f os.FileInfo, err error) error {
+		if err != nil {
+			return err
+		}
+		if filepath.Ext(path) == "."+tsm1.TSMFileExtension {
+			v.files = append(v.files, path)
+		}
+		return nil
+	})
+
+	if err != nil {
+		return errors.Wrap(err, "could not load storage files (use -dir for custom storage root)")
+	}
+
+	return nil
+}
+
+func (v *verifyTSM) Next() bool {
+	if len(v.files) == 0 {
+		return false
+	}
+
+	v.f, v.files = v.files[0], v.files[1:]
+	return true
+}
+
+func (v *verifyTSM) TSMReader() (string, *tsm1.TSMReader) {
+	file, err := os.OpenFile(v.f, os.O_RDONLY, 0600)
+	if err != nil {
+		v.err = err
+		return "", nil
+	}
+
+	reader, err := tsm1.NewTSMReader(file)
+	if err != nil {
+		file.Close()
+		v.err = err
+		return "", nil
+	}
+
+	return v.f, reader
+}
+
+func (v *verifyTSM) Start() {
+	v.start = time.Now()
+}
+
+func (v *verifyTSM) Elapsed() time.Duration {
+	return time.Since(v.start)
+}
+
+type verifyChecksums struct {
+	verifyTSM
+	totalErrors int
+	total       int
+}
+
+func (v *verifyChecksums) Run(w io.Writer, dataPath string) error {
+	if err := v.loadFiles(dataPath); err != nil {
+		return err
+	}
+
+	v.Start()
+
+	for v.Next() {
+		f, reader := v.TSMReader()
+		if reader == nil {
+			break
+		}
+
+		blockItr := reader.BlockIterator()
+		fileErrors := 0
+		count := 0
+		for blockItr.Next() {
+			v.total++
+			key, _, _, _, checksum, buf, err := blockItr.Read()
+			if err != nil {
+				v.totalErrors++
+				fileErrors++
+				fmt.Fprintf(w, "%s: could not get checksum for key %v block %d due to error: %q\n", f, key, count, err)
+			} else if expected := crc32.ChecksumIEEE(buf); checksum != expected {
+				v.totalErrors++
+				fileErrors++
+				fmt.Fprintf(w, "%s: got %d but expected %d for key %v, block %d\n", f, checksum, expected, key, count)
+			}
+			count++
+		}
+		if fileErrors == 0 {
+			fmt.Fprintf(w, "%s: healthy\n", f)
+		}
+		reader.Close()
+	}
+
+	fmt.Fprintf(w, "Broken Blocks: %d / %d, in %vs\n", v.totalErrors, v.total, v.Elapsed().Seconds())
+
+	return v.err
+}
+
+type verifyUTF8 struct {
+	verifyTSM
+	totalErrors int
+	total       int
+}
+
+func (v *verifyUTF8) Run(w io.Writer, dataPath string) error {
+	if err := v.loadFiles(dataPath); err != nil {
+		return err
+	}
+
+	v.Start()
+
+	for v.Next() {
+		f, reader := v.TSMReader()
+		if reader == nil {
+			break
+		}
+
+		n := reader.KeyCount()
+		fileErrors := 0
+		v.total += n
+		for i := 0; i < n; i++ {
+			key, _ := reader.KeyAt(i)
+			if !utf8.Valid(key) {
+				v.totalErrors++
+				fileErrors++
+				fmt.Fprintf(w, "%s: key #%d is not valid UTF-8\n", f, i)
+			}
+		}
+		if fileErrors == 0 {
+			fmt.Fprintf(w, "%s: healthy\n", f)
+		}
+	}
+
+	fmt.Fprintf(w, "Invalid Keys: %d / %d, in %vs\n", v.totalErrors, v.total, v.Elapsed().Seconds())
+	if v.totalErrors > 0 && v.err == nil {
+		v.err = errors.New("check-utf8: failed")
+	}
+
+	return v.err
+}
+
+type verifier interface {
+	Run(w io.Writer, dataPath string) error
+}

cmd/influx_inspect/verify/tsm/verify_test.go

@@ -0,0 +1,3 @@
+package tsm_test
+
+// TODO: write some tests

cmd/influxd/generate/generator.go

@@ -14,10 +14,9 @@ import (
 	"github.com/influxdata/influxdb/v2/pkg/data/gen"
 	"github.com/influxdata/influxdb/v2/pkg/limiter"
 	"github.com/influxdata/influxdb/v2/storage"
-	"github.com/influxdata/influxdb/v2/tsdb"
-	"github.com/influxdata/influxdb/v2/tsdb/seriesfile"
-	"github.com/influxdata/influxdb/v2/tsdb/tsi1"
-	"github.com/influxdata/influxdb/v2/tsdb/tsm1"
+	"github.com/influxdata/influxdb/v2/v1/tsdb"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/index/tsi1"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/engine/tsm1"
 )
 
 type Generator struct {

cmd/influxd/generate/internal/shard/writer.go

@@ -6,7 +6,7 @@ import (
 	"path/filepath"
 
 	"github.com/influxdata/influxdb/v2/pkg/data/gen"
-	"github.com/influxdata/influxdb/v2/tsdb/tsm1"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/engine/tsm1"
 )
 
 const (

cmd/influxd/inspect/build_tsi.go

@@ -13,9 +13,9 @@ import (
 	"github.com/influxdata/influxdb/v2/cmd/influx_inspect/buildtsi"
 	"github.com/influxdata/influxdb/v2/logger"
 	"github.com/influxdata/influxdb/v2/storage"
-	"github.com/influxdata/influxdb/v2/tsdb/seriesfile"
-	"github.com/influxdata/influxdb/v2/tsdb/tsi1"
-	"github.com/influxdata/influxdb/v2/tsdb/tsm1"
+	"github.com/influxdata/influxdb/v2/v1/tsdb"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/index/tsi1"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/engine/tsm1"
 	"github.com/spf13/cobra"
 )
 

cmd/influxd/inspect/compact_series_file.go

@@ -12,7 +12,6 @@ import (
 
 	"github.com/influxdata/influxdb/v2/internal/fs"
 	"github.com/influxdata/influxdb/v2/storage"
-	"github.com/influxdata/influxdb/v2/tsdb/seriesfile"
 	"github.com/spf13/cobra"
 	"golang.org/x/sync/errgroup"
 )

cmd/influxd/inspect/dump_tsi1.go

@@ -8,7 +8,7 @@ import (
 	"regexp"
 
 	"github.com/influxdata/influxdb/v2/internal/fs"
-	"github.com/influxdata/influxdb/v2/tsdb/tsi1"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/index/tsi1"
 	"github.com/spf13/cobra"
 	"go.uber.org/zap"
 )

cmd/influxd/inspect/export_blocks.go

@@ -3,7 +3,7 @@ package inspect
 import (
 	"os"
 
-	"github.com/influxdata/influxdb/v2/tsdb/tsm1"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/engine/tsm1"
 	"github.com/spf13/cobra"
 )
 

cmd/influxd/inspect/export_index.go

@@ -7,8 +7,8 @@ import (
 	"path/filepath"
 
 	"github.com/influxdata/influxdb/v2/internal/fs"
-	"github.com/influxdata/influxdb/v2/tsdb/seriesfile"
-	"github.com/influxdata/influxdb/v2/tsdb/tsi1"
+	"github.com/influxdata/influxdb/v2/v1/tsdb"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/index/tsi1"
 	"github.com/spf13/cobra"
 )
 
@@ -32,7 +32,7 @@ SQL format for easier inspection and debugging.`,
 
 	cmd.RunE = func(cmd *cobra.Command, args []string) error {
 		// Initialize series file.
-		sfile := seriesfile.NewSeriesFile(seriesFilePath)
+		sfile := tsdb.NewSeriesFile(seriesFilePath)
 		if err := sfile.Open(context.Background()); err != nil {
 			return err
 		}

cmd/influxd/inspect/report_tsi1.go

@@ -6,7 +6,7 @@ import (
 	"os"
 
 	"github.com/influxdata/influxdb/v2"
-	"github.com/influxdata/influxdb/v2/tsdb/tsi1"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/index/tsi1"
 	"github.com/spf13/cobra"
 )
 

cmd/influxd/inspect/report_tsm.go

@@ -5,7 +5,7 @@ import (
 	"github.com/influxdata/influxdb/v2"
 	"github.com/influxdata/influxdb/v2/internal/fs"
 	"github.com/influxdata/influxdb/v2/kit/errors"
-	"github.com/influxdata/influxdb/v2/tsdb/tsm1"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/engine/tsm1"
 	"github.com/spf13/cobra"
 	"os"
 	"path/filepath"

cmd/influxd/inspect/verify_series.go

@@ -4,8 +4,8 @@ import (
 	"os"
 	"runtime"
 
+	"github.com/influxdata/influxdb/v2/cmd/influx_inspect/verify/seriesfile"
 	"github.com/influxdata/influxdb/v2/logger"
-	"github.com/influxdata/influxdb/v2/tsdb/seriesfile"
 	"github.com/spf13/cobra"
 	"go.uber.org/zap/zapcore"
 )

cmd/influxd/inspect/verify_tsm.go

@@ -6,7 +6,7 @@ import (
 	"path/filepath"
 
 	"github.com/influxdata/influxdb/v2/kit/cli"
-	"github.com/influxdata/influxdb/v2/tsdb/tsm1"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/engine/tsm1"
 	"github.com/spf13/cobra"
 )
 

cmd/influxd/launcher/engine.go

@@ -13,7 +13,7 @@ import (
 	"github.com/influxdata/influxdb/v2/models"
 	"github.com/influxdata/influxdb/v2/storage"
 	"github.com/influxdata/influxdb/v2/storage/reads"
-	"github.com/influxdata/influxdb/v2/tsdb/cursors"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/cursors"
 	"github.com/influxdata/influxql"
 	"github.com/prometheus/client_golang/prometheus"
 	"go.uber.org/zap"

cmd/influxd/launcher/launcher.go

@@ -62,10 +62,10 @@ import (
 	"github.com/influxdata/influxdb/v2/task/backend/scheduler"
 	"github.com/influxdata/influxdb/v2/telemetry"
 	"github.com/influxdata/influxdb/v2/tenant"
-	_ "github.com/influxdata/influxdb/v2/tsdb/tsi1" // needed for tsi1
-	_ "github.com/influxdata/influxdb/v2/tsdb/tsm1" // needed for tsm1
 	storage2 "github.com/influxdata/influxdb/v2/v1/services/storage"
 	"github.com/influxdata/influxdb/v2/v1/storage/reads"
+	_ "github.com/influxdata/influxdb/v2/v1/tsdb/index/tsi1" // needed for tsi1
+	_ "github.com/influxdata/influxdb/v2/v1/tsdb/engine/tsm1" // needed for tsm1
 	"github.com/influxdata/influxdb/v2/vault"
 	pzap "github.com/influxdata/influxdb/v2/zap"
 	"github.com/opentracing/opentracing-go"

cmd/influxd/launcher/storage_test.go

@@ -13,7 +13,7 @@ import (
 	"github.com/influxdata/influxdb/v2/cmd/influxd/launcher"
 	"github.com/influxdata/influxdb/v2/http"
 	"github.com/influxdata/influxdb/v2/toml"
-	"github.com/influxdata/influxdb/v2/tsdb/tsm1"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/engine/tsm1"
 )
 
 func TestStorage_WriteAndQuery(t *testing.T) {

cmd/influxd/main.go

@@ -13,8 +13,8 @@ import (
 	"github.com/influxdata/influxdb/v2/cmd/influxd/launcher"
 	"github.com/influxdata/influxdb/v2/cmd/influxd/restore"
 	_ "github.com/influxdata/influxdb/v2/query/builtin"
-	_ "github.com/influxdata/influxdb/v2/tsdb/tsi1"
-	_ "github.com/influxdata/influxdb/v2/tsdb/tsm1"
+	_ "github.com/influxdata/influxdb/v2/v1/tsdb/index/tsi1"
+	_ "github.com/influxdata/influxdb/v2/v1/tsdb/engine/tsm1"
 	"github.com/spf13/cobra"
 )
 

gather/recorder.go

@@ -6,7 +6,7 @@ import (
 
 	"github.com/influxdata/influxdb/v2/nats"
 	"github.com/influxdata/influxdb/v2/storage"
-	"github.com/influxdata/influxdb/v2/tsdb"
+	"github.com/influxdata/influxdb/v2/v1/tsdb"
 	"go.uber.org/zap"
 )
 

kit/cli/idflag.go

@@ -2,7 +2,7 @@ package cli
 
 import (
 	"github.com/influxdata/influxdb/v2"
-	"github.com/influxdata/influxdb/v2/tsdb"
+	"github.com/influxdata/influxdb/v2/v1/tsdb"
 	"github.com/spf13/cobra"
 	"github.com/spf13/pflag"
 )

mock/reads_resultset.go

@@ -4,7 +4,7 @@ import (
 	"github.com/influxdata/influxdb/v2/models"
 	"github.com/influxdata/influxdb/v2/pkg/data/gen"
 	"github.com/influxdata/influxdb/v2/storage/reads"
-	"github.com/influxdata/influxdb/v2/tsdb/cursors"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/cursors"
 )
 
 type GeneratorResultSet struct {

mock/reads_resultset_test.go

@@ -9,7 +9,7 @@ import (
 	"github.com/influxdata/influxdb/v2/mock"
 	"github.com/influxdata/influxdb/v2/pkg/data/gen"
 	"github.com/influxdata/influxdb/v2/storage/reads"
-	"github.com/influxdata/influxdb/v2/tsdb/cursors"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/cursors"
 )
 
 func mustNewSpecFromToml(tb testing.TB, toml string) *gen.Spec {

models/points_test.go

@@ -19,7 +19,7 @@ import (
 	"github.com/google/go-cmp/cmp"
 	"github.com/influxdata/influxdb/v2"
 	"github.com/influxdata/influxdb/v2/models"
-	"github.com/influxdata/influxdb/v2/tsdb"
+	"github.com/influxdata/influxdb/v2/v1/tsdb"
 	"github.com/stretchr/testify/assert"
 )
 

pkg/data/gen/arrays.gen.go

@@ -7,8 +7,8 @@
 package gen
 
 import (
-	"github.com/influxdata/influxdb/v2/tsdb/cursors"
-	"github.com/influxdata/influxdb/v2/tsdb/tsm1"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/cursors"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/engine/tsm1"
 )
 
 type FloatValues interface {

pkg/data/gen/arrays.gen.go.tmpl

@@ -1,8 +1,8 @@
 package gen
 
 import (
-	"github.com/influxdata/influxdb/v2/tsdb/cursors"
-	"github.com/influxdata/influxdb/v2/tsdb/tsm1"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/cursors"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/engine/tsm1"
 )
 
 {{range .}}

pkg/data/gen/merged_series_generator_test.go

@@ -10,7 +10,7 @@ import (
 	"github.com/google/go-cmp/cmp"
 	"github.com/influxdata/influxdb/v2"
 	"github.com/influxdata/influxdb/v2/models"
-	"github.com/influxdata/influxdb/v2/tsdb"
+	"github.com/influxdata/influxdb/v2/v1/tsdb"
 )
 
 var (

pkg/data/gen/specs.go

@@ -13,7 +13,7 @@ import (
 	"github.com/BurntSushi/toml"
 	"github.com/influxdata/influxdb/v2"
 	"github.com/influxdata/influxdb/v2/models"
-	"github.com/influxdata/influxdb/v2/tsdb"
+	"github.com/influxdata/influxdb/v2/v1/tsdb"
 	"github.com/pkg/errors"
 )
 

pkg/data/gen/values_sequence.gen.go

@@ -8,7 +8,7 @@ package gen
 
 import (
 	"github.com/influxdata/influxdb/v2/models"
-	"github.com/influxdata/influxdb/v2/tsdb/cursors"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/cursors"
 )
 
 type FloatValuesSequence interface {

pkg/data/gen/values_sequence.gen.go.tmpl

@@ -2,7 +2,7 @@ package gen
 
 import (
 	"github.com/influxdata/influxdb/v2/models"
-	"github.com/influxdata/influxdb/v2/tsdb/cursors"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/cursors"
 )
 
 {{range .}}

predicate/predicate.go

@@ -3,7 +3,7 @@ package predicate
 import (
 	"github.com/influxdata/influxdb/v2"
 	"github.com/influxdata/influxdb/v2/storage/reads/datatypes"
-	"github.com/influxdata/influxdb/v2/tsdb/tsm1"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/engine/tsm1"
 )
 
 // Node is a predicate node.

query/storage.go

@@ -10,7 +10,7 @@ import (
 	"github.com/influxdata/flux/plan"
 	"github.com/influxdata/influxdb/v2"
 	"github.com/influxdata/influxdb/v2/storage/reads/datatypes"
-	"github.com/influxdata/influxdb/v2/tsdb/cursors"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/cursors"
 )
 
 // StorageReader is an interface for reading tables from the storage subsystem.

storage/compat/compat.go

@@ -10,7 +10,7 @@ package compat
 import (
 	"github.com/influxdata/influxdb/v2/storage"
 	"github.com/influxdata/influxdb/v2/toml"
-	"github.com/influxdata/influxdb/v2/tsdb/tsm1"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/engine/tsm1"
 )
 
 // Config matches the old toml layout from the influxdb repo, so that we can read

storage/config.go

@@ -5,9 +5,8 @@ import (
 	"time"
 
 	"github.com/influxdata/influxdb/v2/toml"
-	"github.com/influxdata/influxdb/v2/tsdb/seriesfile"
-	"github.com/influxdata/influxdb/v2/tsdb/tsi1"
-	"github.com/influxdata/influxdb/v2/tsdb/tsm1"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/index/tsi1"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/engine/tsm1"
 )
 
 // Default configuration values.

storage/engine.go

@@ -14,11 +14,11 @@ import (
 	"github.com/influxdata/influxdb/v2"
 	"github.com/influxdata/influxdb/v2/kit/tracing"
 	"github.com/influxdata/influxdb/v2/logger"
-	"github.com/influxdata/influxdb/v2/tsdb/cursors"
 	"github.com/influxdata/influxdb/v2/v1/coordinator"
 	"github.com/influxdata/influxdb/v2/v1/models"
 	"github.com/influxdata/influxdb/v2/v1/services/meta"
 	"github.com/influxdata/influxdb/v2/v1/tsdb"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/cursors"
 	"github.com/influxdata/influxql"
 	"github.com/pkg/errors"
 	"github.com/prometheus/client_golang/prometheus"

storage/engine_measurement_notime_schema.go

@@ -4,7 +4,7 @@ import (
 	"context"
 
 	"github.com/influxdata/influxdb/v2"
-	"github.com/influxdata/influxdb/v2/tsdb/cursors"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/cursors"
 	"github.com/influxdata/influxql"
 )
 

storage/engine_measurement_schema.go

@@ -4,7 +4,7 @@ import (
 	"context"
 
 	"github.com/influxdata/influxdb/v2"
-	"github.com/influxdata/influxdb/v2/tsdb/cursors"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/cursors"
 	"github.com/influxdata/influxql"
 )
 

storage/engine_schema.go

@@ -4,7 +4,7 @@ import (
 	"context"
 
 	"github.com/influxdata/influxdb/v2"
-	"github.com/influxdata/influxdb/v2/tsdb/cursors"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/cursors"
 	"github.com/influxdata/influxql"
 )
 

storage/engine_test.go

@@ -15,8 +15,8 @@ import (
 	"github.com/influxdata/influxdb/v2/models"
 	"github.com/influxdata/influxdb/v2/storage"
 	"github.com/influxdata/influxdb/v2/storage/reads/datatypes"
-	"github.com/influxdata/influxdb/v2/tsdb"
-	"github.com/influxdata/influxdb/v2/tsdb/tsm1"
+	"github.com/influxdata/influxdb/v2/v1/tsdb"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/engine/tsm1"
 	"github.com/prometheus/client_golang/prometheus"
 )
 

storage/flux/reader.go

@@ -16,7 +16,7 @@ import (
 	"github.com/influxdata/influxdb/v2/query"
 	storage "github.com/influxdata/influxdb/v2/storage/reads"
 	"github.com/influxdata/influxdb/v2/storage/reads/datatypes"
-	"github.com/influxdata/influxdb/v2/tsdb/cursors"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/cursors"
 )
 
 // GroupCursorError is returned when two different cursor types

storage/flux/table.gen.go

@@ -18,8 +18,7 @@ import (
 	"github.com/influxdata/influxdb/v2"
 	"github.com/influxdata/influxdb/v2/models"
 	storage "github.com/influxdata/influxdb/v2/storage/reads"
-	"github.com/influxdata/influxdb/v2/storage/reads/datatypes"
-	"github.com/influxdata/influxdb/v2/tsdb/cursors"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/cursors"
 )
 
 //
@@ -308,7 +307,7 @@ func (t *floatWindowTable) advance() bool {
 	} else {
 		cr.cols[startColIdx] = start
 		cr.cols[stopColIdx] = stop
-		cr.cols[valueColIdxWithoutTime] = values
+		cr.cols[windowedValueColIdx] = values
 	}
 	t.appendTags(cr)
 	return true
@@ -711,93 +710,26 @@ func (t *floatGroupTable) Do(f func(flux.ColReader) error) error {
 }
 
 func (t *floatGroupTable) advance() bool {
-	if t.cur == nil {
-		// For group aggregates, we will try to get all the series and all table buffers within those series
-		// all at once and merge them into one row when this advance() function is first called.
-		// At the end of this process, t.advanceCursor() already returns false and t.cur becomes nil.
-		// But we still need to return true to indicate that there is data to be returned.
-		// The second time when we call this advance(), t.cur is already nil, so we directly return false.
+RETRY:
+	a := t.cur.Next()
+	l := a.Len()
+	if l == 0 {
+		if t.advanceCursor() {
+			goto RETRY
+		}
+
 		return false
 	}
-	var arr *cursors.FloatArray
-	var len int
-	for {
-		arr = t.cur.Next()
-		len = arr.Len()
-		if len > 0 {
-			break
-		}
-		if !t.advanceCursor() {
-			return false
-		}
-	}
 
-	// handle the group without aggregate case
-	if t.gc.Aggregate() == nil {
-		// Retrieve the buffer for the data to avoid allocating
-		// additional slices. If the buffer is still being used
-		// because the references were retained, then we will
-		// allocate a new buffer.
-		colReader := t.allocateBuffer(len)
-		colReader.cols[timeColIdx] = arrow.NewInt(arr.Timestamps, t.alloc)
-		colReader.cols[valueColIdx] = t.toArrowBuffer(arr.Values)
-		t.appendTags(colReader)
-		t.appendBounds(colReader)
-		return true
-	}
-
-	// handle the group with aggregate case
-	var value float64
-	// For group count, sum, min, and max, the timestamp here is always math.MaxInt64.
-	// their final result does not contain _time, so this timestamp value can be anything
-	// and it won't matter.
-	// For group first, we need to assign the initial value to math.MaxInt64 so
-	// we can find the row with the smallest timestamp.
-	// Do not worry about data with math.MaxInt64 as its real timestamp.
-	// In OSS we require a |> range() call in the query and a math.MaxInt64 timestamp
-	// cannot make it through.
-	var timestamp int64 = math.MaxInt64
-	if t.gc.Aggregate().Type == datatypes.AggregateTypeLast {
-		timestamp = math.MinInt64
-	}
-	for {
-		// note that for the group aggregate case, len here should always be 1
-		for i := 0; i < len; i++ {
-			switch t.gc.Aggregate().Type {
-			case datatypes.AggregateTypeCount:
-				panic("unsupported for aggregate count: Float")
-			case datatypes.AggregateTypeSum:
-				value += arr.Values[i]
-			case datatypes.AggregateTypeFirst:
-				if arr.Timestamps[i] < timestamp {
-					timestamp = arr.Timestamps[i]
-					value = arr.Values[i]
-				}
-			case datatypes.AggregateTypeLast:
-				if arr.Timestamps[i] > timestamp {
-					timestamp = arr.Timestamps[i]
-					value = arr.Values[i]
-				}
-			}
-		}
-		arr = t.cur.Next()
-		len = arr.Len()
-		if len > 0 {
-			continue
-		}
-		if !t.advanceCursor() {
-			break
-		}
-	}
-	colReader := t.allocateBuffer(1)
-	if IsSelector(t.gc.Aggregate()) {
-		colReader.cols[timeColIdx] = arrow.NewInt([]int64{timestamp}, t.alloc)
-		colReader.cols[valueColIdx] = t.toArrowBuffer([]float64{value})
-	} else {
-		colReader.cols[valueColIdxWithoutTime] = t.toArrowBuffer([]float64{value})
-	}
-	t.appendTags(colReader)
-	t.appendBounds(colReader)
+	// Retrieve the buffer for the data to avoid allocating
+	// additional slices. If the buffer is still being used
+	// because the references were retained, then we will
+	// allocate a new buffer.
+	cr := t.allocateBuffer(l)
+	cr.cols[timeColIdx] = arrow.NewInt(a.Timestamps, t.alloc)
+	cr.cols[valueColIdx] = t.toArrowBuffer(a.Values)
+	t.appendTags(cr)
+	t.appendBounds(cr)
 	return true
 }
 
@@ -1129,7 +1061,7 @@ func (t *integerWindowTable) advance() bool {
 	} else {
 		cr.cols[startColIdx] = start
 		cr.cols[stopColIdx] = stop
-		cr.cols[valueColIdxWithoutTime] = values
+		cr.cols[windowedValueColIdx] = values
 	}
 	t.appendTags(cr)
 	return true
@@ -1532,93 +1464,26 @@ func (t *integerGroupTable) Do(f func(flux.ColReader) error) error {
 }
 
 func (t *integerGroupTable) advance() bool {
-	if t.cur == nil {
-		// For group aggregates, we will try to get all the series and all table buffers within those series
-		// all at once and merge them into one row when this advance() function is first called.
-		// At the end of this process, t.advanceCursor() already returns false and t.cur becomes nil.
-		// But we still need to return true to indicate that there is data to be returned.
-		// The second time when we call this advance(), t.cur is already nil, so we directly return false.
+RETRY:
+	a := t.cur.Next()
+	l := a.Len()
+	if l == 0 {
+		if t.advanceCursor() {
+			goto RETRY
+		}
+
 		return false
 	}
-	var arr *cursors.IntegerArray
-	var len int
-	for {
-		arr = t.cur.Next()
-		len = arr.Len()
-		if len > 0 {
-			break
-		}
-		if !t.advanceCursor() {
-			return false
-		}
-	}
 
-	// handle the group without aggregate case
-	if t.gc.Aggregate() == nil {
-		// Retrieve the buffer for the data to avoid allocating
-		// additional slices. If the buffer is still being used
-		// because the references were retained, then we will
-		// allocate a new buffer.
-		colReader := t.allocateBuffer(len)
-		colReader.cols[timeColIdx] = arrow.NewInt(arr.Timestamps, t.alloc)
-		colReader.cols[valueColIdx] = t.toArrowBuffer(arr.Values)
-		t.appendTags(colReader)
-		t.appendBounds(colReader)
-		return true
-	}
-
-	// handle the group with aggregate case
-	var value int64
-	// For group count, sum, min, and max, the timestamp here is always math.MaxInt64.
-	// their final result does not contain _time, so this timestamp value can be anything
-	// and it won't matter.
-	// For group first, we need to assign the initial value to math.MaxInt64 so
-	// we can find the row with the smallest timestamp.
-	// Do not worry about data with math.MaxInt64 as its real timestamp.
-	// In OSS we require a |> range() call in the query and a math.MaxInt64 timestamp
-	// cannot make it through.
-	var timestamp int64 = math.MaxInt64
-	if t.gc.Aggregate().Type == datatypes.AggregateTypeLast {
-		timestamp = math.MinInt64
-	}
-	for {
-		// note that for the group aggregate case, len here should always be 1
-		for i := 0; i < len; i++ {
-			switch t.gc.Aggregate().Type {
-			case datatypes.AggregateTypeCount:
-				fallthrough
-			case datatypes.AggregateTypeSum:
-				value += arr.Values[i]
-			case datatypes.AggregateTypeFirst:
-				if arr.Timestamps[i] < timestamp {
-					timestamp = arr.Timestamps[i]
-					value = arr.Values[i]
-				}
-			case datatypes.AggregateTypeLast:
-				if arr.Timestamps[i] > timestamp {
-					timestamp = arr.Timestamps[i]
-					value = arr.Values[i]
-				}
-			}
-		}
-		arr = t.cur.Next()
-		len = arr.Len()
-		if len > 0 {
-			continue
-		}
-		if !t.advanceCursor() {
-			break
-		}
-	}
-	colReader := t.allocateBuffer(1)
-	if IsSelector(t.gc.Aggregate()) {
-		colReader.cols[timeColIdx] = arrow.NewInt([]int64{timestamp}, t.alloc)
-		colReader.cols[valueColIdx] = t.toArrowBuffer([]int64{value})
-	} else {
-		colReader.cols[valueColIdxWithoutTime] = t.toArrowBuffer([]int64{value})
-	}
-	t.appendTags(colReader)
-	t.appendBounds(colReader)
+	// Retrieve the buffer for the data to avoid allocating
+	// additional slices. If the buffer is still being used
+	// because the references were retained, then we will
+	// allocate a new buffer.
+	cr := t.allocateBuffer(l)
+	cr.cols[timeColIdx] = arrow.NewInt(a.Timestamps, t.alloc)
+	cr.cols[valueColIdx] = t.toArrowBuffer(a.Values)
+	t.appendTags(cr)
+	t.appendBounds(cr)
 	return true
 }
 
@@ -1948,7 +1813,7 @@ func (t *unsignedWindowTable) advance() bool {
 	} else {
 		cr.cols[startColIdx] = start
 		cr.cols[stopColIdx] = stop
-		cr.cols[valueColIdxWithoutTime] = values
+		cr.cols[windowedValueColIdx] = values
 	}
 	t.appendTags(cr)
 	return true
@@ -2351,93 +2216,26 @@ func (t *unsignedGroupTable) Do(f func(flux.ColReader) error) error {
 }
 
 func (t *unsignedGroupTable) advance() bool {
-	if t.cur == nil {
-		// For group aggregates, we will try to get all the series and all table buffers within those series
-		// all at once and merge them into one row when this advance() function is first called.
-		// At the end of this process, t.advanceCursor() already returns false and t.cur becomes nil.
-		// But we still need to return true to indicate that there is data to be returned.
-		// The second time when we call this advance(), t.cur is already nil, so we directly return false.
+RETRY:
+	a := t.cur.Next()
+	l := a.Len()
+	if l == 0 {
+		if t.advanceCursor() {
+			goto RETRY
+		}
+
 		return false
 	}
-	var arr *cursors.UnsignedArray
-	var len int
-	for {
-		arr = t.cur.Next()
-		len = arr.Len()
-		if len > 0 {
-			break
-		}
-		if !t.advanceCursor() {
-			return false
-		}
-	}
 
-	// handle the group without aggregate case
-	if t.gc.Aggregate() == nil {
-		// Retrieve the buffer for the data to avoid allocating
-		// additional slices. If the buffer is still being used
-		// because the references were retained, then we will
-		// allocate a new buffer.
-		colReader := t.allocateBuffer(len)
-		colReader.cols[timeColIdx] = arrow.NewInt(arr.Timestamps, t.alloc)
-		colReader.cols[valueColIdx] = t.toArrowBuffer(arr.Values)
-		t.appendTags(colReader)
-		t.appendBounds(colReader)
-		return true
-	}
-
-	// handle the group with aggregate case
-	var value uint64
-	// For group count, sum, min, and max, the timestamp here is always math.MaxInt64.
-	// their final result does not contain _time, so this timestamp value can be anything
-	// and it won't matter.
-	// For group first, we need to assign the initial value to math.MaxInt64 so
-	// we can find the row with the smallest timestamp.
-	// Do not worry about data with math.MaxInt64 as its real timestamp.
-	// In OSS we require a |> range() call in the query and a math.MaxInt64 timestamp
-	// cannot make it through.
-	var timestamp int64 = math.MaxInt64
-	if t.gc.Aggregate().Type == datatypes.AggregateTypeLast {
-		timestamp = math.MinInt64
-	}
-	for {
-		// note that for the group aggregate case, len here should always be 1
-		for i := 0; i < len; i++ {
-			switch t.gc.Aggregate().Type {
-			case datatypes.AggregateTypeCount:
-				panic("unsupported for aggregate count: Unsigned")
-			case datatypes.AggregateTypeSum:
-				value += arr.Values[i]
-			case datatypes.AggregateTypeFirst:
-				if arr.Timestamps[i] < timestamp {
-					timestamp = arr.Timestamps[i]
-					value = arr.Values[i]
-				}
-			case datatypes.AggregateTypeLast:
-				if arr.Timestamps[i] > timestamp {
-					timestamp = arr.Timestamps[i]
-					value = arr.Values[i]
-				}
-			}
-		}
-		arr = t.cur.Next()
-		len = arr.Len()
-		if len > 0 {
-			continue
-		}
-		if !t.advanceCursor() {
-			break
-		}
-	}
-	colReader := t.allocateBuffer(1)
-	if IsSelector(t.gc.Aggregate()) {
-		colReader.cols[timeColIdx] = arrow.NewInt([]int64{timestamp}, t.alloc)
-		colReader.cols[valueColIdx] = t.toArrowBuffer([]uint64{value})
-	} else {
-		colReader.cols[valueColIdxWithoutTime] = t.toArrowBuffer([]uint64{value})
-	}
-	t.appendTags(colReader)
-	t.appendBounds(colReader)
+	// Retrieve the buffer for the data to avoid allocating
+	// additional slices. If the buffer is still being used
+	// because the references were retained, then we will
+	// allocate a new buffer.
+	cr := t.allocateBuffer(l)
+	cr.cols[timeColIdx] = arrow.NewInt(a.Timestamps, t.alloc)
+	cr.cols[valueColIdx] = t.toArrowBuffer(a.Values)
+	t.appendTags(cr)
+	t.appendBounds(cr)
 	return true
 }
 
@@ -2767,7 +2565,7 @@ func (t *stringWindowTable) advance() bool {
 	} else {
 		cr.cols[startColIdx] = start
 		cr.cols[stopColIdx] = stop
-		cr.cols[valueColIdxWithoutTime] = values
+		cr.cols[windowedValueColIdx] = values
 	}
 	t.appendTags(cr)
 	return true
@@ -3170,93 +2968,26 @@ func (t *stringGroupTable) Do(f func(flux.ColReader) error) error {
 }
 
 func (t *stringGroupTable) advance() bool {
-	if t.cur == nil {
-		// For group aggregates, we will try to get all the series and all table buffers within those series
-		// all at once and merge them into one row when this advance() function is first called.
-		// At the end of this process, t.advanceCursor() already returns false and t.cur becomes nil.
-		// But we still need to return true to indicate that there is data to be returned.
-		// The second time when we call this advance(), t.cur is already nil, so we directly return false.
+RETRY:
+	a := t.cur.Next()
+	l := a.Len()
+	if l == 0 {
+		if t.advanceCursor() {
+			goto RETRY
+		}
+
 		return false
 	}
-	var arr *cursors.StringArray
-	var len int
-	for {
-		arr = t.cur.Next()
-		len = arr.Len()
-		if len > 0 {
-			break
-		}
-		if !t.advanceCursor() {
-			return false
-		}
-	}
 
-	// handle the group without aggregate case
-	if t.gc.Aggregate() == nil {
-		// Retrieve the buffer for the data to avoid allocating
-		// additional slices. If the buffer is still being used
-		// because the references were retained, then we will
-		// allocate a new buffer.
-		colReader := t.allocateBuffer(len)
-		colReader.cols[timeColIdx] = arrow.NewInt(arr.Timestamps, t.alloc)
-		colReader.cols[valueColIdx] = t.toArrowBuffer(arr.Values)
-		t.appendTags(colReader)
-		t.appendBounds(colReader)
-		return true
-	}
-
-	// handle the group with aggregate case
-	var value string
-	// For group count, sum, min, and max, the timestamp here is always math.MaxInt64.
-	// their final result does not contain _time, so this timestamp value can be anything
-	// and it won't matter.
-	// For group first, we need to assign the initial value to math.MaxInt64 so
-	// we can find the row with the smallest timestamp.
-	// Do not worry about data with math.MaxInt64 as its real timestamp.
-	// In OSS we require a |> range() call in the query and a math.MaxInt64 timestamp
-	// cannot make it through.
-	var timestamp int64 = math.MaxInt64
-	if t.gc.Aggregate().Type == datatypes.AggregateTypeLast {
-		timestamp = math.MinInt64
-	}
-	for {
-		// note that for the group aggregate case, len here should always be 1
-		for i := 0; i < len; i++ {
-			switch t.gc.Aggregate().Type {
-			case datatypes.AggregateTypeCount:
-				panic("unsupported for aggregate count: String")
-			case datatypes.AggregateTypeSum:
-				panic("unsupported for aggregate sum: String")
-			case datatypes.AggregateTypeFirst:
-				if arr.Timestamps[i] < timestamp {
-					timestamp = arr.Timestamps[i]
-					value = arr.Values[i]
-				}
-			case datatypes.AggregateTypeLast:
-				if arr.Timestamps[i] > timestamp {
-					timestamp = arr.Timestamps[i]
-					value = arr.Values[i]
-				}
-			}
-		}
-		arr = t.cur.Next()
-		len = arr.Len()
-		if len > 0 {
-			continue
-		}
-		if !t.advanceCursor() {
-			break
-		}
-	}
-	colReader := t.allocateBuffer(1)
-	if IsSelector(t.gc.Aggregate()) {
-		colReader.cols[timeColIdx] = arrow.NewInt([]int64{timestamp}, t.alloc)
-		colReader.cols[valueColIdx] = t.toArrowBuffer([]string{value})
-	} else {
-		colReader.cols[valueColIdxWithoutTime] = t.toArrowBuffer([]string{value})
-	}
-	t.appendTags(colReader)
-	t.appendBounds(colReader)
+	// Retrieve the buffer for the data to avoid allocating
+	// additional slices. If the buffer is still being used
+	// because the references were retained, then we will
+	// allocate a new buffer.
+	cr := t.allocateBuffer(l)
+	cr.cols[timeColIdx] = arrow.NewInt(a.Timestamps, t.alloc)
+	cr.cols[valueColIdx] = t.toArrowBuffer(a.Values)
+	t.appendTags(cr)
+	t.appendBounds(cr)
 	return true
 }
 
@@ -3586,7 +3317,7 @@ func (t *booleanWindowTable) advance() bool {
 	} else {
 		cr.cols[startColIdx] = start
 		cr.cols[stopColIdx] = stop
-		cr.cols[valueColIdxWithoutTime] = values
+		cr.cols[windowedValueColIdx] = values
 	}
 	t.appendTags(cr)
 	return true
@@ -3989,93 +3720,26 @@ func (t *booleanGroupTable) Do(f func(flux.ColReader) error) error {
 }
 
 func (t *booleanGroupTable) advance() bool {
-	if t.cur == nil {
-		// For group aggregates, we will try to get all the series and all table buffers within those series
-		// all at once and merge them into one row when this advance() function is first called.
-		// At the end of this process, t.advanceCursor() already returns false and t.cur becomes nil.
-		// But we still need to return true to indicate that there is data to be returned.
-		// The second time when we call this advance(), t.cur is already nil, so we directly return false.
+RETRY:
+	a := t.cur.Next()
+	l := a.Len()
+	if l == 0 {
+		if t.advanceCursor() {
+			goto RETRY
+		}
+
 		return false
 	}
-	var arr *cursors.BooleanArray
-	var len int
-	for {
-		arr = t.cur.Next()
-		len = arr.Len()
-		if len > 0 {
-			break
-		}
-		if !t.advanceCursor() {
-			return false
-		}
-	}
 
-	// handle the group without aggregate case
-	if t.gc.Aggregate() == nil {
-		// Retrieve the buffer for the data to avoid allocating
-		// additional slices. If the buffer is still being used
-		// because the references were retained, then we will
-		// allocate a new buffer.
-		colReader := t.allocateBuffer(len)
-		colReader.cols[timeColIdx] = arrow.NewInt(arr.Timestamps, t.alloc)
-		colReader.cols[valueColIdx] = t.toArrowBuffer(arr.Values)
-		t.appendTags(colReader)
-		t.appendBounds(colReader)
-		return true
-	}
-
-	// handle the group with aggregate case
-	var value bool
-	// For group count, sum, min, and max, the timestamp here is always math.MaxInt64.
-	// their final result does not contain _time, so this timestamp value can be anything
-	// and it won't matter.
-	// For group first, we need to assign the initial value to math.MaxInt64 so
-	// we can find the row with the smallest timestamp.
-	// Do not worry about data with math.MaxInt64 as its real timestamp.
-	// In OSS we require a |> range() call in the query and a math.MaxInt64 timestamp
-	// cannot make it through.
-	var timestamp int64 = math.MaxInt64
-	if t.gc.Aggregate().Type == datatypes.AggregateTypeLast {
-		timestamp = math.MinInt64
-	}
-	for {
-		// note that for the group aggregate case, len here should always be 1
-		for i := 0; i < len; i++ {
-			switch t.gc.Aggregate().Type {
-			case datatypes.AggregateTypeCount:
-				panic("unsupported for aggregate count: Boolean")
-			case datatypes.AggregateTypeSum:
-				panic("unsupported for aggregate sum: Boolean")
-			case datatypes.AggregateTypeFirst:
-				if arr.Timestamps[i] < timestamp {
-					timestamp = arr.Timestamps[i]
-					value = arr.Values[i]
-				}
-			case datatypes.AggregateTypeLast:
-				if arr.Timestamps[i] > timestamp {
-					timestamp = arr.Timestamps[i]
-					value = arr.Values[i]
-				}
-			}
-		}
-		arr = t.cur.Next()
-		len = arr.Len()
-		if len > 0 {
-			continue
-		}
-		if !t.advanceCursor() {
-			break
-		}
-	}
-	colReader := t.allocateBuffer(1)
-	if IsSelector(t.gc.Aggregate()) {
-		colReader.cols[timeColIdx] = arrow.NewInt([]int64{timestamp}, t.alloc)
-		colReader.cols[valueColIdx] = t.toArrowBuffer([]bool{value})
-	} else {
-		colReader.cols[valueColIdxWithoutTime] = t.toArrowBuffer([]bool{value})
-	}
-	t.appendTags(colReader)
-	t.appendBounds(colReader)
+	// Retrieve the buffer for the data to avoid allocating
+	// additional slices. If the buffer is still being used
+	// because the references were retained, then we will
+	// allocate a new buffer.
+	cr := t.allocateBuffer(l)
+	cr.cols[timeColIdx] = arrow.NewInt(a.Timestamps, t.alloc)
+	cr.cols[valueColIdx] = t.toArrowBuffer(a.Values)
+	t.appendTags(cr)
+	t.appendBounds(cr)
 	return true
 }
 

storage/flux/table.gen.go.tmpl

@@ -12,8 +12,7 @@ import (
 	"github.com/influxdata/influxdb/v2"
 	"github.com/influxdata/influxdb/v2/models"
 	storage "github.com/influxdata/influxdb/v2/storage/reads"
-	"github.com/influxdata/influxdb/v2/storage/reads/datatypes"
-	"github.com/influxdata/influxdb/v2/tsdb/cursors"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/cursors"
 )
 {{range .}}
 //
@@ -304,7 +303,7 @@ func (t *{{.name}}WindowTable) advance() bool {
 	} else {
 		cr.cols[startColIdx] = start
 		cr.cols[stopColIdx] = stop
-		cr.cols[valueColIdxWithoutTime] = values
+		cr.cols[windowedValueColIdx] = values
 	}
 	t.appendTags(cr)
 	return true
@@ -707,93 +706,26 @@ func (t *{{.name}}GroupTable) Do(f func(flux.ColReader) error) error {
 }
 
 func (t *{{.name}}GroupTable) advance() bool {
-	if t.cur == nil {
-		// For group aggregates, we will try to get all the series and all table buffers within those series
-		// all at once and merge them into one row when this advance() function is first called.
-		// At the end of this process, t.advanceCursor() already returns false and t.cur becomes nil.
-		// But we still need to return true to indicate that there is data to be returned.
-		// The second time when we call this advance(), t.cur is already nil, so we directly return false.
+RETRY:
+	a := t.cur.Next()
+	l := a.Len()
+	if l == 0 {
+		if t.advanceCursor() {
+			goto RETRY
+		}
+
 		return false
 	}
-	var arr *cursors.{{.Name}}Array
-	var len int
-	for {
-		arr = t.cur.Next()
-		len = arr.Len()
-		if len > 0 {
-			break
-		}
-		if !t.advanceCursor() {
-			return false
-		}
-	}
 
-	// handle the group without aggregate case
-	if t.gc.Aggregate() == nil {
-		// Retrieve the buffer for the data to avoid allocating
-		// additional slices. If the buffer is still being used
-		// because the references were retained, then we will
-		// allocate a new buffer.
-		colReader := t.allocateBuffer(len)
-		colReader.cols[timeColIdx] = arrow.NewInt(arr.Timestamps, t.alloc)
-		colReader.cols[valueColIdx] = t.toArrowBuffer(arr.Values)
-		t.appendTags(colReader)
-		t.appendBounds(colReader)
-		return true
-	}
-
-	// handle the group with aggregate case
-	var value {{.Type}}
-	// For group count, sum, min, and max, the timestamp here is always math.MaxInt64.
-	// their final result does not contain _time, so this timestamp value can be anything
-	// and it won't matter.
-	// For group first, we need to assign the initial value to math.MaxInt64 so
-	// we can find the row with the smallest timestamp.
-	// Do not worry about data with math.MaxInt64 as its real timestamp.
-	// In OSS we require a |> range() call in the query and a math.MaxInt64 timestamp
-	// cannot make it through.
-	var timestamp int64 = math.MaxInt64
-	if t.gc.Aggregate().Type == datatypes.AggregateTypeLast {
-		timestamp = math.MinInt64
-	}
-	for {
-		// note that for the group aggregate case, len here should always be 1
-		for i := 0; i < len; i++ {
-			switch t.gc.Aggregate().Type {
-			case datatypes.AggregateTypeCount:
-				{{if eq .Name "Integer"}}fallthrough{{else}}panic("unsupported for aggregate count: {{.Name}}"){{end}}
-			case datatypes.AggregateTypeSum:
-				{{if or (eq .Name "String") (eq .Name "Boolean")}}panic("unsupported for aggregate sum: {{.Name}}"){{else}}value += arr.Values[i]{{end}}
-			case datatypes.AggregateTypeFirst:
-				if arr.Timestamps[i] < timestamp {
-					timestamp = arr.Timestamps[i]
-					value = arr.Values[i]
-				}
-			case datatypes.AggregateTypeLast:
-				if arr.Timestamps[i] > timestamp {
-					timestamp = arr.Timestamps[i]
-					value = arr.Values[i]
-				}
-			}
-		}
-		arr = t.cur.Next()
-		len = arr.Len()
-		if len > 0 {
-			continue
-		}
-		if !t.advanceCursor() {
-			break
-		}
-	}
-	colReader := t.allocateBuffer(1)
-	if IsSelector(t.gc.Aggregate()) {
-		colReader.cols[timeColIdx] = arrow.NewInt([]int64{timestamp}, t.alloc)
-		colReader.cols[valueColIdx] = t.toArrowBuffer([]{{.Type}}{value})
-	} else {
-		colReader.cols[valueColIdxWithoutTime] = t.toArrowBuffer([]{{.Type}}{value})
-	}
-	t.appendTags(colReader)
-	t.appendBounds(colReader)
+	// Retrieve the buffer for the data to avoid allocating
+	// additional slices. If the buffer is still being used
+	// because the references were retained, then we will
+	// allocate a new buffer.
+	cr := t.allocateBuffer(l)
+	cr.cols[timeColIdx] = arrow.NewInt(a.Timestamps, t.alloc)
+	cr.cols[valueColIdx] = t.toArrowBuffer(a.Values)
+	t.appendTags(cr)
+	t.appendBounds(cr)
 	return true
 }
 

storage/points_writer.go

@@ -6,7 +6,6 @@ import (
 	"time"
 
 	"github.com/influxdata/influxdb/v2"
-	"github.com/influxdata/influxdb/v2/tsdb"
 	"github.com/influxdata/influxdb/v2/v1/models"
 )
 
@@ -52,20 +51,16 @@ func (w *LoggingPointsWriter) WritePoints(ctx context.Context, orgID influxdb.ID
 	}
 
 	// Log error to bucket.
-	name := tsdb.EncodeName(orgID, bkts[0].ID)
 	pt, e := models.NewPoint(
-		string(name[:]),
-		models.NewTags(map[string]string{
-			models.MeasurementTagKey: "write_errors",
-			models.FieldKeyTagKey:    "error"},
-		),
+		"write_errors",
+		nil,
 		models.Fields{"error": err.Error()},
 		time.Now(),
 	)
 	if e != nil {
 		return e
 	}
-	if e := w.Underlying.WritePoints(ctx, orgID, bucketID, []models.Point{pt}); e != nil {
+	if e := w.Underlying.WritePoints(ctx, orgID, bkts[0].ID, []models.Point{pt}); e != nil {
 		return e
 	}
 

storage/points_writer_test.go

@@ -13,7 +13,7 @@ import (
 	"github.com/influxdata/influxdb/v2/mock"
 	"github.com/influxdata/influxdb/v2/models"
 	"github.com/influxdata/influxdb/v2/storage"
-	"github.com/influxdata/influxdb/v2/tsdb"
+	"github.com/influxdata/influxdb/v2/v1/tsdb"
 )
 
 func TestLoggingPointsWriter(t *testing.T) {

storage/reads/aggregate_resultset.go

@@ -8,7 +8,7 @@ import (
 	"github.com/influxdata/influxdb/v2/kit/tracing"
 	"github.com/influxdata/influxdb/v2/models"
 	"github.com/influxdata/influxdb/v2/storage/reads/datatypes"
-	"github.com/influxdata/influxdb/v2/tsdb/cursors"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/cursors"
 )
 
 type windowAggregateResultSet struct {

storage/reads/aggregate_resultset_test.go

@@ -8,7 +8,7 @@ import (
 	"github.com/influxdata/influxdb/v2/models"
 	"github.com/influxdata/influxdb/v2/storage/reads"
 	"github.com/influxdata/influxdb/v2/storage/reads/datatypes"
-	"github.com/influxdata/influxdb/v2/tsdb/cursors"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/cursors"
 )
 
 func TestNewWindowAggregateResultSet_Tags(t *testing.T) {

storage/reads/array_cursor.gen.go

@@ -11,7 +11,7 @@ import (
 	"fmt"
 	"math"
 
-	"github.com/influxdata/influxdb/v2/tsdb/cursors"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/cursors"
 )
 
 const (

storage/reads/array_cursor.gen.go.tmpl

@@ -5,7 +5,7 @@ import (
 	"fmt"
 	"math"
 
-	"github.com/influxdata/influxdb/v2/tsdb/cursors"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/cursors"
 )
 
 const (

storage/reads/array_cursor.go

@@ -5,7 +5,7 @@ import (
 	"fmt"
 
 	"github.com/influxdata/influxdb/v2/storage/reads/datatypes"
-	"github.com/influxdata/influxdb/v2/tsdb/cursors"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/cursors"
 )
 
 type singleValue struct {

storage/reads/array_cursor_test.go

@@ -6,7 +6,7 @@ import (
 	"time"
 
 	"github.com/google/go-cmp/cmp"
-	"github.com/influxdata/influxdb/v2/tsdb/cursors"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/cursors"
 )
 
 func TestIntegerFilterArrayCursor(t *testing.T) {

storage/reads/group_resultset.go

@@ -9,7 +9,7 @@ import (
 	"github.com/influxdata/influxdb/v2/kit/tracing"
 	"github.com/influxdata/influxdb/v2/models"
 	"github.com/influxdata/influxdb/v2/storage/reads/datatypes"
-	"github.com/influxdata/influxdb/v2/tsdb/cursors"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/cursors"
 )
 
 type groupResultSet struct {

storage/reads/resultset.go

@@ -5,7 +5,7 @@ import (
 
 	"github.com/influxdata/influxdb/v2/models"
 	"github.com/influxdata/influxdb/v2/storage/reads/datatypes"
-	"github.com/influxdata/influxdb/v2/tsdb/cursors"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/cursors"
 )
 
 type resultSet struct {

storage/reads/resultset_lineprotocol.go

@@ -6,7 +6,7 @@ import (
 	"strconv"
 
 	"github.com/influxdata/influxdb/v2/models"
-	"github.com/influxdata/influxdb/v2/tsdb/cursors"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/cursors"
 )
 
 // ResultSetToLineProtocol transforms rs to line protocol and writes the

storage/reads/series_cursor.go

@@ -10,7 +10,7 @@ import (
 	"github.com/influxdata/influxdb/v2/query"
 	"github.com/influxdata/influxdb/v2/storage"
 	"github.com/influxdata/influxdb/v2/storage/reads/datatypes"
-	"github.com/influxdata/influxdb/v2/tsdb/cursors"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/cursors"
 	"github.com/influxdata/influxql"
 )
 

storage/reads/store.go

@@ -7,7 +7,7 @@ import (
 	"github.com/influxdata/influxdb/v2/models"
 	"github.com/influxdata/influxdb/v2/query"
 	"github.com/influxdata/influxdb/v2/storage/reads/datatypes"
-	"github.com/influxdata/influxdb/v2/tsdb/cursors"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/cursors"
 )
 
 type ResultSet interface {

storage/reads/store_test.go

@@ -8,7 +8,7 @@ import (
 
 	"github.com/influxdata/influxdb/v2/models"
 	"github.com/influxdata/influxdb/v2/storage/reads"
-	"github.com/influxdata/influxdb/v2/tsdb/cursors"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/cursors"
 )
 
 func cursorToString(wr io.Writer, cur cursors.Cursor) {

storage/reads/viewer.go

@@ -5,7 +5,7 @@ import (
 
 	"github.com/influxdata/influxdb/v2"
 	"github.com/influxdata/influxdb/v2/storage"
-	"github.com/influxdata/influxdb/v2/tsdb/cursors"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/cursors"
 	"github.com/influxdata/influxql"
 )
 

storage/readservice/store.go

@@ -9,7 +9,7 @@ import (
 	"github.com/influxdata/influxdb/v2/models"
 	"github.com/influxdata/influxdb/v2/storage/reads"
 	"github.com/influxdata/influxdb/v2/storage/reads/datatypes"
-	"github.com/influxdata/influxdb/v2/tsdb/cursors"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/cursors"
 	"github.com/influxdata/influxql"
 )
 

storage/retention.go

@@ -9,7 +9,7 @@ import (
 	"github.com/influxdata/influxdb/v2"
 	"github.com/influxdata/influxdb/v2/kit/tracing"
 	"github.com/influxdata/influxdb/v2/logger"
-	"github.com/influxdata/influxdb/v2/tsdb/tsm1"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/engine/tsm1"
 	"github.com/prometheus/client_golang/prometheus"
 	"go.uber.org/zap"
 	"go.uber.org/zap/zapcore"

storage/retention_test.go

@@ -18,8 +18,8 @@ import (
 	"github.com/influxdata/influxdb/v2/kit/prom/promtest"
 	"github.com/influxdata/influxdb/v2/logger"
 	"github.com/influxdata/influxdb/v2/toml"
-	"github.com/influxdata/influxdb/v2/tsdb"
-	"github.com/influxdata/influxdb/v2/tsdb/tsm1"
+	"github.com/influxdata/influxdb/v2/v1/tsdb"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/engine/tsm1"
 	"github.com/prometheus/client_golang/prometheus"
 	"go.uber.org/zap"
 )

storage/series_cursor.go

@@ -7,9 +7,8 @@ import (
 	"github.com/influxdata/influxdb/v2"
 	"github.com/influxdata/influxdb/v2/models"
 	"github.com/influxdata/influxdb/v2/pkg/lifecycle"
-	"github.com/influxdata/influxdb/v2/tsdb"
-	"github.com/influxdata/influxdb/v2/tsdb/seriesfile"
-	"github.com/influxdata/influxdb/v2/tsdb/tsi1"
+	"github.com/influxdata/influxdb/v2/v1/tsdb"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/index/tsi1"
 	"github.com/influxdata/influxql"
 )
 

storage/series_cursor_test.go

@@ -5,8 +5,7 @@ import (
 
 	"github.com/google/go-cmp/cmp"
 	"github.com/influxdata/influxdb/v2"
-	"github.com/influxdata/influxdb/v2/tsdb"
-	"github.com/influxdata/influxdb/v2/tsdb/seriesfile"
+	"github.com/influxdata/influxdb/v2/v1/tsdb"
 )
 
 func Test_NewSeriesCursor_UnexpectedOrg(t *testing.T) {

storage/wal/dump.go

@@ -10,8 +10,8 @@ import (
 	"text/tabwriter"
 
 	"github.com/influxdata/influxdb/v2/storage/reads/datatypes"
-	"github.com/influxdata/influxdb/v2/tsdb"
-	"github.com/influxdata/influxdb/v2/tsdb/value"
+	"github.com/influxdata/influxdb/v2/v1/tsdb"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/engine/tsm1"
 )
 
 // Command represents the program execution for "influxd inspect dumpmwal

storage/wal/dump_test.go

@@ -12,8 +12,8 @@ import (
 	"github.com/google/go-cmp/cmp"
 	"github.com/google/go-cmp/cmp/cmpopts"
 	"github.com/influxdata/influxdb/v2"
-	"github.com/influxdata/influxdb/v2/tsdb"
-	"github.com/influxdata/influxdb/v2/tsdb/value"
+	"github.com/influxdata/influxdb/v2/v1/tsdb"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/engine/tsm1"
 )
 
 func TestWalDump_RunWriteEntries(t *testing.T) {

storage/wal/verifier_test.go

@@ -2,14 +2,15 @@ package wal
 
 import (
 	"context"
-	"github.com/google/go-cmp/cmp"
-	"github.com/google/go-cmp/cmp/cmpopts"
-	"github.com/influxdata/influxdb/v2/kit/errors"
-	"github.com/influxdata/influxdb/v2/tsdb/value"
 	"io/ioutil"
 	"math/rand"
 	"os"
 	"testing"
+
+	"github.com/google/go-cmp/cmp"
+	"github.com/google/go-cmp/cmp/cmpopts"
+	"github.com/influxdata/influxdb/v2/kit/errors"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/engine/tsm1"
 )
 
 type Test struct {

storage/wal/wal.go

@@ -25,7 +25,7 @@ import (
 	"github.com/influxdata/influxdb/v2/kit/tracing"
 	"github.com/influxdata/influxdb/v2/pkg/limiter"
 	"github.com/influxdata/influxdb/v2/pkg/pool"
-	"github.com/influxdata/influxdb/v2/tsdb/value"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/engine/tsm1"
 )
 
 const (

storage/wal/wal_test.go

@@ -12,7 +12,7 @@ import (
 	"github.com/golang/snappy"
 
 	"github.com/influxdata/influxdb/v2"
-	"github.com/influxdata/influxdb/v2/tsdb/value"
+	"github.com/influxdata/influxdb/v2/v1/tsdb/engine/tsm1/value"
 )
 
 func TestWALWriter_WriteMulti_Single(t *testing.T) {

task/backend/run_recorder.go

@@ -9,7 +9,7 @@ import (
 	"github.com/influxdata/influxdb/v2"
 	"github.com/influxdata/influxdb/v2/models"
 	"github.com/influxdata/influxdb/v2/storage"
-	"github.com/influxdata/influxdb/v2/tsdb"
+	"github.com/influxdata/influxdb/v2/v1/tsdb"
 	"go.uber.org/zap"
 )
 

tsdb/cursors/arrayvalues.gen.go.tmpl

@@ -1,259 +0,0 @@
-package cursors
-
-{{range .}}
-{{- $typename := print .Name "Array" }}
-{{- $hasType  := or (and .Type true) false }}
-
-type {{ $typename }} struct {
-	Timestamps []int64
-{{- if $hasType }}
-	Values     []{{.Type}}
-{{- end }}
-}
-
-func New{{$typename}}Len(sz int) *{{$typename}} {
-	return &{{$typename}}{
-		Timestamps: make([]int64, sz),
-{{- if $hasType }}
-		Values: make([]{{.Type}}, sz),
-{{- end }}
-	}
-}
-
-func (a *{{ $typename }}) MinTime() int64 {
-	return a.Timestamps[0]
-}
-
-func (a *{{ $typename }}) MaxTime() int64 {
-	return a.Timestamps[len(a.Timestamps)-1]
-}
-
-func (a *{{ $typename}}) Len() int {
-	return len(a.Timestamps)
-}
-
-// search performs a binary search for UnixNano() v in a
-// and returns the position, i, where v would be inserted.
-// An additional check of a.Timestamps[i] == v is necessary
-// to determine if the value v exists.
-func (a *{{ $typename }}) search(v int64) int {
-	// Define: f(x) → a.Timestamps[x] < v
-	// Define: f(-1) == true, f(n) == false
-	// Invariant: f(lo-1) == true, f(hi) == false
-	lo := 0
-	hi := a.Len()
-	for lo < hi {
-		mid := int(uint(lo+hi) >> 1)
-		if a.Timestamps[mid] < v {
-			lo = mid + 1 // preserves f(lo-1) == true
-		} else {
-			hi = mid // preserves f(hi) == false
-		}
-	}
-
-	// lo == hi
-	return lo
-}
-
-// FindRange returns the positions where min and max would be
-// inserted into the array. If a[0].UnixNano() > max or
-// a[len-1].UnixNano() < min then FindRange returns (-1, -1)
-// indicating the array is outside the [min, max]. The values must
-// be deduplicated and sorted before calling FindRange or the results
-// are undefined.
-func (a *{{ $typename }}) FindRange(min, max int64) (int, int) {
-	if a.Len() == 0 || min > max {
-		return -1, -1
-	}
-
-	minVal := a.MinTime()
-	maxVal := a.MaxTime()
-
-	if maxVal < min || minVal > max {
-		return -1, -1
-	}
-
-	return a.search(min), a.search(max)
-}
-
-{{- if $hasType }}
-// Exclude removes the subset of values in [min, max]. The values must
-// be deduplicated and sorted before calling Exclude or the results are undefined.
-func (a *{{ $typename }}) Exclude(min, max int64) {
-	rmin, rmax := a.FindRange(min, max)
-	if rmin == -1 && rmax == -1 {
-		return
-	}
-
-	// a.Timestamps[rmin] ≥ min
-	// a.Timestamps[rmax] ≥ max
-
-	if rmax < a.Len() {
-		if a.Timestamps[rmax] == max {
-			rmax++
-		}
-		rest := a.Len()-rmax
-		if rest > 0 {
-			ts := a.Timestamps[:rmin+rest]
-			copy(ts[rmin:], a.Timestamps[rmax:])
-			a.Timestamps = ts
-
-			vs := a.Values[:rmin+rest]
-			copy(vs[rmin:], a.Values[rmax:])
-			a.Values = vs
-			return
-		}
-	}
-
-	a.Timestamps = a.Timestamps[:rmin]
-	a.Values = a.Values[:rmin]
-}
-
-// Include returns the subset values between min and max inclusive. The values must
-// be deduplicated and sorted before calling Include or the results are undefined.
-func (a *{{ $typename }}) Include(min, max int64) {
-	rmin, rmax := a.FindRange(min, max)
-	if rmin == -1 && rmax == -1 {
-		a.Timestamps = a.Timestamps[:0]
-		a.Values = a.Values[:0]
-		return
-	}
-
-	// a.Timestamps[rmin] ≥ min
-	// a.Timestamps[rmax] ≥ max
-
-	if rmax < a.Len() && a.Timestamps[rmax] == max {
-		rmax++
-	}
-
-	if rmin > -1 {
-		ts := a.Timestamps[:rmax-rmin]
-		copy(ts, a.Timestamps[rmin:rmax])
-		a.Timestamps = ts
-		vs := a.Values[:rmax-rmin]
-		copy(vs, a.Values[rmin:rmax])
-		a.Values = vs
-	} else {
-		a.Timestamps = a.Timestamps[:rmax]
-		a.Values = a.Values[:rmax]
-	}
-}
-
-// Merge overlays b to top of a.  If two values conflict with
-// the same timestamp, b is used.  Both a and b must be sorted
-// in ascending order.
-func (a *{{ $typename }}) Merge(b *{{ $typename }}) {
-	if a.Len() == 0 {
-		*a = *b
-		return
-	}
-
-	if b.Len() == 0 {
-		return
-	}
-
-	// Normally, both a and b should not contain duplicates.  Due to a bug in older versions, it's
-	// possible stored blocks might contain duplicate values.  Remove them if they exists before
-	// merging.
-	// a = a.Deduplicate()
-	// b = b.Deduplicate()
-
-	if a.MaxTime() < b.MinTime() {
-		a.Timestamps = append(a.Timestamps, b.Timestamps...)
-		a.Values = append(a.Values, b.Values...)
-		return
-	}
-
-	if b.MaxTime() < a.MinTime() {
-		var tmp {{$typename}}
-		tmp.Timestamps = append(b.Timestamps, a.Timestamps...)
-		tmp.Values = append(b.Values, a.Values...)
-		*a = tmp
-		return
-	}
-
-	out := New{{$typename}}Len(a.Len()+b.Len())
-	i, j, k := 0, 0, 0
-	for i < len(a.Timestamps) && j < len(b.Timestamps) {
-		if a.Timestamps[i] < b.Timestamps[j] {
-			out.Timestamps[k] = a.Timestamps[i]
-			out.Values[k] = a.Values[i]
-			i++
-		} else if a.Timestamps[i] == b.Timestamps[j] {
-			out.Timestamps[k] = b.Timestamps[j]
-			out.Values[k] = b.Values[j]
-			i++
-			j++
-		} else {
-			out.Timestamps[k] = b.Timestamps[j]
-			out.Values[k] = b.Values[j]
-			j++
-		}
-		k++
-	}
-
-	if i < len(a.Timestamps) {
-		n := copy(out.Timestamps[k:], a.Timestamps[i:])
-		copy(out.Values[k:], a.Values[i:])
-		k += n
-	} else if j < len(b.Timestamps) {
-		n := copy(out.Timestamps[k:], b.Timestamps[j:])
-		copy(out.Values[k:], b.Values[j:])
-		k += n
-	}
-
-	a.Timestamps = out.Timestamps[:k]
-	a.Values = out.Values[:k]
-}
-{{ else }}
-// Exclude removes the subset of timestamps in [min, max]. The timestamps must
-// be deduplicated and sorted before calling Exclude or the results are undefined.
-func (a *{{ $typename }}) Exclude(min, max int64) {
-	rmin, rmax := a.FindRange(min, max)
-	if rmin == -1 && rmax == -1 {
-		return
-	}
-
-	// a.Timestamps[rmin] ≥ min
-	// a.Timestamps[rmax] ≥ max
-
-	if rmax < a.Len() {
-		if a.Timestamps[rmax] == max {
-			rmax++
-		}
-		rest := a.Len()-rmax
-		if rest > 0 {
-			ts := a.Timestamps[:rmin+rest]
-			copy(ts[rmin:], a.Timestamps[rmax:])
-			a.Timestamps = ts
-			return
-		}
-	}
-
-	a.Timestamps = a.Timestamps[:rmin]
-}
-
-// Contains returns true if values exist between min and max inclusive. The
-// values must be sorted before calling Contains or the results are undefined.
-func (a *{{ $typename }}) Contains(min, max int64) bool {
-	rmin, rmax := a.FindRange(min, max)
-	if rmin == -1 && rmax == -1 {
-		return false
-	}
-
-	// a.Timestamps[rmin] ≥ min
-	// a.Timestamps[rmax] ≥ max
-
-	if a.Timestamps[rmin] == min {
-		return true
-	}
-
-	if rmax < a.Len() && a.Timestamps[rmax] == max {
-		return true
-	}
-
-	return rmax-rmin > 0
-}
-{{ end }}
-
-{{ end }}

tsdb/cursors/arrayvalues.gen.go.tmpldata

@@ -1,26 +0,0 @@
-[
-	{
-		"Name":"Float",
-		"Type":"float64"
-	},
-	{
-		"Name":"Integer",
-		"Type":"int64"
-	},
-	{
-		"Name":"Unsigned",
-		"Type":"uint64"
-	},
-	{
-		"Name":"String",
-		"Type":"string"
-	},
-	{
-		"Name":"Boolean",
-		"Type":"bool"
-	},
-	{
-		"Name":"Timestamp",
-		"Type": null
-	}
-]

tsdb/cursors/arrayvalues.gen_test.go

@@ -1,254 +0,0 @@
-package cursors
-
-import (
-	"fmt"
-	"testing"
-
-	"github.com/google/go-cmp/cmp"
-)
-
-func makeIntegerArray(count int, min, max int64) *IntegerArray {
-	vals := NewIntegerArrayLen(count)
-
-	ts := min
-	inc := (max - min) / int64(count)
-
-	for i := 0; i < count; i++ {
-		vals.Timestamps[i] = ts
-		ts += inc
-	}
-
-	return vals
-}
-
-func makeIntegerArrayFromSlice(t []int64) *IntegerArray {
-	iv := NewIntegerArrayLen(len(t))
-	copy(iv.Timestamps, t)
-	return iv
-}
-
-func TestIntegerArray_FindRangeNoValues(t *testing.T) {
-	var vals IntegerArray
-	l, r := vals.FindRange(0, 100)
-	if exp := -1; l != exp {
-		t.Errorf("invalid l; exp=%d, got=%d", exp, l)
-	}
-	if exp := -1; r != exp {
-		t.Errorf("invalid r; exp=%d, got=%d", exp, r)
-	}
-}
-
-func TestIntegerArray_FindRange(t *testing.T) {
-	vals := makeIntegerArrayFromSlice([]int64{10, 11, 13, 15, 17, 20, 21})
-
-	cases := []struct {
-		min, max int64
-		l, r     int
-	}{
-		{12, 20, 2, 5},
-		{22, 40, -1, -1},
-		{1, 9, -1, -1},
-		{1, 10, 0, 0},
-		{1, 11, 0, 1},
-		{15, 15, 3, 3},
-	}
-
-	for _, tc := range cases {
-		t.Run(fmt.Sprintf("%d→%d", tc.min, tc.max), func(t *testing.T) {
-			l, r := vals.FindRange(tc.min, tc.max)
-			if l != tc.l {
-				t.Errorf("left: got %d, exp %d", l, tc.l)
-			}
-			if r != tc.r {
-				t.Errorf("right: got %d, exp %d", r, tc.r)
-			}
-		})
-	}
-}
-
-func TestIntegerArray_Exclude(t *testing.T) {
-	cases := []struct {
-		n        string
-		min, max int64
-		exp      []int64
-	}{
-		{"excl bad range", 18, 11, []int64{10, 12, 14, 16, 18}},
-		{"excl none-lo", 0, 9, []int64{10, 12, 14, 16, 18}},
-		{"excl none-hi", 19, 30, []int64{10, 12, 14, 16, 18}},
-		{"excl first", 0, 10, []int64{12, 14, 16, 18}},
-		{"excl last", 18, 20, []int64{10, 12, 14, 16}},
-		{"excl all but first and last", 12, 16, []int64{10, 18}},
-		{"excl none in middle", 13, 13, []int64{10, 12, 14, 16, 18}},
-		{"excl middle", 14, 14, []int64{10, 12, 16, 18}},
-		{"excl suffix", 14, 18, []int64{10, 12}},
-	}
-
-	for _, tc := range cases {
-		t.Run(fmt.Sprintf("%s[%d,%d]", tc.n, tc.min, tc.max), func(t *testing.T) {
-			vals := makeIntegerArray(5, 10, 20)
-			vals.Exclude(tc.min, tc.max)
-			got := vals.Timestamps
-			if !cmp.Equal(got, tc.exp) {
-				t.Errorf("unexpected values -got/+exp\n%s", cmp.Diff(got, tc.exp))
-			}
-		})
-	}
-}
-
-func TestIntegerArray_Include(t *testing.T) {
-	cases := []struct {
-		n        string
-		min, max int64
-		exp      []int64
-	}{
-		{"incl none-lo", 0, 9, []int64{}},
-		{"incl none-hi", 19, 30, []int64{}},
-		{"incl first", 0, 10, []int64{10}},
-		{"incl last", 18, 20, []int64{18}},
-		{"incl all but first and last", 12, 16, []int64{12, 14, 16}},
-		{"incl none in middle", 13, 13, []int64{}},
-		{"incl middle", 14, 14, []int64{14}},
-	}
-
-	for _, tc := range cases {
-		t.Run(fmt.Sprintf("%s[%d,%d]", tc.n, tc.min, tc.max), func(t *testing.T) {
-			vals := makeIntegerArray(5, 10, 20)
-			vals.Include(tc.min, tc.max)
-			got := vals.Timestamps
-			if !cmp.Equal(got, tc.exp) {
-				t.Errorf("unexpected values -got/+exp\n%s", cmp.Diff(got, tc.exp))
-			}
-		})
-	}
-}
-
-func makeTimestampArray(count int, min, max int64) *TimestampArray {
-	vals := NewTimestampArrayLen(count)
-
-	ts := min
-	inc := (max - min) / int64(count)
-
-	for i := 0; i < count; i++ {
-		vals.Timestamps[i] = ts
-		ts += inc
-	}
-
-	return vals
-}
-
-func TestTimestampArray_Contains(t *testing.T) {
-	cases := []struct {
-		n        string
-		min, max int64
-		exp      bool
-	}{
-		{"no/lo", 0, 9, false},
-		{"no/hi", 19, 30, false},
-		{"no/middle", 13, 13, false},
-
-		{"yes/first", 0, 10, true},
-		{"yes/first-eq", 10, 10, true},
-		{"yes/last", 18, 20, true},
-		{"yes/last-eq", 18, 18, true},
-		{"yes/all but first and last", 12, 16, true},
-		{"yes/middle-eq", 14, 14, true},
-		{"yes/middle-overlap", 13, 15, true},
-		{"yes/covers", 8, 22, true},
-	}
-
-	for _, tc := range cases {
-		t.Run(fmt.Sprintf("%s[%d,%d]", tc.n, tc.min, tc.max), func(t *testing.T) {
-			vals := makeTimestampArray(5, 10, 20)
-			if got := vals.Contains(tc.min, tc.max); got != tc.exp {
-				t.Errorf("Contains -got/+exp\n%s", cmp.Diff(got, tc.exp))
-			}
-		})
-	}
-}
-
-func benchExclude(b *testing.B, vals *IntegerArray, min, max int64) {
-	b.ResetTimer()
-
-	for i := 0; i < b.N; i++ {
-		vals.Exclude(min, max)
-	}
-}
-
-func BenchmarkIntegerArray_ExcludeNone_1000(b *testing.B) {
-	benchExclude(b, makeIntegerArray(1000, 1000, 2000), 0, 500)
-}
-
-func BenchmarkIntegerArray_ExcludeMiddleHalf_1000(b *testing.B) {
-	benchExclude(b, makeIntegerArray(1000, 1000, 2000), 1250, 1750)
-}
-
-func BenchmarkIntegerArray_ExcludeFirst_1000(b *testing.B) {
-	benchExclude(b, makeIntegerArray(1000, 1000, 2000), 0, 1000)
-}
-
-func BenchmarkIntegerArray_ExcludeLast_1000(b *testing.B) {
-	benchExclude(b, makeIntegerArray(1000, 1000, 2000), 1999, 2000)
-}
-
-func BenchmarkIntegerArray_ExcludeNone_10000(b *testing.B) {
-	benchExclude(b, makeIntegerArray(10000, 10000, 20000), 00, 5000)
-}
-
-func BenchmarkIntegerArray_ExcludeMiddleHalf_10000(b *testing.B) {
-	benchExclude(b, makeIntegerArray(10000, 10000, 20000), 12500, 17500)
-}
-
-func BenchmarkIntegerArray_ExcludeFirst_10000(b *testing.B) {
-	benchExclude(b, makeIntegerArray(10000, 10000, 20000), 0, 10000)
-}
-
-func BenchmarkIntegerArray_ExcludeLast_10000(b *testing.B) {
-	benchExclude(b, makeIntegerArray(10000, 10000, 20000), 19999, 20000)
-}
-
-func benchInclude(b *testing.B, vals *IntegerArray, min, max int64) {
-	src := *vals
-	tmp := NewIntegerArrayLen(vals.Len())
-	copy(tmp.Timestamps, vals.Timestamps)
-	copy(tmp.Values, vals.Values)
-	b.ResetTimer()
-
-	for i := 0; i < b.N; i++ {
-		vals.Include(min, max)
-		*vals = src
-		copy(vals.Timestamps, tmp.Timestamps)
-		copy(vals.Values, tmp.Values)
-	}
-}
-
-func BenchmarkIntegerArray_IncludeNone_1000(b *testing.B) {
-	benchInclude(b, makeIntegerArray(1000, 1000, 2000), 0, 500)
-}
-
-func BenchmarkIntegerArray_IncludeMiddleHalf_1000(b *testing.B) {
-	benchInclude(b, makeIntegerArray(1000, 1000, 2000), 1250, 1750)
-}
-
-func BenchmarkIntegerArray_IncludeFirst_1000(b *testing.B) {
-	benchInclude(b, makeIntegerArray(1000, 1000, 2000), 0, 1000)
-}
-
-func BenchmarkIntegerArray_IncludeLast_1000(b *testing.B) {
-	benchInclude(b, makeIntegerArray(1000, 1000, 2000), 1999, 2000)
-}
-
-func BenchmarkIntegerArray_IncludeNone_10000(b *testing.B) {
-	benchInclude(b, makeIntegerArray(10000, 10000, 20000), 00, 5000)
-}
-
-func BenchmarkIntegerArray_IncludeMiddleHalf_10000(b *testing.B) {
-	benchInclude(b, makeIntegerArray(10000, 10000, 20000), 12500, 17500)
-}
-
-func BenchmarkIntegerArray_IncludeFirst_10000(b *testing.B) {
-	benchInclude(b, makeIntegerArray(10000, 10000, 20000), 0, 10000)
-}
-
-func BenchmarkIntegerArray_IncludeLast_10000(b *testing.B) {
-	benchInclude(b, makeIntegerArray(10000, 10000, 20000), 19999, 20000)
-}

tsdb/cursors/arrayvalues.go

@@ -1,41 +0,0 @@
-package cursors
-
-import "sort"
-
-func (a *FloatArray) Size() int {
-	// size of timestamps + values
-	return len(a.Timestamps)*8 + len(a.Values)*8
-}
-
-func (a *IntegerArray) Size() int {
-	// size of timestamps + values
-	return len(a.Timestamps)*8 + len(a.Values)*8
-}
-
-func (a *UnsignedArray) Size() int {
-	// size of timestamps + values
-	return len(a.Timestamps)*8 + len(a.Values)*8
-}
-
-func (a *StringArray) Size() int {
-	sz := len(a.Timestamps) * 8
-	for _, s := range a.Values {
-		sz += len(s)
-	}
-	return sz
-}
-
-func (a *BooleanArray) Size() int {
-	// size of timestamps + values
-	return len(a.Timestamps)*8 + len(a.Values)
-}
-
-var _ sort.Interface = (*TimestampArray)(nil)
-
-func (a *TimestampArray) Less(i, j int) bool {
-	return a.Timestamps[i] < a.Timestamps[j]
-}
-
-func (a *TimestampArray) Swap(i, j int) {
-	a.Timestamps[i], a.Timestamps[j] = a.Timestamps[j], a.Timestamps[i]
-}

tsdb/cursors/arrayvalues_test.go

@@ -1,459 +0,0 @@
-package cursors_test
-
-import (
-	"strconv"
-	"testing"
-
-	"github.com/google/go-cmp/cmp"
-	"github.com/influxdata/influxdb/v2/tsdb/cursors"
-)
-
-func makeBooleanArray(v ...interface{}) *cursors.BooleanArray {
-	if len(v)&1 == 1 {
-		panic("invalid array length")
-	}
-	a := cursors.NewBooleanArrayLen(len(v) / 2)
-	for i := 0; i < len(v); i += 2 {
-		a.Timestamps[i/2] = int64(v[i].(int))
-		a.Values[i/2] = v[i+1].(bool)
-	}
-	return a
-}
-
-func makeFloatArray(v ...interface{}) *cursors.FloatArray {
-	if len(v)&1 == 1 {
-		panic("invalid array length")
-	}
-	a := cursors.NewFloatArrayLen(len(v) / 2)
-	for i := 0; i < len(v); i += 2 {
-		a.Timestamps[i/2] = int64(v[i].(int))
-		a.Values[i/2] = v[i+1].(float64)
-	}
-	return a
-}
-
-func makeIntegerArray(v ...interface{}) *cursors.IntegerArray {
-	if len(v)&1 == 1 {
-		panic("invalid array length")
-	}
-	a := cursors.NewIntegerArrayLen(len(v) / 2)
-	for i := 0; i < len(v); i += 2 {
-		a.Timestamps[i/2] = int64(v[i].(int))
-		a.Values[i/2] = int64(v[i+1].(int))
-	}
-	return a
-}
-
-func makeUnsignedArray(v ...interface{}) *cursors.UnsignedArray {
-	if len(v)&1 == 1 {
-		panic("invalid array length")
-	}
-	a := cursors.NewUnsignedArrayLen(len(v) / 2)
-	for i := 0; i < len(v); i += 2 {
-		a.Timestamps[i/2] = int64(v[i].(int))
-		a.Values[i/2] = uint64(v[i+1].(int))
-	}
-	return a
-}
-
-func makeStringArray(v ...interface{}) *cursors.StringArray {
-	if len(v)&1 == 1 {
-		panic("invalid array length")
-	}
-	a := cursors.NewStringArrayLen(len(v) / 2)
-	for i := 0; i < len(v); i += 2 {
-		a.Timestamps[i/2] = int64(v[i].(int))
-		a.Values[i/2] = strconv.Itoa(v[i+1].(int))
-	}
-	return a
-}
-
-func TestBooleanArray_Merge(t *testing.T) {
-	tests := []struct {
-		name      string
-		a, b, exp *cursors.BooleanArray
-	}{
-		{
-			name: "empty a",
-
-			a:   makeBooleanArray(),
-			b:   makeBooleanArray(1, true, 2, true),
-			exp: makeBooleanArray(1, true, 2, true),
-		},
-		{
-			name: "empty b",
-
-			a:   makeBooleanArray(1, true, 2, true),
-			b:   makeBooleanArray(),
-			exp: makeBooleanArray(1, true, 2, true),
-		},
-		{
-			name: "b replaces a",
-
-			a: makeBooleanArray(1, true),
-			b: makeBooleanArray(
-				0, false,
-				1, false, // overwrites a
-				2, false,
-				3, false,
-				4, false,
-			),
-			exp: makeBooleanArray(0, false, 1, false, 2, false, 3, false, 4, false),
-		},
-		{
-			name: "b replaces partial a",
-
-			a: makeBooleanArray(1, true, 2, true, 3, true, 4, true),
-			b: makeBooleanArray(
-				1, false, // overwrites a
-				2, false, // overwrites a
-			),
-			exp: makeBooleanArray(
-				1, false, // overwrites a
-				2, false, // overwrites a
-				3, true,
-				4, true,
-			),
-		},
-		{
-			name: "b replaces all a",
-
-			a:   makeBooleanArray(1, true, 2, true, 3, true, 4, true),
-			b:   makeBooleanArray(1, false, 2, false, 3, false, 4, false),
-			exp: makeBooleanArray(1, false, 2, false, 3, false, 4, false),
-		},
-		{
-			name: "b replaces a interleaved",
-			a:    makeBooleanArray(0, true, 1, true, 2, true, 3, true, 4, true),
-			b:    makeBooleanArray(0, false, 2, false, 4, false),
-			exp:  makeBooleanArray(0, false, 1, true, 2, false, 3, true, 4, false),
-		},
-		{
-			name: "b merges a interleaved",
-			a:    makeBooleanArray(0, true, 2, true, 4, true),
-			b:    makeBooleanArray(1, false, 3, false, 5, false),
-			exp:  makeBooleanArray(0, true, 1, false, 2, true, 3, false, 4, true, 5, false),
-		},
-	}
-
-	for _, test := range tests {
-		t.Run(test.name, func(t *testing.T) {
-			test.a.Merge(test.b)
-			if !cmp.Equal(test.a, test.exp) {
-				t.Fatalf("unexpected values -got/+exp\n%s", cmp.Diff(test.a, test.exp))
-			}
-		})
-	}
-}
-
-func TestFloatArray_Merge(t *testing.T) {
-	tests := []struct {
-		name      string
-		a, b, exp *cursors.FloatArray
-	}{
-		{
-			name: "empty a",
-
-			a:   makeFloatArray(),
-			b:   makeFloatArray(1, 1.1, 2, 2.1),
-			exp: makeFloatArray(1, 1.1, 2, 2.1),
-		},
-		{
-			name: "empty b",
-
-			a:   makeFloatArray(1, 1.0, 2, 2.0),
-			b:   makeFloatArray(),
-			exp: makeFloatArray(1, 1.0, 2, 2.0),
-		},
-		{
-			name: "b replaces a",
-
-			a: makeFloatArray(1, 1.0),
-			b: makeFloatArray(
-				0, 0.1,
-				1, 1.1, // overwrites a
-				2, 2.1,
-				3, 3.1,
-				4, 4.1,
-			),
-			exp: makeFloatArray(0, 0.1, 1, 1.1, 2, 2.1, 3, 3.1, 4, 4.1),
-		},
-		{
-			name: "b replaces partial a",
-
-			a: makeFloatArray(1, 1.0, 2, 2.0, 3, 3.0, 4, 4.0),
-			b: makeFloatArray(
-				1, 1.1, // overwrites a
-				2, 2.1, // overwrites a
-			),
-			exp: makeFloatArray(
-				1, 1.1, // overwrites a
-				2, 2.1, // overwrites a
-				3, 3.0,
-				4, 4.0,
-			),
-		},
-		{
-			name: "b replaces all a",
-
-			a:   makeFloatArray(1, 1.0, 2, 2.0, 3, 3.0, 4, 4.0),
-			b:   makeFloatArray(1, 1.1, 2, 2.1, 3, 3.1, 4, 4.1),
-			exp: makeFloatArray(1, 1.1, 2, 2.1, 3, 3.1, 4, 4.1),
-		},
-		{
-			name: "b replaces a interleaved",
-			a:    makeFloatArray(0, 0.0, 1, 1.0, 2, 2.0, 3, 3.0, 4, 4.0),
-			b:    makeFloatArray(0, 0.1, 2, 2.1, 4, 4.1),
-			exp:  makeFloatArray(0, 0.1, 1, 1.0, 2, 2.1, 3, 3.0, 4, 4.1),
-		},
-		{
-			name: "b merges a interleaved",
-			a:    makeFloatArray(0, 0.0, 2, 2.0, 4, 4.0),
-			b:    makeFloatArray(1, 1.1, 3, 3.1, 5, 5.1),
-			exp:  makeFloatArray(0, 0.0, 1, 1.1, 2, 2.0, 3, 3.1, 4, 4.0, 5, 5.1),
-		},
-	}
-
-	for _, test := range tests {
-		t.Run(test.name, func(t *testing.T) {
-			test.a.Merge(test.b)
-			if !cmp.Equal(test.a, test.exp) {
-				t.Fatalf("unexpected values -got/+exp\n%s", cmp.Diff(test.a, test.exp))
-			}
-		})
-	}
-}
-
-func TestIntegerArray_Merge(t *testing.T) {
-	tests := []struct {
-		name      string
-		a, b, exp *cursors.IntegerArray
-	}{
-		{
-			name: "empty a",
-
-			a:   makeIntegerArray(),
-			b:   makeIntegerArray(1, 11, 2, 21),
-			exp: makeIntegerArray(1, 11, 2, 21),
-		},
-		{
-			name: "empty b",
-
-			a:   makeIntegerArray(1, 10, 2, 20),
-			b:   makeIntegerArray(),
-			exp: makeIntegerArray(1, 10, 2, 20),
-		},
-		{
-			name: "b replaces a",
-
-			a: makeIntegerArray(1, 10),
-			b: makeIntegerArray(
-				0, 1,
-				1, 11, // overwrites a
-				2, 21,
-				3, 31,
-				4, 41,
-			),
-			exp: makeIntegerArray(0, 1, 1, 11, 2, 21, 3, 31, 4, 41),
-		},
-		{
-			name: "b replaces partial a",
-
-			a: makeIntegerArray(1, 10, 2, 20, 3, 30, 4, 40),
-			b: makeIntegerArray(
-				1, 11, // overwrites a
-				2, 21, // overwrites a
-			),
-			exp: makeIntegerArray(
-				1, 11, // overwrites a
-				2, 21, // overwrites a
-				3, 30,
-				4, 40,
-			),
-		},
-		{
-			name: "b replaces all a",
-
-			a:   makeIntegerArray(1, 10, 2, 20, 3, 30, 4, 40),
-			b:   makeIntegerArray(1, 11, 2, 21, 3, 31, 4, 41),
-			exp: makeIntegerArray(1, 11, 2, 21, 3, 31, 4, 41),
-		},
-		{
-			name: "b replaces a interleaved",
-			a:    makeIntegerArray(0, 0, 1, 10, 2, 20, 3, 30, 4, 40),
-			b:    makeIntegerArray(0, 1, 2, 21, 4, 41),
-			exp:  makeIntegerArray(0, 1, 1, 10, 2, 21, 3, 30, 4, 41),
-		},
-		{
-			name: "b merges a interleaved",
-			a:    makeIntegerArray(0, 00, 2, 20, 4, 40),
-			b:    makeIntegerArray(1, 11, 3, 31, 5, 51),
-			exp:  makeIntegerArray(0, 00, 1, 11, 2, 20, 3, 31, 4, 40, 5, 51),
-		},
-	}
-
-	for _, test := range tests {
-		t.Run(test.name, func(t *testing.T) {
-			test.a.Merge(test.b)
-			if !cmp.Equal(test.a, test.exp) {
-				t.Fatalf("unexpected values -got/+exp\n%s", cmp.Diff(test.a, test.exp))
-			}
-		})
-	}
-}
-
-func TestUnsignedArray_Merge(t *testing.T) {
-	tests := []struct {
-		name      string
-		a, b, exp *cursors.UnsignedArray
-	}{
-		{
-			name: "empty a",
-
-			a:   makeUnsignedArray(),
-			b:   makeUnsignedArray(1, 11, 2, 21),
-			exp: makeUnsignedArray(1, 11, 2, 21),
-		},
-		{
-			name: "empty b",
-
-			a:   makeUnsignedArray(1, 10, 2, 20),
-			b:   makeUnsignedArray(),
-			exp: makeUnsignedArray(1, 10, 2, 20),
-		},
-		{
-			name: "b replaces a",
-
-			a: makeUnsignedArray(1, 10),
-			b: makeUnsignedArray(
-				0, 1,
-				1, 11, // overwrites a
-				2, 21,
-				3, 31,
-				4, 41,
-			),
-			exp: makeUnsignedArray(0, 1, 1, 11, 2, 21, 3, 31, 4, 41),
-		},
-		{
-			name: "b replaces partial a",
-
-			a: makeUnsignedArray(1, 10, 2, 20, 3, 30, 4, 40),
-			b: makeUnsignedArray(
-				1, 11, // overwrites a
-				2, 21, // overwrites a
-			),
-			exp: makeUnsignedArray(
-				1, 11, // overwrites a
-				2, 21, // overwrites a
-				3, 30,
-				4, 40,
-			),
-		},
-		{
-			name: "b replaces all a",
-
-			a:   makeUnsignedArray(1, 10, 2, 20, 3, 30, 4, 40),
-			b:   makeUnsignedArray(1, 11, 2, 21, 3, 31, 4, 41),
-			exp: makeUnsignedArray(1, 11, 2, 21, 3, 31, 4, 41),
-		},
-		{
-			name: "b replaces a interleaved",
-			a:    makeUnsignedArray(0, 0, 1, 10, 2, 20, 3, 30, 4, 40),
-			b:    makeUnsignedArray(0, 1, 2, 21, 4, 41),
-			exp:  makeUnsignedArray(0, 1, 1, 10, 2, 21, 3, 30, 4, 41),
-		},
-		{
-			name: "b merges a interleaved",
-			a:    makeUnsignedArray(0, 00, 2, 20, 4, 40),
-			b:    makeUnsignedArray(1, 11, 3, 31, 5, 51),
-			exp:  makeUnsignedArray(0, 00, 1, 11, 2, 20, 3, 31, 4, 40, 5, 51),
-		},
-	}
-
-	for _, test := range tests {
-		t.Run(test.name, func(t *testing.T) {
-			test.a.Merge(test.b)
-			if !cmp.Equal(test.a, test.exp) {
-				t.Fatalf("unexpected values -got/+exp\n%s", cmp.Diff(test.a, test.exp))
-			}
-		})
-	}
-}
-
-func TestStringArray_Merge(t *testing.T) {
-	tests := []struct {
-		name      string
-		a, b, exp *cursors.StringArray
-	}{
-		{
-			name: "empty a",
-
-			a:   makeStringArray(),
-			b:   makeStringArray(1, 11, 2, 21),
-			exp: makeStringArray(1, 11, 2, 21),
-		},
-		{
-			name: "empty b",
-
-			a:   makeStringArray(1, 10, 2, 20),
-			b:   makeStringArray(),
-			exp: makeStringArray(1, 10, 2, 20),
-		},
-		{
-			name: "b replaces a",
-
-			a: makeStringArray(1, 10),
-			b: makeStringArray(
-				0, 1,
-				1, 11, // overwrites a
-				2, 21,
-				3, 31,
-				4, 41,
-			),
-			exp: makeStringArray(0, 1, 1, 11, 2, 21, 3, 31, 4, 41),
-		},
-		{
-			name: "b replaces partial a",
-
-			a: makeStringArray(1, 10, 2, 20, 3, 30, 4, 40),
-			b: makeStringArray(
-				1, 11, // overwrites a
-				2, 21, // overwrites a
-			),
-			exp: makeStringArray(
-				1, 11, // overwrites a
-				2, 21, // overwrites a
-				3, 30,
-				4, 40,
-			),
-		},
-		{
-			name: "b replaces all a",
-
-			a:   makeStringArray(1, 10, 2, 20, 3, 30, 4, 40),
-			b:   makeStringArray(1, 11, 2, 21, 3, 31, 4, 41),
-			exp: makeStringArray(1, 11, 2, 21, 3, 31, 4, 41),
-		},
-		{
-			name: "b replaces a interleaved",
-			a:    makeStringArray(0, 0, 1, 10, 2, 20, 3, 30, 4, 40),
-			b:    makeStringArray(0, 1, 2, 21, 4, 41),
-			exp:  makeStringArray(0, 1, 1, 10, 2, 21, 3, 30, 4, 41),
-		},
-		{
-			name: "b merges a interleaved",
-			a:    makeStringArray(0, 00, 2, 20, 4, 40),
-			b:    makeStringArray(1, 11, 3, 31, 5, 51),
-			exp:  makeStringArray(0, 00, 1, 11, 2, 20, 3, 31, 4, 40, 5, 51),
-		},
-	}
-
-	for _, test := range tests {
-		t.Run(test.name, func(t *testing.T) {
-			test.a.Merge(test.b)
-			if !cmp.Equal(test.a, test.exp) {
-				t.Fatalf("unexpected values -got/+exp\n%s", cmp.Diff(test.a, test.exp))
-			}
-		})
-	}
-}

tsdb/cursors/cursor.go

@@ -1,68 +0,0 @@
-package cursors
-
-import (
-	"context"
-
-	"github.com/influxdata/influxdb/v2/models"
-)
-
-const DefaultMaxPointsPerBlock = 1000
-
-type Cursor interface {
-	Close()
-	Err() error
-	Stats() CursorStats
-}
-
-type IntegerArrayCursor interface {
-	Cursor
-	Next() *IntegerArray
-}
-
-type FloatArrayCursor interface {
-	Cursor
-	Next() *FloatArray
-}
-
-type UnsignedArrayCursor interface {
-	Cursor
-	Next() *UnsignedArray
-}
-
-type StringArrayCursor interface {
-	Cursor
-	Next() *StringArray
-}
-
-type BooleanArrayCursor interface {
-	Cursor
-	Next() *BooleanArray
-}
-
-type CursorRequest struct {
-	Name      []byte
-	Tags      models.Tags
-	Field     string
-	Ascending bool
-	StartTime int64
-	EndTime   int64
-}
-
-type CursorIterator interface {
-	Next(ctx context.Context, r *CursorRequest) (Cursor, error)
-	Stats() CursorStats
-}
-
-type CursorIterators []CursorIterator
-
-// CursorStats represents stats collected by a cursor.
-type CursorStats struct {
-	ScannedValues int // number of values scanned
-	ScannedBytes  int // number of uncompressed bytes scanned
-}
-
-// Add adds other to s and updates s.
-func (s *CursorStats) Add(other CursorStats) {
-	s.ScannedValues += other.ScannedValues
-	s.ScannedBytes += other.ScannedBytes
-}

tsdb/cursors/gen.go

@@ -1,4 +0,0 @@
-package cursors
-
-//go:generate env GO111MODULE=on go run github.com/benbjohnson/tmpl -data=@arrayvalues.gen.go.tmpldata arrayvalues.gen.go.tmpl
-//go:generate stringer -type FieldType

tsdb/cursors/string.go

@@ -1,81 +0,0 @@
-package cursors
-
-// StringIterator describes the behavior for enumerating a sequence of
-// string values.
-type StringIterator interface {
-	// Next advances the StringIterator to the next value. It returns false
-	// when there are no more values.
-	Next() bool
-
-	// Value returns the current value.
-	Value() string
-
-	Stats() CursorStats
-}
-
-// EmptyStringIterator is an implementation of StringIterator that returns
-// no values.
-var EmptyStringIterator StringIterator = &stringIterator{}
-
-type stringIterator struct{}
-
-func (*stringIterator) Next() bool         { return false }
-func (*stringIterator) Value() string      { return "" }
-func (*stringIterator) Stats() CursorStats { return CursorStats{} }
-
-type StringSliceIterator struct {
-	s     []string
-	v     string
-	i     int
-	stats CursorStats
-}
-
-func NewStringSliceIterator(s []string) *StringSliceIterator {
-	return &StringSliceIterator{s: s, i: 0}
-}
-
-func NewStringSliceIteratorWithStats(s []string, stats CursorStats) *StringSliceIterator {
-	return &StringSliceIterator{s: s, i: 0, stats: stats}
-}
-
-func (s *StringSliceIterator) Next() bool {
-	if s.i < len(s.s) {
-		s.v = s.s[s.i]
-		s.i++
-		return true
-	}
-	s.v = ""
-	return false
-}
-
-func (s *StringSliceIterator) Value() string {
-	return s.v
-}
-
-func (s *StringSliceIterator) Stats() CursorStats {
-	return s.stats
-}
-
-func (s *StringSliceIterator) toSlice() []string {
-	if s.i < len(s.s) {
-		return s.s[s.i:]
-	}
-	return nil
-}
-
-// StringIteratorToSlice reads the remainder of i into a slice and
-// returns the result.
-func StringIteratorToSlice(i StringIterator) []string {
-	if i == nil {
-		return nil
-	}
-
-	if si, ok := i.(*StringSliceIterator); ok {
-		return si.toSlice()
-	}
-	var a []string
-	for i.Next() {
-		a = append(a, i.Value())
-	}
-	return a
-}

tsdb/errors.go

@@ -1,19 +0,0 @@
-package tsdb
-
-import (
-	"fmt"
-)
-
-// PartialWriteError indicates a write request could only write a portion of the
-// requested values.
-type PartialWriteError struct {
-	Reason  string
-	Dropped int
-
-	// A sorted slice of series keys that were dropped.
-	DroppedKeys [][]byte
-}
-
-func (e PartialWriteError) Error() string {
-	return fmt.Sprintf("partial write: %s dropped=%d", e.Reason, e.Dropped)
-}

tsdb/explode.go

@@ -1,106 +0,0 @@
-package tsdb
-
-import (
-	"encoding/binary"
-
-	"github.com/influxdata/influxdb/v2"
-	"github.com/influxdata/influxdb/v2/models"
-)
-
-// DecodeName converts tsdb internal serialization back to organization and bucket IDs.
-func DecodeName(name [16]byte) (org, bucket influxdb.ID) {
-	org = influxdb.ID(binary.BigEndian.Uint64(name[0:8]))
-	bucket = influxdb.ID(binary.BigEndian.Uint64(name[8:16]))
-	return
-}
-
-// DecodeNameSlice converts tsdb internal serialization back to organization and bucket IDs.
-func DecodeNameSlice(name []byte) (org, bucket influxdb.ID) {
-	return influxdb.ID(binary.BigEndian.Uint64(name[0:8])), influxdb.ID(binary.BigEndian.Uint64(name[8:16]))
-}
-
-// EncodeName converts org/bucket pairs to the tsdb internal serialization
-func EncodeName(org, bucket influxdb.ID) [16]byte {
-	var nameBytes [16]byte
-	binary.BigEndian.PutUint64(nameBytes[0:8], uint64(org))
-	binary.BigEndian.PutUint64(nameBytes[8:16], uint64(bucket))
-	return nameBytes
-}
-
-// EncodeNameSlice converts org/bucket pairs to the tsdb internal serialization but returns a byte slice.
-func EncodeNameSlice(org, bucket influxdb.ID) []byte {
-	buf := EncodeName(org, bucket)
-	return buf[:]
-}
-
-// EncodeOrgName converts org to the tsdb internal serialization that may be used
-// as a prefix when searching for keys matching a specific organization.
-func EncodeOrgName(org influxdb.ID) [8]byte {
-	var orgBytes [8]byte
-	binary.BigEndian.PutUint64(orgBytes[0:8], uint64(org))
-	return orgBytes
-}
-
-// EncodeNameString converts org/bucket pairs to the tsdb internal serialization
-func EncodeNameString(org, bucket influxdb.ID) string {
-	name := EncodeName(org, bucket)
-	return string(name[:])
-}
-
-// ExplodePoints creates a list of points that only contains one field per point. It also
-// moves the measurement to a tag, and changes the measurement to be the provided argument.
-func ExplodePoints(org, bucket influxdb.ID, points []models.Point) ([]models.Point, error) {
-	out := make([]models.Point, 0, len(points))
-
-	// TODO(jeff): We should add a RawEncode() method or something to the influxdb.ID type
-	// or we should use hex encoded measurement names. Either way, we shouldn't be doing a
-	// decode of the encode here, and we don't want to depend on details of how the ID type
-	// is represented.
-	ob := EncodeName(org, bucket)
-	name := string(ob[:])
-
-	tags := make(models.Tags, 1)
-	for _, pt := range points {
-		tags = tags[:1] // reset buffer for next point.
-
-		tags[0] = models.NewTag(models.MeasurementTagKeyBytes, pt.Name())
-		pt.ForEachTag(func(k, v []byte) bool {
-			tags = append(tags, models.NewTag(k, v))
-			return true
-		})
-
-		t := pt.Time()
-		itr := pt.FieldIterator()
-		tags = append(tags, models.Tag{}) // Make room for field key and value.
-
-		for itr.Next() {
-			tags[len(tags)-1] = models.NewTag(models.FieldKeyTagKeyBytes, itr.FieldKey())
-
-			var err error
-			field := make(models.Fields, 1)
-			switch itr.Type() {
-			case models.Float:
-				field[string(itr.FieldKey())], err = itr.FloatValue()
-			case models.Integer:
-				field[string(itr.FieldKey())], err = itr.IntegerValue()
-			case models.Boolean:
-				field[string(itr.FieldKey())], err = itr.BooleanValue()
-			case models.String:
-				field[string(itr.FieldKey())] = itr.StringValue()
-			case models.Unsigned:
-				field[string(itr.FieldKey())], err = itr.UnsignedValue()
-			}
-			if err != nil {
-				return nil, err
-			}
-
-			pt, err := models.NewPoint(name, tags, field, t)
-			if err != nil {
-				return nil, err
-			}
-			out = append(out, pt)
-		}
-	}
-
-	return out, nil
-}

tsdb/explode_test.go

@@ -1,42 +0,0 @@
-package tsdb_test
-
-import (
-	"fmt"
-	"testing"
-
-	"github.com/influxdata/influxdb/v2"
-	"github.com/influxdata/influxdb/v2/tsdb"
-)
-
-func TestNames(t *testing.T) {
-	goodExamples := []struct {
-		Org    uint64
-		Bucket uint64
-		Name   [16]byte
-	}{
-		{Org: 12345678, Bucket: 87654321, Name: [16]byte{0, 0, 0, 0, 0, 188, 97, 78, 0, 0, 0, 0, 5, 57, 127, 177}},
-		{Org: 1234567891011, Bucket: 87654321, Name: [16]byte{0, 0, 1, 31, 113, 251, 8, 67, 0, 0, 0, 0, 5, 57, 127, 177}},
-		{Org: 12345678, Bucket: 8765432100000, Name: [16]byte{0, 0, 0, 0, 0, 188, 97, 78, 0, 0, 7, 248, 220, 119, 116, 160}},
-		{Org: 123456789929, Bucket: 8765432100000, Name: [16]byte{0, 0, 0, 28, 190, 153, 29, 169, 0, 0, 7, 248, 220, 119, 116, 160}},
-	}
-
-	for _, example := range goodExamples {
-		t.Run(fmt.Sprintf("%d%d", example.Org, example.Bucket), func(t *testing.T) {
-
-			name := tsdb.EncodeName(influxdb.ID(example.Org), influxdb.ID(example.Bucket))
-
-			if got, exp := name, example.Name; got != exp {
-				t.Errorf("got name %q, expected %q", got, exp)
-			}
-
-			org, bucket := tsdb.DecodeName(name)
-
-			if gotOrg, expOrg := org, example.Org; gotOrg != influxdb.ID(expOrg) {
-				t.Errorf("got organization ID %q, expected %q", gotOrg, expOrg)
-			}
-			if gotBucket, expBucket := bucket, example.Bucket; gotBucket != influxdb.ID(expBucket) {
-				t.Errorf("got organization ID %q, expected %q", gotBucket, expBucket)
-			}
-		})
-	}
-}

tsdb/gen_test.go

@@ -1,14 +0,0 @@
-//go:generate sh -c "curl -L https://github.com/influxdata/testdata/raw/2020.07.20.1/tsdbtestdata.tar.gz | tar xz"
-package tsdb_test
-
-import (
-	"fmt"
-	"os"
-)
-
-func init() {
-	if _, err := os.Stat("./testdata"); err != nil {
-		fmt.Println("Run go generate to download testdata directory.")
-		os.Exit(1)
-	}
-}

tsdb/meta.go

@@ -1,59 +0,0 @@
-package tsdb
-
-import (
-	"github.com/influxdata/influxdb/v2/models"
-)
-
-// MakeTagsKey converts a tag set to bytes for use as a lookup key.
-func MakeTagsKey(keys []string, tags models.Tags) []byte {
-	// precondition: keys is sorted
-	// precondition: models.Tags is sorted
-
-	// Empty maps marshal to empty bytes.
-	if len(keys) == 0 || len(tags) == 0 {
-		return nil
-	}
-
-	sel := make([]int, 0, len(keys))
-
-	sz := 0
-	i, j := 0, 0
-	for i < len(keys) && j < len(tags) {
-		if keys[i] < string(tags[j].Key) {
-			i++
-		} else if keys[i] > string(tags[j].Key) {
-			j++
-		} else {
-			sel = append(sel, j)
-			sz += len(keys[i]) + len(tags[j].Value)
-			i++
-			j++
-		}
-	}
-
-	if len(sel) == 0 {
-		// no tags matched the requested keys
-		return nil
-	}
-
-	sz += (len(sel) * 2) - 1 // selected tags, add separators
-
-	// Generate marshaled bytes.
-	b := make([]byte, sz)
-	buf := b
-	for _, k := range sel {
-		copy(buf, tags[k].Key)
-		buf[len(tags[k].Key)] = '|'
-		buf = buf[len(tags[k].Key)+1:]
-	}
-
-	for i, k := range sel {
-		copy(buf, tags[k].Value)
-		if i < len(sel)-1 {
-			buf[len(tags[k].Value)] = '|'
-			buf = buf[len(tags[k].Value)+1:]
-		}
-	}
-
-	return b
-}

tsdb/meta_test.go

@@ -1,89 +0,0 @@
-package tsdb_test
-
-import (
-	"bytes"
-	"fmt"
-	"testing"
-
-	"github.com/influxdata/influxdb/v2/models"
-	"github.com/influxdata/influxdb/v2/tsdb"
-)
-
-// Ensure tags can be marshaled into a byte slice.
-func TestMakeTagsKey(t *testing.T) {
-	for i, tt := range []struct {
-		keys   []string
-		tags   models.Tags
-		result []byte
-	}{
-		{
-			keys:   nil,
-			tags:   nil,
-			result: nil,
-		},
-		{
-			keys:   []string{"foo"},
-			tags:   models.NewTags(map[string]string{"foo": "bar"}),
-			result: []byte(`foo|bar`),
-		},
-		{
-			keys:   []string{"foo"},
-			tags:   models.NewTags(map[string]string{"baz": "battttt"}),
-			result: []byte(``),
-		},
-		{
-			keys:   []string{"baz", "foo"},
-			tags:   models.NewTags(map[string]string{"baz": "battttt"}),
-			result: []byte(`baz|battttt`),
-		},
-		{
-			keys:   []string{"baz", "foo", "zzz"},
-			tags:   models.NewTags(map[string]string{"foo": "bar"}),
-			result: []byte(`foo|bar`),
-		},
-		{
-			keys:   []string{"baz", "foo"},
-			tags:   models.NewTags(map[string]string{"foo": "bar", "baz": "battttt"}),
-			result: []byte(`baz|foo|battttt|bar`),
-		},
-		{
-			keys:   []string{"baz"},
-			tags:   models.NewTags(map[string]string{"baz": "battttt", "foo": "bar"}),
-			result: []byte(`baz|battttt`),
-		},
-	} {
-		result := tsdb.MakeTagsKey(tt.keys, tt.tags)
-		if !bytes.Equal(result, tt.result) {
-			t.Fatalf("%d. unexpected result: exp=%s, got=%s", i, tt.result, result)
-		}
-	}
-}
-
-func BenchmarkMakeTagsKey_KeyN1(b *testing.B)  { benchmarkMakeTagsKey(b, 1) }
-func BenchmarkMakeTagsKey_KeyN3(b *testing.B)  { benchmarkMakeTagsKey(b, 3) }
-func BenchmarkMakeTagsKey_KeyN5(b *testing.B)  { benchmarkMakeTagsKey(b, 5) }
-func BenchmarkMakeTagsKey_KeyN10(b *testing.B) { benchmarkMakeTagsKey(b, 10) }
-
-func makeTagsAndKeys(keyN int) ([]string, models.Tags) {
-	const keySize, valueSize = 8, 15
-
-	// Generate tag map.
-	keys := make([]string, keyN)
-	tags := make(map[string]string)
-	for i := 0; i < keyN; i++ {
-		keys[i] = fmt.Sprintf("%0*d", keySize, i)
-		tags[keys[i]] = fmt.Sprintf("%0*d", valueSize, i)
-	}
-
-	return keys, models.NewTags(tags)
-}
-
-func benchmarkMakeTagsKey(b *testing.B, keyN int) {
-	keys, tags := makeTagsAndKeys(keyN)
-
-	// Unmarshal map into byte slice.
-	b.ReportAllocs()
-	for i := 0; i < b.N; i++ {
-		tsdb.MakeTagsKey(keys, tags)
-	}
-}

tsdb/series_collection.go

@@ -1,333 +0,0 @@
-package tsdb
-
-import (
-	"sync"
-	"sync/atomic"
-	"unsafe"
-
-	"github.com/influxdata/influxdb/v2/models"
-	"github.com/influxdata/influxdb/v2/pkg/bytesutil"
-)
-
-// SeriesCollection is a struct of arrays representation of a collection of series that allows
-// for efficient filtering.
-type SeriesCollection struct {
-	Points     []models.Point
-	Keys       [][]byte
-	SeriesKeys [][]byte
-	Names      [][]byte
-	Tags       []models.Tags
-	Types      []models.FieldType
-	SeriesIDs  []SeriesID
-
-	// Keeps track of invalid entries.
-	Dropped     uint64
-	DroppedKeys [][]byte
-	Reason      string
-
-	// Used by the concurrent iterators to stage drops. Inefficient, but should be
-	// very infrequently used.
-	state *seriesCollectionState
-}
-
-// seriesCollectionState keeps track of concurrent iterator state.
-type seriesCollectionState struct {
-	mu     sync.Mutex
-	reason string
-	index  map[int]struct{}
-}
-
-// NewSeriesCollection builds a SeriesCollection from a slice of points. It does some filtering
-// of invalid points.
-func NewSeriesCollection(points []models.Point) *SeriesCollection {
-	out := &SeriesCollection{
-		Points: append([]models.Point(nil), points...),
-		Keys:   make([][]byte, 0, len(points)),
-		Names:  make([][]byte, 0, len(points)),
-		Tags:   make([]models.Tags, 0, len(points)),
-		Types:  make([]models.FieldType, 0, len(points)),
-	}
-
-	for _, pt := range points {
-		out.Keys = append(out.Keys, pt.Key())
-		out.Names = append(out.Names, pt.Name())
-		out.Tags = append(out.Tags, pt.Tags())
-
-		fi := pt.FieldIterator()
-		fi.Next()
-		out.Types = append(out.Types, fi.Type())
-	}
-
-	return out
-}
-
-// Duplicate returns a copy of the SeriesCollection. The slices are shared. Appending to any of
-// them may or may not be reflected.
-func (s SeriesCollection) Duplicate() *SeriesCollection { return &s }
-
-// Length returns the length of the first non-nil slice in the collection, or 0 if there is no
-// non-nil slice.
-func (s *SeriesCollection) Length() int {
-	switch {
-	case s.Points != nil:
-		return len(s.Points)
-	case s.Keys != nil:
-		return len(s.Keys)
-	case s.SeriesKeys != nil:
-		return len(s.SeriesKeys)
-	case s.Names != nil:
-		return len(s.Names)
-	case s.Tags != nil:
-		return len(s.Tags)
-	case s.Types != nil:
-		return len(s.Types)
-	case s.SeriesIDs != nil:
-		return len(s.SeriesIDs)
-	default:
-		return 0
-	}
-}
-
-// Copy will copy the element at src into dst in all slices that can: x[dst] = x[src].
-func (s *SeriesCollection) Copy(dst, src int) {
-	if dst == src {
-		return
-	}
-	udst, usrc := uint(dst), uint(src)
-	if n := uint(len(s.Points)); udst < n && usrc < n {
-		s.Points[udst] = s.Points[usrc]
-	}
-	if n := uint(len(s.Keys)); udst < n && usrc < n {
-		s.Keys[udst] = s.Keys[usrc]
-	}
-	if n := uint(len(s.SeriesKeys)); udst < n && usrc < n {
-		s.SeriesKeys[udst] = s.SeriesKeys[usrc]
-	}
-	if n := uint(len(s.Names)); udst < n && usrc < n {
-		s.Names[udst] = s.Names[usrc]
-	}
-	if n := uint(len(s.Tags)); udst < n && usrc < n {
-		s.Tags[udst] = s.Tags[usrc]
-	}
-	if n := uint(len(s.Types)); udst < n && usrc < n {
-		s.Types[udst] = s.Types[usrc]
-	}
-	if n := uint(len(s.SeriesIDs)); udst < n && usrc < n {
-		s.SeriesIDs[udst] = s.SeriesIDs[usrc]
-	}
-}
-
-// Swap will swap the elements at i and j in all slices that can: x[i], x[j] = x[j], x[i].
-func (s *SeriesCollection) Swap(i, j int) {
-	if i == j {
-		return
-	}
-	ui, uj := uint(i), uint(j)
-	if n := uint(len(s.Points)); ui < n && uj < n {
-		s.Points[ui], s.Points[uj] = s.Points[uj], s.Points[ui]
-	}
-	if n := uint(len(s.Keys)); ui < n && uj < n {
-		s.Keys[ui], s.Keys[uj] = s.Keys[uj], s.Keys[ui]
-	}
-	if n := uint(len(s.SeriesKeys)); ui < n && uj < n {
-		s.SeriesKeys[ui], s.SeriesKeys[uj] = s.SeriesKeys[uj], s.SeriesKeys[ui]
-	}
-	if n := uint(len(s.Names)); ui < n && uj < n {
-		s.Names[ui], s.Names[uj] = s.Names[uj], s.Names[ui]
-	}
-	if n := uint(len(s.Tags)); ui < n && uj < n {
-		s.Tags[ui], s.Tags[uj] = s.Tags[uj], s.Tags[ui]
-	}
-	if n := uint(len(s.Types)); ui < n && uj < n {
-		s.Types[ui], s.Types[uj] = s.Types[uj], s.Types[ui]
-	}
-	if n := uint(len(s.SeriesIDs)); ui < n && uj < n {
-		s.SeriesIDs[ui], s.SeriesIDs[uj] = s.SeriesIDs[uj], s.SeriesIDs[ui]
-	}
-}
-
-// Truncate will truncate all of the slices that can down to length: x = x[:length].
-func (s *SeriesCollection) Truncate(length int) {
-	ulength := uint(length)
-	if ulength < uint(len(s.Points)) {
-		s.Points = s.Points[:ulength]
-	}
-	if ulength < uint(len(s.Keys)) {
-		s.Keys = s.Keys[:ulength]
-	}
-	if ulength < uint(len(s.SeriesKeys)) {
-		s.SeriesKeys = s.SeriesKeys[:ulength]
-	}
-	if ulength < uint(len(s.Names)) {
-		s.Names = s.Names[:ulength]
-	}
-	if ulength < uint(len(s.Tags)) {
-		s.Tags = s.Tags[:ulength]
-	}
-	if ulength < uint(len(s.Types)) {
-		s.Types = s.Types[:ulength]
-	}
-	if ulength < uint(len(s.SeriesIDs)) {
-		s.SeriesIDs = s.SeriesIDs[:ulength]
-	}
-}
-
-// Advance will advance all of the slices that can length elements: x = x[length:].
-func (s *SeriesCollection) Advance(length int) {
-	ulength := uint(length)
-	if ulength < uint(len(s.Points)) {
-		s.Points = s.Points[ulength:]
-	}
-	if ulength < uint(len(s.Keys)) {
-		s.Keys = s.Keys[ulength:]
-	}
-	if ulength < uint(len(s.SeriesKeys)) {
-		s.SeriesKeys = s.SeriesKeys[ulength:]
-	}
-	if ulength < uint(len(s.Names)) {
-		s.Names = s.Names[ulength:]
-	}
-	if ulength < uint(len(s.Tags)) {
-		s.Tags = s.Tags[ulength:]
-	}
-	if ulength < uint(len(s.Types)) {
-		s.Types = s.Types[ulength:]
-	}
-	if ulength < uint(len(s.SeriesIDs)) {
-		s.SeriesIDs = s.SeriesIDs[ulength:]
-	}
-}
-
-// InvalidateAll causes all of the entries to become invalid.
-func (s *SeriesCollection) InvalidateAll(reason string) {
-	if s.Reason == "" {
-		s.Reason = reason
-	}
-	s.Dropped += uint64(len(s.Keys))
-	s.DroppedKeys = append(s.DroppedKeys, s.Keys...)
-	s.Truncate(0)
-}
-
-// ApplyConcurrentDrops will remove all of the dropped values during concurrent iteration. It should
-// not be called concurrently with any calls to Invalid.
-func (s *SeriesCollection) ApplyConcurrentDrops() {
-	state := s.getState(false)
-	if state == nil {
-		return
-	}
-
-	length, j := s.Length(), 0
-	for i := 0; i < length; i++ {
-		if _, ok := state.index[i]; ok {
-			s.Dropped++
-
-			if i < len(s.Keys) {
-				s.DroppedKeys = append(s.DroppedKeys, s.Keys[i])
-			}
-
-			continue
-		}
-
-		s.Copy(j, i)
-		j++
-	}
-	s.Truncate(j)
-
-	if s.Reason == "" {
-		s.Reason = state.reason
-	}
-
-	// clear concurrent state
-	atomic.StorePointer((*unsafe.Pointer)(unsafe.Pointer(&s.state)), nil)
-}
-
-// getState returns the SeriesCollection's concurrent state. If alloc is true and there
-// is no state, it will attempt to allocate one and set it. It is safe to call concurrently, but
-// not with ApplyConcurrentDrops.
-func (s *SeriesCollection) getState(alloc bool) *seriesCollectionState {
-	addr := (*unsafe.Pointer)(unsafe.Pointer(&s.state))
-
-	// fast path: load pointer and it already exists. always return the result if we can't alloc.
-	if ptr := atomic.LoadPointer(addr); ptr != nil || !alloc {
-		return (*seriesCollectionState)(ptr)
-	}
-
-	// nothing there. make a new state and try to swap it in.
-	atomic.CompareAndSwapPointer(addr, nil, unsafe.Pointer(new(seriesCollectionState)))
-
-	// reload the pointer. this way we always end up with the winner of the race.
-	return (*seriesCollectionState)(atomic.LoadPointer(addr))
-}
-
-// invalidIndex stages the index as invalid with the reason. It will be removed when
-// ApplyConcurrentDrops is called.
-func (s *SeriesCollection) invalidIndex(index int, reason string) {
-	state := s.getState(true)
-
-	state.mu.Lock()
-	if state.index == nil {
-		state.index = make(map[int]struct{})
-	}
-	state.index[index] = struct{}{}
-	if state.reason == "" {
-		state.reason = reason
-	}
-	state.mu.Unlock()
-}
-
-// PartialWriteError returns a PartialWriteError if any entries have been marked as invalid. It
-// returns an error to avoid `return collection.PartialWriteError()` always being non-nil.
-func (s *SeriesCollection) PartialWriteError() error {
-	if s.Dropped == 0 {
-		return nil
-	}
-	droppedKeys := bytesutil.SortDedup(s.DroppedKeys)
-	return PartialWriteError{
-		Reason:      s.Reason,
-		Dropped:     len(droppedKeys),
-		DroppedKeys: droppedKeys,
-	}
-}
-
-// Iterator returns a new iterator over the entries in the collection. Multiple iterators
-// can exist at the same time. Marking entries as invalid/skipped is more expensive, but thread
-// safe. You must call ApplyConcurrentDrops after all of the iterators are finished.
-func (s *SeriesCollection) Iterator() SeriesCollectionIterator {
-	return SeriesCollectionIterator{
-		s:      s,
-		length: s.Length(),
-		index:  -1,
-	}
-}
-
-// SeriesCollectionIterator is an iterator over the collection of series.
-type SeriesCollectionIterator struct {
-	s      *SeriesCollection
-	length int
-	index  int
-}
-
-// Next advances the iterator and returns false if it's done.
-func (i *SeriesCollectionIterator) Next() bool {
-	i.index++
-	return i.index < i.length
-}
-
-// Helpers that return the current state of the iterator.
-
-func (i SeriesCollectionIterator) Index() int             { return i.index }
-func (i SeriesCollectionIterator) Length() int            { return i.length }
-func (i SeriesCollectionIterator) Point() models.Point    { return i.s.Points[i.index] }
-func (i SeriesCollectionIterator) Key() []byte            { return i.s.Keys[i.index] }
-func (i SeriesCollectionIterator) SeriesKey() []byte      { return i.s.SeriesKeys[i.index] }
-func (i SeriesCollectionIterator) Name() []byte           { return i.s.Names[i.index] }
-func (i SeriesCollectionIterator) Tags() models.Tags      { return i.s.Tags[i.index] }
-func (i SeriesCollectionIterator) Type() models.FieldType { return i.s.Types[i.index] }
-func (i SeriesCollectionIterator) SeriesID() SeriesID     { return i.s.SeriesIDs[i.index] }
-
-// Invalid flags the current entry as invalid, including it in the set of dropped keys and
-// recording a reason. Only the first reason is kept. This is safe for concurrent callers,
-// but ApplyConcurrentDrops must be called after all iterators are finished.
-func (i *SeriesCollectionIterator) Invalid(reason string) {
-	i.s.invalidIndex(i.index, reason)
-}

tsdb/series_collection_test.go

@@ -1,149 +0,0 @@
-package tsdb
-
-import (
-	"reflect"
-	"testing"
-	"time"
-
-	"github.com/influxdata/influxdb/v2/models"
-)
-
-func TestSeriesCollection(t *testing.T) {
-	// some helper functions. short names because local scope and frequently used.
-	var (
-		equal = reflect.DeepEqual
-		b     = func(s string) []byte { return []byte(s) }
-		bs    = func(s ...string) [][]byte {
-			out := make([][]byte, len(s))
-			for i := range s {
-				out[i] = b(s[i])
-			}
-			return out
-		}
-
-		assertEqual = func(t *testing.T, name string, got, wanted interface{}) {
-			t.Helper()
-			if !equal(got, wanted) {
-				t.Fatalf("bad %s: got: %v but wanted: %v", name, got, wanted)
-			}
-		}
-	)
-
-	t.Run("New", func(t *testing.T) {
-		points := []models.Point{
-			models.MustNewPoint("a", models.Tags{}, models.Fields{"f": 1.0}, time.Now()),
-			models.MustNewPoint("b", models.Tags{}, models.Fields{"b": true}, time.Now()),
-			models.MustNewPoint("c", models.Tags{}, models.Fields{"i": int64(1)}, time.Now()),
-		}
-		collection := NewSeriesCollection(points)
-
-		assertEqual(t, "length", collection.Length(), 3)
-
-		for iter := collection.Iterator(); iter.Next(); {
-			ipt, spt := iter.Point(), points[iter.Index()]
-			fi := spt.FieldIterator()
-			fi.Next()
-
-			assertEqual(t, "point", ipt, spt)
-			assertEqual(t, "key", iter.Key(), spt.Key())
-			assertEqual(t, "name", iter.Name(), spt.Name())
-			assertEqual(t, "tags", iter.Tags(), spt.Tags())
-			assertEqual(t, "type", iter.Type(), fi.Type())
-		}
-	})
-
-	t.Run("Copy", func(t *testing.T) {
-		collection := &SeriesCollection{
-			Keys:  bs("ka", "kb", "kc"),
-			Names: bs("na", "nb", "nc"),
-		}
-
-		collection.Copy(0, 2)
-		assertEqual(t, "keys", collection.Keys, bs("kc", "kb", "kc"))
-		assertEqual(t, "names", collection.Names, bs("nc", "nb", "nc"))
-
-		collection.Copy(0, 4) // out of bounds
-		assertEqual(t, "keys", collection.Keys, bs("kc", "kb", "kc"))
-		assertEqual(t, "names", collection.Names, bs("nc", "nb", "nc"))
-	})
-
-	t.Run("Swap", func(t *testing.T) {
-		collection := &SeriesCollection{
-			Keys:  bs("ka", "kb", "kc"),
-			Names: bs("na", "nb", "nc"),
-		}
-
-		collection.Swap(0, 2)
-		assertEqual(t, "keys", collection.Keys, bs("kc", "kb", "ka"))
-		assertEqual(t, "names", collection.Names, bs("nc", "nb", "na"))
-
-		collection.Swap(0, 4) // out of bounds
-		assertEqual(t, "keys", collection.Keys, bs("kc", "kb", "ka"))
-		assertEqual(t, "names", collection.Names, bs("nc", "nb", "na"))
-	})
-
-	t.Run("Truncate", func(t *testing.T) {
-		collection := &SeriesCollection{
-			Keys:  bs("ka", "kb", "kc"),
-			Names: bs("na", "nb", "nc"),
-		}
-
-		collection.Truncate(1)
-		assertEqual(t, "keys", collection.Keys, bs("ka"))
-		assertEqual(t, "names", collection.Names, bs("na"))
-
-		collection.Truncate(0)
-		assertEqual(t, "keys", collection.Keys, bs())
-		assertEqual(t, "names", collection.Names, bs())
-	})
-
-	t.Run("Advance", func(t *testing.T) {
-		collection := &SeriesCollection{
-			Keys:  bs("ka", "kb", "kc"),
-			Names: bs("na", "nb", "nc"),
-		}
-
-		collection.Advance(1)
-		assertEqual(t, "keys", collection.Keys, bs("kb", "kc"))
-		assertEqual(t, "names", collection.Names, bs("nb", "nc"))
-
-		collection.Advance(1)
-		assertEqual(t, "keys", collection.Keys, bs("kc"))
-		assertEqual(t, "names", collection.Names, bs("nc"))
-	})
-
-	t.Run("InvalidateAll", func(t *testing.T) {
-		collection := &SeriesCollection{Keys: bs("ka", "kb", "kc")}
-
-		collection.InvalidateAll("test reason")
-		assertEqual(t, "length", collection.Length(), 0)
-		assertEqual(t, "error", collection.PartialWriteError(), PartialWriteError{
-			Reason:      "test reason",
-			Dropped:     3,
-			DroppedKeys: bs("ka", "kb", "kc"),
-		})
-	})
-
-	t.Run("Invalid", func(t *testing.T) {
-		collection := &SeriesCollection{Keys: bs("ka", "kb", "kc")}
-
-		// invalidate half the entries
-		for iter := collection.Iterator(); iter.Next(); {
-			if iter.Index()%2 == 0 {
-				iter.Invalid("test reason")
-			}
-		}
-
-		// nothing happens yet: all values are staged
-		assertEqual(t, "length", collection.Length(), 3)
-
-		// apply all of the invalid calls
-		collection.ApplyConcurrentDrops()
-		assertEqual(t, "length", collection.Length(), 1)
-		assertEqual(t, "error", collection.PartialWriteError(), PartialWriteError{
-			Reason:      "test reason",
-			Dropped:     2,
-			DroppedKeys: bs("ka", "kc"),
-		})
-	})
-}

tsdb/series_id.go

@@ -1,78 +0,0 @@
-package tsdb
-
-import (
-	"unsafe"
-
-	"github.com/influxdata/influxdb/v2/models"
-)
-
-const (
-	// constants describing bit layout of id and type info
-	seriesIDTypeFlag  = 1 << 63                   // a flag marking that the id contains type info
-	seriesIDValueMask = 0xFFFFFFFF                // series ids numerically are 32 bits
-	seriesIDTypeShift = 32                        // we put the type right after the value info
-	seriesIDTypeMask  = 0xFF << seriesIDTypeShift // a mask for the type byte
-	seriesIDSize      = 8
-)
-
-// SeriesID is the type of a series id. It is logically a uint64, but encoded as a struct so
-// that we gain more type checking when changing operations on it. The field is exported only
-// so that tests that use reflection based comparisons still work; no one should use the field
-// directly.
-type SeriesID struct{ ID uint64 }
-
-// NewSeriesID constructs a series id from the raw value. It discards any type information.
-func NewSeriesID(id uint64) SeriesID { return SeriesID{ID: id & seriesIDValueMask} }
-
-// IsZero returns if the SeriesID is zero.
-func (s SeriesID) IsZero() bool { return s.ID == 0 }
-
-// ID returns the raw id for the SeriesID.
-func (s SeriesID) RawID() uint64 { return s.ID }
-
-// WithType constructs a SeriesIDTyped with the given type.
-func (s SeriesID) WithType(typ models.FieldType) SeriesIDTyped {
-	return NewSeriesIDTyped(s.ID | seriesIDTypeFlag | (uint64(typ&0xFF) << seriesIDTypeShift))
-}
-
-// Greater returns if the SeriesID is greater than the passed in value.
-func (s SeriesID) Greater(o SeriesID) bool { return s.ID > o.ID }
-
-// Less returns if the SeriesID is less than the passed in value.
-func (s SeriesID) Less(o SeriesID) bool { return s.ID < o.ID }
-
-// SeriesIDType represents a series id with a type. It is logically a uint64, but encoded as
-// a struct so that we gain more type checking when changing operations on it. The field is
-// exported only so that tests that use reflection based comparisons still work; no one should
-// use the field directly.
-type SeriesIDTyped struct{ ID uint64 }
-
-// NewSeriesIDTyped constructs a typed series id from the raw values.
-func NewSeriesIDTyped(id uint64) SeriesIDTyped { return SeriesIDTyped{ID: id} }
-
-// IsZero returns if the SeriesIDTyped is zero. It ignores any type information.
-func (s SeriesIDTyped) IsZero() bool { return s.ID&seriesIDValueMask == 0 }
-
-// ID returns the raw id for the SeriesIDTyped.
-func (s SeriesIDTyped) RawID() uint64 { return s.ID }
-
-// SeriesID constructs a SeriesID, discarding any type information.
-func (s SeriesIDTyped) SeriesID() SeriesID { return NewSeriesID(s.ID) }
-
-// HasType returns if the id actually contains a type.
-func (s SeriesIDTyped) HasType() bool { return s.ID&seriesIDTypeFlag > 0 }
-
-// Type returns the associated type.
-func (s SeriesIDTyped) Type() models.FieldType {
-	return models.FieldType((s.ID & seriesIDTypeMask) >> seriesIDTypeShift)
-}
-
-type (
-	// some static assertions that the SeriesIDSize matches the structs we defined.
-	// if the values are not the same, at least one will be negative causing a compilation failure
-	_ [seriesIDSize - unsafe.Sizeof(SeriesID{})]byte
-	_ [unsafe.Sizeof(SeriesID{}) - seriesIDSize]byte
-
-	_ [seriesIDSize - unsafe.Sizeof(SeriesIDTyped{})]byte
-	_ [unsafe.Sizeof(SeriesIDTyped{}) - seriesIDSize]byte
-)

tsdb/series_id_test.go

@@ -1,31 +0,0 @@
-package tsdb
-
-import (
-	"math/rand"
-	"testing"
-
-	"github.com/influxdata/influxdb/v2/models"
-)
-
-func TestSeriesID(t *testing.T) {
-	types := []models.FieldType{
-		models.Integer,
-		models.Float,
-		models.Boolean,
-		models.String,
-		models.Unsigned,
-	}
-
-	for i := 0; i < 1000000; i++ {
-		id := NewSeriesID(uint64(rand.Int31()))
-		for _, typ := range types {
-			typed := id.WithType(typ)
-			if got := typed.Type(); got != typ {
-				t.Fatalf("wanted: %v got: %v", typ, got)
-			}
-			if got := typed.SeriesID(); id != got {
-				t.Fatalf("wanted: %016x got: %016x", id, got)
-			}
-		}
-	}
-}

tsdb/series_iterators.go

@@ -1,699 +0,0 @@
-package tsdb
-
-import (
-	"bytes"
-
-	"github.com/influxdata/influxql"
-)
-
-// SeriesIDElem represents a single series and optional expression.
-type SeriesIDElem struct {
-	SeriesID SeriesID
-	Expr     influxql.Expr
-}
-
-// SeriesIDIterator represents a iterator over a list of series ids.
-type SeriesIDIterator interface {
-	Next() (SeriesIDElem, error)
-	Close() error
-}
-
-// SeriesIDSetIterator represents an iterator that can produce a SeriesIDSet.
-type SeriesIDSetIterator interface {
-	SeriesIDIterator
-	SeriesIDSet() *SeriesIDSet
-}
-
-type seriesIDSetIterator struct {
-	ss  *SeriesIDSet
-	itr SeriesIDSetIterable
-}
-
-func NewSeriesIDSetIterator(ss *SeriesIDSet) SeriesIDSetIterator {
-	if ss == nil || ss.bitmap == nil {
-		return nil
-	}
-	return &seriesIDSetIterator{ss: ss, itr: ss.Iterator()}
-}
-
-func (itr *seriesIDSetIterator) Next() (SeriesIDElem, error) {
-	if !itr.itr.HasNext() {
-		return SeriesIDElem{}, nil
-	}
-	return SeriesIDElem{SeriesID: NewSeriesID(uint64(itr.itr.Next()))}, nil
-}
-
-func (itr *seriesIDSetIterator) Close() error { return nil }
-
-func (itr *seriesIDSetIterator) SeriesIDSet() *SeriesIDSet { return itr.ss }
-
-// NewSeriesIDSetIterators returns a slice of SeriesIDSetIterator if all itrs
-// can be type casted. Otherwise returns nil.
-func NewSeriesIDSetIterators(itrs []SeriesIDIterator) []SeriesIDSetIterator {
-	if len(itrs) == 0 {
-		return nil
-	}
-
-	a := make([]SeriesIDSetIterator, len(itrs))
-	for i := range itrs {
-		if itr, ok := itrs[i].(SeriesIDSetIterator); ok {
-			a[i] = itr
-		} else {
-			return nil
-		}
-	}
-	return a
-}
-
-// NewSeriesIDSliceIterator returns a SeriesIDIterator that iterates over a slice.
-func NewSeriesIDSliceIterator(ids []SeriesID) *SeriesIDSliceIterator {
-	return &SeriesIDSliceIterator{ids: ids}
-}
-
-// SeriesIDSliceIterator iterates over a slice of series ids.
-type SeriesIDSliceIterator struct {
-	ids []SeriesID
-}
-
-// Next returns the next series id in the slice.
-func (itr *SeriesIDSliceIterator) Next() (SeriesIDElem, error) {
-	if len(itr.ids) == 0 {
-		return SeriesIDElem{}, nil
-	}
-	id := itr.ids[0]
-	itr.ids = itr.ids[1:]
-	return SeriesIDElem{SeriesID: id}, nil
-}
-
-func (itr *SeriesIDSliceIterator) Close() error { return nil }
-
-// SeriesIDSet returns a set of all remaining ids.
-func (itr *SeriesIDSliceIterator) SeriesIDSet() *SeriesIDSet {
-	s := NewSeriesIDSet()
-	for _, id := range itr.ids {
-		s.AddNoLock(id)
-	}
-	return s
-}
-
-type SeriesIDIterators []SeriesIDIterator
-
-func (a SeriesIDIterators) Close() (err error) {
-	for i := range a {
-		if e := a[i].Close(); e != nil && err == nil {
-			err = e
-		}
-	}
-	return err
-}
-
-// seriesIDExprIterator is an iterator that attaches an associated expression.
-type SeriesIDExprIterator struct {
-	itr  SeriesIDIterator
-	expr influxql.Expr
-}
-
-// newSeriesIDExprIterator returns a new instance of seriesIDExprIterator.
-func NewSeriesIDExprIterator(itr SeriesIDIterator, expr influxql.Expr) SeriesIDIterator {
-	if itr == nil {
-		return nil
-	}
-
-	return &SeriesIDExprIterator{
-		itr:  itr,
-		expr: expr,
-	}
-}
-
-func (itr *SeriesIDExprIterator) Close() error {
-	return itr.itr.Close()
-}
-
-// Next returns the next element in the iterator.
-func (itr *SeriesIDExprIterator) Next() (SeriesIDElem, error) {
-	elem, err := itr.itr.Next()
-	if err != nil {
-		return SeriesIDElem{}, err
-	} else if elem.SeriesID.IsZero() {
-		return SeriesIDElem{}, nil
-	}
-	elem.Expr = itr.expr
-	return elem, nil
-}
-
-// MergeSeriesIDIterators returns an iterator that merges a set of iterators.
-// Iterators that are first in the list take precedence and a deletion by those
-// early iterators will invalidate elements by later iterators.
-func MergeSeriesIDIterators(itrs ...SeriesIDIterator) SeriesIDIterator {
-	if n := len(itrs); n == 0 {
-		return nil
-	} else if n == 1 {
-		return itrs[0]
-	}
-
-	// Merge as series id sets, if available.
-	if a := NewSeriesIDSetIterators(itrs); a != nil {
-		sets := make([]*SeriesIDSet, len(a))
-		for i := range a {
-			sets[i] = a[i].SeriesIDSet()
-		}
-
-		ss := NewSeriesIDSet()
-		ss.Merge(sets...)
-		SeriesIDIterators(itrs).Close()
-		return NewSeriesIDSetIterator(ss)
-	}
-
-	return &seriesIDMergeIterator{
-		buf:  make([]SeriesIDElem, len(itrs)),
-		itrs: itrs,
-	}
-}
-
-// seriesIDMergeIterator is an iterator that merges multiple iterators together.
-type seriesIDMergeIterator struct {
-	buf  []SeriesIDElem
-	itrs []SeriesIDIterator
-}
-
-func (itr *seriesIDMergeIterator) Close() (err error) {
-	return SeriesIDIterators(itr.itrs).Close()
-}
-
-// Next returns the element with the next lowest name/tags across the iterators.
-func (itr *seriesIDMergeIterator) Next() (SeriesIDElem, error) {
-	// Find next lowest id amongst the buffers.
-	var elem SeriesIDElem
-	for i := range itr.buf {
-		buf := &itr.buf[i]
-
-		// Fill buffer.
-		if buf.SeriesID.IsZero() {
-			elem, err := itr.itrs[i].Next()
-			if err != nil {
-				return SeriesIDElem{}, nil
-			} else if elem.SeriesID.IsZero() {
-				continue
-			}
-			itr.buf[i] = elem
-		}
-
-		if elem.SeriesID.IsZero() || buf.SeriesID.Less(elem.SeriesID) {
-			elem = *buf
-		}
-	}
-
-	// Return EOF if no elements remaining.
-	if elem.SeriesID.IsZero() {
-		return SeriesIDElem{}, nil
-	}
-
-	// Clear matching buffers.
-	for i := range itr.buf {
-		if itr.buf[i].SeriesID == elem.SeriesID {
-			itr.buf[i].SeriesID = SeriesID{}
-		}
-	}
-	return elem, nil
-}
-
-// IntersectSeriesIDIterators returns an iterator that only returns series which
-// occur in both iterators. If both series have associated expressions then
-// they are combined together.
-func IntersectSeriesIDIterators(itr0, itr1 SeriesIDIterator) SeriesIDIterator {
-	if itr0 == nil || itr1 == nil {
-		if itr0 != nil {
-			itr0.Close()
-		}
-		if itr1 != nil {
-			itr1.Close()
-		}
-		return nil
-	}
-
-	// Create series id set, if available.
-	if a := NewSeriesIDSetIterators([]SeriesIDIterator{itr0, itr1}); a != nil {
-		itr0.Close()
-		itr1.Close()
-		return NewSeriesIDSetIterator(a[0].SeriesIDSet().And(a[1].SeriesIDSet()))
-	}
-
-	return &seriesIDIntersectIterator{itrs: [2]SeriesIDIterator{itr0, itr1}}
-}
-
-// seriesIDIntersectIterator is an iterator that merges two iterators together.
-type seriesIDIntersectIterator struct {
-	buf  [2]SeriesIDElem
-	itrs [2]SeriesIDIterator
-}
-
-func (itr *seriesIDIntersectIterator) Close() (err error) {
-	if e := itr.itrs[0].Close(); e != nil && err == nil {
-		err = e
-	}
-	if e := itr.itrs[1].Close(); e != nil && err == nil {
-		err = e
-	}
-	return err
-}
-
-// Next returns the next element which occurs in both iterators.
-func (itr *seriesIDIntersectIterator) Next() (_ SeriesIDElem, err error) {
-	for {
-		// Fill buffers.
-		if itr.buf[0].SeriesID.IsZero() {
-			if itr.buf[0], err = itr.itrs[0].Next(); err != nil {
-				return SeriesIDElem{}, err
-			}
-		}
-		if itr.buf[1].SeriesID.IsZero() {
-			if itr.buf[1], err = itr.itrs[1].Next(); err != nil {
-				return SeriesIDElem{}, err
-			}
-		}
-
-		// Exit if either buffer is still empty.
-		if itr.buf[0].SeriesID.IsZero() || itr.buf[1].SeriesID.IsZero() {
-			return SeriesIDElem{}, nil
-		}
-
-		// Skip if both series are not equal.
-		if a, b := itr.buf[0].SeriesID, itr.buf[1].SeriesID; a.Less(b) {
-			itr.buf[0].SeriesID = SeriesID{}
-			continue
-		} else if a.Greater(b) {
-			itr.buf[1].SeriesID = SeriesID{}
-			continue
-		}
-
-		// Merge series together if equal.
-		elem := itr.buf[0]
-
-		// Attach expression.
-		expr0 := itr.buf[0].Expr
-		expr1 := itr.buf[1].Expr
-		if expr0 == nil {
-			elem.Expr = expr1
-		} else if expr1 == nil {
-			elem.Expr = expr0
-		} else {
-			elem.Expr = influxql.Reduce(&influxql.BinaryExpr{
-				Op:  influxql.AND,
-				LHS: expr0,
-				RHS: expr1,
-			}, nil)
-		}
-
-		itr.buf[0].SeriesID, itr.buf[1].SeriesID = SeriesID{}, SeriesID{}
-		return elem, nil
-	}
-}
-
-// UnionSeriesIDIterators returns an iterator that returns series from both
-// both iterators. If both series have associated expressions then they are
-// combined together.
-func UnionSeriesIDIterators(itr0, itr1 SeriesIDIterator) SeriesIDIterator {
-	// Return other iterator if either one is nil.
-	if itr0 == nil {
-		return itr1
-	} else if itr1 == nil {
-		return itr0
-	}
-
-	// Create series id set, if available.
-	if a := NewSeriesIDSetIterators([]SeriesIDIterator{itr0, itr1}); a != nil {
-		itr0.Close()
-		itr1.Close()
-		ss := NewSeriesIDSet()
-		ss.Merge(a[0].SeriesIDSet(), a[1].SeriesIDSet())
-		return NewSeriesIDSetIterator(ss)
-	}
-
-	return &seriesIDUnionIterator{itrs: [2]SeriesIDIterator{itr0, itr1}}
-}
-
-// seriesIDUnionIterator is an iterator that unions two iterators together.
-type seriesIDUnionIterator struct {
-	buf  [2]SeriesIDElem
-	itrs [2]SeriesIDIterator
-}
-
-func (itr *seriesIDUnionIterator) Close() (err error) {
-	if e := itr.itrs[0].Close(); e != nil && err == nil {
-		err = e
-	}
-	if e := itr.itrs[1].Close(); e != nil && err == nil {
-		err = e
-	}
-	return err
-}
-
-// Next returns the next element which occurs in both iterators.
-func (itr *seriesIDUnionIterator) Next() (_ SeriesIDElem, err error) {
-	// Fill buffers.
-	if itr.buf[0].SeriesID.IsZero() {
-		if itr.buf[0], err = itr.itrs[0].Next(); err != nil {
-			return SeriesIDElem{}, err
-		}
-	}
-	if itr.buf[1].SeriesID.IsZero() {
-		if itr.buf[1], err = itr.itrs[1].Next(); err != nil {
-			return SeriesIDElem{}, err
-		}
-	}
-
-	// Return non-zero or lesser series.
-	if a, b := itr.buf[0].SeriesID, itr.buf[1].SeriesID; a.IsZero() && b.IsZero() {
-		return SeriesIDElem{}, nil
-	} else if b.IsZero() || (!a.IsZero() && a.Less(b)) {
-		elem := itr.buf[0]
-		itr.buf[0].SeriesID = SeriesID{}
-		return elem, nil
-	} else if a.IsZero() || (!b.IsZero() && a.Greater(b)) {
-		elem := itr.buf[1]
-		itr.buf[1].SeriesID = SeriesID{}
-		return elem, nil
-	}
-
-	// Attach element.
-	elem := itr.buf[0]
-
-	// Attach expression.
-	expr0 := itr.buf[0].Expr
-	expr1 := itr.buf[1].Expr
-	if expr0 != nil && expr1 != nil {
-		elem.Expr = influxql.Reduce(&influxql.BinaryExpr{
-			Op:  influxql.OR,
-			LHS: expr0,
-			RHS: expr1,
-		}, nil)
-	} else {
-		elem.Expr = nil
-	}
-
-	itr.buf[0].SeriesID, itr.buf[1].SeriesID = SeriesID{}, SeriesID{}
-	return elem, nil
-}
-
-// DifferenceSeriesIDIterators returns an iterator that only returns series which
-// occur the first iterator but not the second iterator.
-func DifferenceSeriesIDIterators(itr0, itr1 SeriesIDIterator) SeriesIDIterator {
-	if itr0 == nil && itr1 == nil {
-		return nil
-	} else if itr1 == nil {
-		return itr0
-	} else if itr0 == nil {
-		itr1.Close()
-		return nil
-	}
-
-	// Create series id set, if available.
-	if a := NewSeriesIDSetIterators([]SeriesIDIterator{itr0, itr1}); a != nil {
-		itr0.Close()
-		itr1.Close()
-		return NewSeriesIDSetIterator(NewSeriesIDSetNegate(a[0].SeriesIDSet(), a[1].SeriesIDSet()))
-	}
-
-	return &seriesIDDifferenceIterator{itrs: [2]SeriesIDIterator{itr0, itr1}}
-}
-
-// seriesIDDifferenceIterator is an iterator that merges two iterators together.
-type seriesIDDifferenceIterator struct {
-	buf  [2]SeriesIDElem
-	itrs [2]SeriesIDIterator
-}
-
-func (itr *seriesIDDifferenceIterator) Close() (err error) {
-	if e := itr.itrs[0].Close(); e != nil && err == nil {
-		err = e
-	}
-	if e := itr.itrs[1].Close(); e != nil && err == nil {
-		err = e
-	}
-	return err
-}
-
-// Next returns the next element which occurs only in the first iterator.
-func (itr *seriesIDDifferenceIterator) Next() (_ SeriesIDElem, err error) {
-	for {
-		// Fill buffers.
-		if itr.buf[0].SeriesID.IsZero() {
-			if itr.buf[0], err = itr.itrs[0].Next(); err != nil {
-				return SeriesIDElem{}, err
-			}
-		}
-		if itr.buf[1].SeriesID.IsZero() {
-			if itr.buf[1], err = itr.itrs[1].Next(); err != nil {
-				return SeriesIDElem{}, err
-			}
-		}
-
-		// Exit if first buffer is still empty.
-		if itr.buf[0].SeriesID.IsZero() {
-			return SeriesIDElem{}, nil
-		} else if itr.buf[1].SeriesID.IsZero() {
-			elem := itr.buf[0]
-			itr.buf[0].SeriesID = SeriesID{}
-			return elem, nil
-		}
-
-		// Return first series if it's less.
-		// If second series is less then skip it.
-		// If both series are equal then skip both.
-		if a, b := itr.buf[0].SeriesID, itr.buf[1].SeriesID; a.Less(b) {
-			elem := itr.buf[0]
-			itr.buf[0].SeriesID = SeriesID{}
-			return elem, nil
-		} else if a.Greater(b) {
-			itr.buf[1].SeriesID = SeriesID{}
-			continue
-		} else {
-			itr.buf[0].SeriesID, itr.buf[1].SeriesID = SeriesID{}, SeriesID{}
-			continue
-		}
-	}
-}
-
-// MeasurementIterator represents a iterator over a list of measurements.
-type MeasurementIterator interface {
-	Close() error
-	Next() ([]byte, error)
-}
-
-// MergeMeasurementIterators returns an iterator that merges a set of iterators.
-// Iterators that are first in the list take precedence and a deletion by those
-// early iterators will invalidate elements by later iterators.
-func MergeMeasurementIterators(itrs ...MeasurementIterator) MeasurementIterator {
-	if len(itrs) == 0 {
-		return nil
-	} else if len(itrs) == 1 {
-		return itrs[0]
-	}
-
-	return &measurementMergeIterator{
-		buf:  make([][]byte, len(itrs)),
-		itrs: itrs,
-	}
-}
-
-type measurementMergeIterator struct {
-	buf  [][]byte
-	itrs []MeasurementIterator
-}
-
-func (itr *measurementMergeIterator) Close() (err error) {
-	for i := range itr.itrs {
-		if e := itr.itrs[i].Close(); e != nil && err == nil {
-			err = e
-		}
-	}
-	return err
-}
-
-// Next returns the element with the next lowest name across the iterators.
-//
-// If multiple iterators contain the same name then the first is returned
-// and the remaining ones are skipped.
-func (itr *measurementMergeIterator) Next() (_ []byte, err error) {
-	// Find next lowest name amongst the buffers.
-	var name []byte
-	for i, buf := range itr.buf {
-		// Fill buffer if empty.
-		if buf == nil {
-			if buf, err = itr.itrs[i].Next(); err != nil {
-				return nil, err
-			} else if buf != nil {
-				itr.buf[i] = buf
-			} else {
-				continue
-			}
-		}
-
-		// Find next lowest name.
-		if name == nil || bytes.Compare(itr.buf[i], name) == -1 {
-			name = itr.buf[i]
-		}
-	}
-
-	// Return nil if no elements remaining.
-	if name == nil {
-		return nil, nil
-	}
-
-	// Merge all elements together and clear buffers.
-	for i, buf := range itr.buf {
-		if buf == nil || !bytes.Equal(buf, name) {
-			continue
-		}
-		itr.buf[i] = nil
-	}
-	return name, nil
-}
-
-// TagKeyIterator represents a iterator over a list of tag keys.
-type TagKeyIterator interface {
-	Close() error
-	Next() ([]byte, error)
-}
-
-// MergeTagKeyIterators returns an iterator that merges a set of iterators.
-func MergeTagKeyIterators(itrs ...TagKeyIterator) TagKeyIterator {
-	if len(itrs) == 0 {
-		return nil
-	} else if len(itrs) == 1 {
-		return itrs[0]
-	}
-
-	return &tagKeyMergeIterator{
-		buf:  make([][]byte, len(itrs)),
-		itrs: itrs,
-	}
-}
-
-type tagKeyMergeIterator struct {
-	buf  [][]byte
-	itrs []TagKeyIterator
-}
-
-func (itr *tagKeyMergeIterator) Close() (err error) {
-	for i := range itr.itrs {
-		if e := itr.itrs[i].Close(); e != nil && err == nil {
-			err = e
-		}
-	}
-	return err
-}
-
-// Next returns the element with the next lowest key across the iterators.
-//
-// If multiple iterators contain the same key then the first is returned
-// and the remaining ones are skipped.
-func (itr *tagKeyMergeIterator) Next() (_ []byte, err error) {
-	// Find next lowest key amongst the buffers.
-	var key []byte
-	for i, buf := range itr.buf {
-		// Fill buffer.
-		if buf == nil {
-			if buf, err = itr.itrs[i].Next(); err != nil {
-				return nil, err
-			} else if buf != nil {
-				itr.buf[i] = buf
-			} else {
-				continue
-			}
-		}
-
-		// Find next lowest key.
-		if key == nil || bytes.Compare(buf, key) == -1 {
-			key = buf
-		}
-	}
-
-	// Return nil if no elements remaining.
-	if key == nil {
-		return nil, nil
-	}
-
-	// Merge elements and clear buffers.
-	for i, buf := range itr.buf {
-		if buf == nil || !bytes.Equal(buf, key) {
-			continue
-		}
-		itr.buf[i] = nil
-	}
-	return key, nil
-}
-
-// TagValueIterator represents a iterator over a list of tag values.
-type TagValueIterator interface {
-	Close() error
-	Next() ([]byte, error)
-}
-
-// MergeTagValueIterators returns an iterator that merges a set of iterators.
-func MergeTagValueIterators(itrs ...TagValueIterator) TagValueIterator {
-	if len(itrs) == 0 {
-		return nil
-	} else if len(itrs) == 1 {
-		return itrs[0]
-	}
-
-	return &tagValueMergeIterator{
-		buf:  make([][]byte, len(itrs)),
-		itrs: itrs,
-	}
-}
-
-type tagValueMergeIterator struct {
-	buf  [][]byte
-	itrs []TagValueIterator
-}
-
-func (itr *tagValueMergeIterator) Close() (err error) {
-	for i := range itr.itrs {
-		if e := itr.itrs[i].Close(); e != nil && err == nil {
-			err = e
-		}
-	}
-	return err
-}
-
-// Next returns the element with the next lowest value across the iterators.
-//
-// If multiple iterators contain the same value then the first is returned
-// and the remaining ones are skipped.
-func (itr *tagValueMergeIterator) Next() (_ []byte, err error) {
-	// Find next lowest value amongst the buffers.
-	var value []byte
-	for i, buf := range itr.buf {
-		// Fill buffer.
-		if buf == nil {
-			if buf, err = itr.itrs[i].Next(); err != nil {
-				return nil, err
-			} else if buf != nil {
-				itr.buf[i] = buf
-			} else {
-				continue
-			}
-		}
-
-		// Find next lowest value.
-		if value == nil || bytes.Compare(buf, value) == -1 {
-			value = buf
-		}
-	}
-
-	// Return nil if no elements remaining.
-	if value == nil {
-		return nil, nil
-	}
-
-	// Merge elements and clear buffers.
-	for i, buf := range itr.buf {
-		if buf == nil || !bytes.Equal(buf, value) {
-			continue
-		}
-		itr.buf[i] = nil
-	}
-	return value, nil
-}

tsdb/series_iterators_test.go

@@ -1,357 +0,0 @@
-package tsdb_test
-
-import (
-	"compress/gzip"
-	"context"
-	"fmt"
-	"io/ioutil"
-	"os"
-	"path/filepath"
-	"reflect"
-	"sync"
-	"testing"
-
-	"github.com/influxdata/influxdb/v2/logger"
-	"github.com/influxdata/influxdb/v2/models"
-	"github.com/influxdata/influxdb/v2/query"
-	"github.com/influxdata/influxdb/v2/tsdb"
-	"github.com/influxdata/influxdb/v2/tsdb/seriesfile"
-	"github.com/influxdata/influxdb/v2/tsdb/tsi1"
-	"github.com/influxdata/influxql"
-)
-
-func toSeriesIDs(ids []uint64) []tsdb.SeriesID {
-	sids := make([]tsdb.SeriesID, 0, len(ids))
-	for _, id := range ids {
-		sids = append(sids, tsdb.NewSeriesID(id))
-	}
-	return sids
-}
-
-// Ensure iterator can merge multiple iterators together.
-func TestMergeSeriesIDIterators(t *testing.T) {
-	itr := tsdb.MergeSeriesIDIterators(
-		tsdb.NewSeriesIDSliceIterator(toSeriesIDs([]uint64{1, 2, 3})),
-		tsdb.NewSeriesIDSliceIterator(nil),
-		tsdb.NewSeriesIDSliceIterator(toSeriesIDs([]uint64{1, 2, 3, 4})),
-	)
-
-	if e, err := itr.Next(); err != nil {
-		t.Fatal(err)
-	} else if !reflect.DeepEqual(e, tsdb.SeriesIDElem{SeriesID: tsdb.NewSeriesID(1)}) {
-		t.Fatalf("unexpected elem(0): %#v", e)
-	}
-	if e, err := itr.Next(); err != nil {
-		t.Fatal(err)
-	} else if !reflect.DeepEqual(e, tsdb.SeriesIDElem{SeriesID: tsdb.NewSeriesID(2)}) {
-		t.Fatalf("unexpected elem(1): %#v", e)
-	}
-	if e, err := itr.Next(); err != nil {
-		t.Fatal(err)
-	} else if !reflect.DeepEqual(e, tsdb.SeriesIDElem{SeriesID: tsdb.NewSeriesID(3)}) {
-		t.Fatalf("unexpected elem(2): %#v", e)
-	}
-	if e, err := itr.Next(); err != nil {
-		t.Fatal(err)
-	} else if !reflect.DeepEqual(e, tsdb.SeriesIDElem{SeriesID: tsdb.NewSeriesID(4)}) {
-		t.Fatalf("unexpected elem(3): %#v", e)
-	}
-	if e, err := itr.Next(); err != nil {
-		t.Fatal(err)
-	} else if !e.SeriesID.IsZero() {
-		t.Fatalf("expected nil elem: %#v", e)
-	}
-}
-
-// Index wraps a series file and index.
-type Index struct {
-	rootPath string
-
-	config tsi1.Config
-	*tsi1.Index
-	sfile *seriesfile.SeriesFile
-}
-
-// MustNewIndex will initialize a new index using the provide type. It creates
-// everything under the same root directory so it can be cleanly removed on Close.
-//
-// The index will not be opened.
-func MustNewIndex(c tsi1.Config) *Index {
-	rootPath, err := ioutil.TempDir("", "influxdb-tsdb")
-	if err != nil {
-		panic(err)
-	}
-
-	seriesPath, err := ioutil.TempDir(rootPath, "_series")
-	if err != nil {
-		panic(err)
-	}
-
-	sfile := seriesfile.NewSeriesFile(seriesPath)
-	if err := sfile.Open(context.Background()); err != nil {
-		panic(err)
-	}
-
-	i := tsi1.NewIndex(sfile, c, tsi1.WithPath(filepath.Join(rootPath, "index")))
-
-	if testing.Verbose() {
-		i.WithLogger(logger.New(os.Stderr))
-	}
-
-	idx := &Index{
-		config:   c,
-		Index:    i,
-		rootPath: rootPath,
-		sfile:    sfile,
-	}
-	return idx
-}
-
-// MustOpenNewIndex will initialize a new index using the provide type and opens
-// it.
-func MustOpenNewIndex(c tsi1.Config) *Index {
-	idx := MustNewIndex(c)
-	idx.MustOpen()
-	return idx
-}
-
-// MustOpen opens the underlying index or panics.
-func (i *Index) MustOpen() {
-	if err := i.Index.Open(context.Background()); err != nil {
-		panic(err)
-	}
-}
-
-// Reopen closes and re-opens the underlying index, without removing any data.
-func (i *Index) Reopen() error {
-	if err := i.Index.Close(); err != nil {
-		return err
-	}
-
-	if err := i.sfile.Close(); err != nil {
-		return err
-	}
-
-	i.sfile = seriesfile.NewSeriesFile(i.sfile.Path())
-	if err := i.sfile.Open(context.Background()); err != nil {
-		return err
-	}
-
-	i.Index = tsi1.NewIndex(i.SeriesFile(), i.config,
-		tsi1.WithPath(filepath.Join(i.rootPath, "index")))
-	return i.Index.Open(context.Background())
-}
-
-// Close closes the index cleanly and removes all on-disk data.
-func (i *Index) Close() error {
-	if err := i.Index.Close(); err != nil {
-		return err
-	}
-
-	if err := i.sfile.Close(); err != nil {
-		return err
-	}
-	return os.RemoveAll(i.rootPath)
-}
-
-// This benchmark compares the TagSets implementation across index types.
-//
-// In the case of the TSI index, TagSets has to merge results across all several
-// index partitions.
-//
-// Typical results on an i7 laptop.
-//
-// BenchmarkIndex_TagSets/1M_series/tsi1-8    	     100	  18995530 ns/op	 5221180 B/op	   20379 allocs/op
-func BenchmarkIndex_TagSets(b *testing.B) {
-	// Read line-protocol and coerce into tsdb format.
-	// 1M series generated with:
-	// $inch -b 10000 -c 1 -t 10,10,10,10,10,10 -f 1 -m 5 -p 1
-	fd, err := os.Open("testdata/line-protocol-1M.txt.gz")
-	if err != nil {
-		b.Fatal(err)
-	}
-
-	gzr, err := gzip.NewReader(fd)
-	if err != nil {
-		fd.Close()
-		b.Fatal(err)
-	}
-
-	data, err := ioutil.ReadAll(gzr)
-	if err != nil {
-		b.Fatal(err)
-	}
-
-	if err := fd.Close(); err != nil {
-		b.Fatal(err)
-	}
-
-	points, err := models.ParsePoints(data, []byte("mm"))
-	if err != nil {
-		b.Fatal(err)
-	}
-
-	// setup writes all of the above points to the index.
-	setup := func(idx *Index) {
-		batchSize := 10000
-		for j := 0; j < 1; j++ {
-			for i := 0; i < len(points); i += batchSize {
-				collection := tsdb.NewSeriesCollection(points[i : i+batchSize])
-				if err := idx.CreateSeriesListIfNotExists(collection); err != nil {
-					b.Fatal(err)
-				}
-			}
-		}
-	}
-
-	var errResult error
-
-	// This benchmark will merge eight bitsets each containing ~10,000 series IDs.
-	b.Run("1M series", func(b *testing.B) {
-		idx := MustOpenNewIndex(tsi1.NewConfig())
-		setup(idx)
-		defer idx.Close()
-
-		name := []byte("m4")
-		opt := query.IteratorOptions{Condition: influxql.MustParseExpr(`"tag5"::tag = 'value0'`)}
-
-		ts := func() ([]*query.TagSet, error) {
-			return idx.Index.TagSets(name, opt)
-		}
-
-		b.ReportAllocs()
-		b.ResetTimer()
-
-		for i := 0; i < b.N; i++ {
-			// Will call TagSets on the appropriate implementation.
-			_, errResult = ts()
-			if errResult != nil {
-				b.Fatal(err)
-			}
-		}
-
-		if err := idx.Close(); err != nil {
-			b.Fatal(err)
-		}
-	})
-}
-
-// This benchmark concurrently writes series to the index and fetches cached bitsets.
-// The idea is to emphasize the performance difference when bitset caching is on and off.
-//
-// Typical results for an i7 laptop
-//
-// BenchmarkIndex_ConcurrentWriteQuery/inmem/queries_100000/cache-8   	  1	5963346204 ns/op	2499655768 B/op	 23964183 allocs/op
-// BenchmarkIndex_ConcurrentWriteQuery/inmem/queries_100000/no_cache-8    1	5314841090 ns/op	2499495280 B/op	 23963322 allocs/op
-// BenchmarkIndex_ConcurrentWriteQuery/tsi1/queries_100000/cache-8        1	1645048376 ns/op	2215402840 B/op	 23048978 allocs/op
-// BenchmarkIndex_ConcurrentWriteQuery/tsi1/queries_100000/no_cache-8     1	22242155616 ns/op	28277544136 B/op 79620463 allocs/op
-func BenchmarkIndex_ConcurrentWriteQuery(b *testing.B) {
-	// Read line-protocol and coerce into tsdb format.
-	// 1M series generated with:
-	// $inch -b 10000 -c 1 -t 10,10,10,10,10,10 -f 1 -m 5 -p 1
-	fd, err := os.Open("testdata/line-protocol-1M.txt.gz")
-	if err != nil {
-		b.Fatal(err)
-	}
-
-	gzr, err := gzip.NewReader(fd)
-	if err != nil {
-		fd.Close()
-		b.Fatal(err)
-	}
-
-	data, err := ioutil.ReadAll(gzr)
-	if err != nil {
-		b.Fatal(err)
-	}
-
-	if err := fd.Close(); err != nil {
-		b.Fatal(err)
-	}
-
-	points, err := models.ParsePoints(data, []byte("mm"))
-	if err != nil {
-		b.Fatal(err)
-	}
-
-	runBenchmark := func(b *testing.B, queryN int, cacheSize uint64) {
-		config := tsi1.NewConfig()
-		config.SeriesIDSetCacheSize = cacheSize
-		idx := MustOpenNewIndex(config)
-		var wg sync.WaitGroup
-		begin := make(chan struct{})
-
-		// Run concurrent iterator...
-		runIter := func() {
-			keys := [][]string{
-				{"m0", "tag2", "value4"},
-				{"m1", "tag3", "value5"},
-				{"m2", "tag4", "value6"},
-				{"m3", "tag0", "value8"},
-				{"m4", "tag5", "value0"},
-			}
-
-			<-begin // Wait for writes to land
-			for i := 0; i < queryN/5; i++ {
-				for _, key := range keys {
-					itr, err := idx.TagValueSeriesIDIterator([]byte(key[0]), []byte(key[1]), []byte(key[2]))
-					if err != nil {
-						b.Fatal(err)
-					}
-
-					if itr == nil {
-						panic("should not happen")
-					}
-
-					if err := itr.Close(); err != nil {
-						b.Fatal(err)
-					}
-				}
-			}
-		}
-
-		batchSize := 10000
-		wg.Add(1)
-		go func() { defer wg.Done(); runIter() }()
-		var once sync.Once
-		for j := 0; j < b.N; j++ {
-			for i := 0; i < len(points); i += batchSize {
-				collection := tsdb.NewSeriesCollection(points[i : i+batchSize])
-				if err := idx.CreateSeriesListIfNotExists(collection); err != nil {
-					b.Fatal(err)
-				}
-				once.Do(func() { close(begin) })
-			}
-
-			// Wait for queries to finish
-			wg.Wait()
-
-			// Reset the index...
-			b.StopTimer()
-			if err := idx.Close(); err != nil {
-				b.Fatal(err)
-			}
-
-			// Re-open everything
-			idx = MustOpenNewIndex(tsi1.NewConfig())
-			wg.Add(1)
-			begin = make(chan struct{})
-			once = sync.Once{}
-			go func() { defer wg.Done(); runIter() }()
-			b.StartTimer()
-		}
-	}
-
-	queries := []int{1e5}
-	for _, queryN := range queries {
-		b.Run(fmt.Sprintf("queries %d", queryN), func(b *testing.B) {
-			b.Run("cache", func(b *testing.B) {
-				runBenchmark(b, queryN, tsi1.DefaultSeriesIDSetCacheSize)
-			})
-
-			b.Run("no cache", func(b *testing.B) {
-				runBenchmark(b, queryN, 0)
-			})
-		})
-	}
-}

tsdb/series_set.go

@@ -1,293 +0,0 @@
-package tsdb
-
-import (
-	"io"
-	"sync"
-	"unsafe"
-
-	"github.com/RoaringBitmap/roaring"
-)
-
-// SeriesIDSet represents a lockable bitmap of series ids.
-type SeriesIDSet struct {
-	sync.RWMutex
-	bitmap *roaring.Bitmap
-}
-
-// NewSeriesIDSet returns a new instance of SeriesIDSet.
-func NewSeriesIDSet(a ...SeriesID) *SeriesIDSet {
-	ss := &SeriesIDSet{bitmap: roaring.NewBitmap()}
-	if len(a) > 0 {
-		a32 := make([]uint32, len(a))
-		for i := range a {
-			a32[i] = uint32(a[i].RawID())
-		}
-		ss.bitmap.AddMany(a32)
-	}
-	return ss
-}
-
-// NewSeriesIDSetNegate returns a new SeriesIDSet containing all the elements in a
-// that are not present in b. That is, the set difference between a and b.
-func NewSeriesIDSetNegate(a, b *SeriesIDSet) *SeriesIDSet {
-	a.RLock()
-	defer a.RUnlock()
-	b.RLock()
-	defer b.RUnlock()
-
-	return &SeriesIDSet{bitmap: roaring.AndNot(a.bitmap, b.bitmap)}
-}
-
-// Bytes estimates the memory footprint of this SeriesIDSet, in bytes.
-func (s *SeriesIDSet) Bytes() int {
-	var b int
-	s.RLock()
-	b += 24 // mu RWMutex is 24 bytes
-	b += int(unsafe.Sizeof(s.bitmap)) + int(s.bitmap.GetSizeInBytes())
-	s.RUnlock()
-	return b
-}
-
-// Add adds the series id to the set.
-func (s *SeriesIDSet) Add(id SeriesID) {
-	s.Lock()
-	defer s.Unlock()
-	s.AddNoLock(id)
-}
-
-// AddNoLock adds the series id to the set. Add is not safe for use from multiple
-// goroutines. Callers must manage synchronization.
-func (s *SeriesIDSet) AddNoLock(id SeriesID) {
-	s.bitmap.Add(uint32(id.RawID()))
-}
-
-// AddMany adds multiple ids to the SeriesIDSet. AddMany takes a lock, so may not be
-// optimal to call many times with few ids.
-func (s *SeriesIDSet) AddMany(ids ...SeriesID) {
-	if len(ids) == 0 {
-		return
-	}
-
-	a32 := make([]uint32, len(ids))
-	for i := range ids {
-		a32[i] = uint32(ids[i].RawID())
-	}
-
-	s.Lock()
-	defer s.Unlock()
-	s.bitmap.AddMany(a32)
-}
-
-// Contains returns true if the id exists in the set.
-func (s *SeriesIDSet) Contains(id SeriesID) bool {
-	s.RLock()
-	x := s.ContainsNoLock(id)
-	s.RUnlock()
-	return x
-}
-
-// ContainsNoLock returns true if the id exists in the set. ContainsNoLock is
-// not safe for use from multiple goroutines. The caller must manage synchronization.
-func (s *SeriesIDSet) ContainsNoLock(id SeriesID) bool {
-	return s.bitmap.Contains(uint32(id.RawID()))
-}
-
-// Remove removes the id from the set.
-func (s *SeriesIDSet) Remove(id SeriesID) {
-	s.Lock()
-	defer s.Unlock()
-	s.RemoveNoLock(id)
-}
-
-// RemoveNoLock removes the id from the set. RemoveNoLock is not safe for use
-// from multiple goroutines. The caller must manage synchronization.
-func (s *SeriesIDSet) RemoveNoLock(id SeriesID) {
-	s.bitmap.Remove(uint32(id.RawID()))
-}
-
-// Cardinality returns the cardinality of the SeriesIDSet.
-func (s *SeriesIDSet) Cardinality() uint64 {
-	s.RLock()
-	defer s.RUnlock()
-	return s.bitmap.GetCardinality()
-}
-
-// Merge merged the contents of others into s. The caller does not need to
-// provide s as an argument, and the contents of s will always be present in s
-// after Merge returns.
-func (s *SeriesIDSet) Merge(others ...*SeriesIDSet) {
-	bms := make([]*roaring.Bitmap, 0, len(others)+1)
-
-	s.RLock()
-	bms = append(bms, s.bitmap) // Add ourself.
-
-	// Add other bitsets.
-	for _, other := range others {
-		other.RLock()
-		defer other.RUnlock() // Hold until we have merged all the bitmaps
-		bms = append(bms, other.bitmap)
-	}
-
-	result := roaring.FastOr(bms...)
-	s.RUnlock()
-
-	s.Lock()
-	s.bitmap = result
-	s.Unlock()
-}
-
-// MergeInPlace merges other into s, modifying s in the process.
-func (s *SeriesIDSet) MergeInPlace(other *SeriesIDSet) {
-	if s == other {
-		return
-	}
-
-	other.RLock()
-	s.Lock()
-	s.bitmap.Or(other.bitmap)
-	s.Unlock()
-	other.RUnlock()
-}
-
-// Equals returns true if other and s are the same set of ids.
-func (s *SeriesIDSet) Equals(other *SeriesIDSet) bool {
-	if s == other {
-		return true
-	}
-
-	s.RLock()
-	defer s.RUnlock()
-	other.RLock()
-	defer other.RUnlock()
-	return s.bitmap.Equals(other.bitmap)
-}
-
-// And returns a new SeriesIDSet containing elements that were present in s and other.
-func (s *SeriesIDSet) And(other *SeriesIDSet) *SeriesIDSet {
-	s.RLock()
-	defer s.RUnlock()
-	other.RLock()
-	defer other.RUnlock()
-	return &SeriesIDSet{bitmap: roaring.And(s.bitmap, other.bitmap)}
-}
-
-// RemoveSet removes all values in other from s, if they exist.
-func (s *SeriesIDSet) RemoveSet(other *SeriesIDSet) {
-	s.RLock()
-	defer s.RUnlock()
-	other.RLock()
-	defer other.RUnlock()
-	s.bitmap.AndNot(other.bitmap)
-}
-
-// ForEach calls f for each id in the set. The function is applied to the IDs
-// in ascending order.
-func (s *SeriesIDSet) ForEach(f func(id SeriesID)) {
-	s.RLock()
-	defer s.RUnlock()
-	itr := s.bitmap.Iterator()
-	for itr.HasNext() {
-		f(NewSeriesID(uint64(itr.Next())))
-	}
-}
-
-// ForEachNoLock calls f for each id in the set without taking a lock.
-func (s *SeriesIDSet) ForEachNoLock(f func(id SeriesID)) {
-	itr := s.bitmap.Iterator()
-	for itr.HasNext() {
-		f(NewSeriesID(uint64(itr.Next())))
-	}
-}
-
-func (s *SeriesIDSet) String() string {
-	s.RLock()
-	defer s.RUnlock()
-	return s.bitmap.String()
-}
-
-// Diff removes from s any elements also present in other.
-func (s *SeriesIDSet) Diff(other *SeriesIDSet) {
-	other.RLock()
-	defer other.RUnlock()
-
-	s.Lock()
-	defer s.Unlock()
-	s.bitmap = roaring.AndNot(s.bitmap, other.bitmap)
-}
-
-// Clone returns a new SeriesIDSet with a deep copy of the underlying bitmap.
-func (s *SeriesIDSet) Clone() *SeriesIDSet {
-	// Cloning the SeriesIDSet involves cloning s's bitmap.
-	// Unfortunately, if the bitmap is set to COW, the original bitmap is modified during clone,
-	// so we have to take a write lock rather than a read lock.
-	// For now, we'll just hold a write lock for clone; if this shows up as a bottleneck later,
-	// we can conditionally RLock if we are not COW.
-	s.Lock()
-	defer s.Unlock()
-	return s.CloneNoLock()
-}
-
-// CloneNoLock calls Clone without taking a lock.
-func (s *SeriesIDSet) CloneNoLock() *SeriesIDSet {
-	new := NewSeriesIDSet()
-	new.bitmap = s.bitmap.Clone()
-	return new
-}
-
-// Iterator returns an iterator to the underlying bitmap.
-// This iterator is not protected by a lock.
-func (s *SeriesIDSet) Iterator() SeriesIDSetIterable {
-	return s.bitmap.Iterator()
-}
-
-// UnmarshalBinary unmarshals data into the set.
-func (s *SeriesIDSet) UnmarshalBinary(data []byte) error {
-	s.Lock()
-	defer s.Unlock()
-	return s.bitmap.UnmarshalBinary(data)
-}
-
-// UnmarshalBinaryUnsafe unmarshals data into the set.
-// References to the underlying data are used so data should not be reused by caller.
-func (s *SeriesIDSet) UnmarshalBinaryUnsafe(data []byte) error {
-	s.Lock()
-	defer s.Unlock()
-	_, err := s.bitmap.FromBuffer(data)
-	return err
-}
-
-// WriteTo writes the set to w.
-func (s *SeriesIDSet) WriteTo(w io.Writer) (int64, error) {
-	s.RLock()
-	defer s.RUnlock()
-	return s.bitmap.WriteTo(w)
-}
-
-// Clear clears the underlying bitmap for re-use. Clear is safe for use by multiple goroutines.
-func (s *SeriesIDSet) Clear() {
-	s.Lock()
-	defer s.Unlock()
-	s.ClearNoLock()
-}
-
-// ClearNoLock clears the underlying bitmap for re-use without taking a lock.
-func (s *SeriesIDSet) ClearNoLock() {
-	s.bitmap.Clear()
-}
-
-// Slice returns a slice of series ids.
-func (s *SeriesIDSet) Slice() []uint64 {
-	s.RLock()
-	defer s.RUnlock()
-
-	a := make([]uint64, 0, s.bitmap.GetCardinality())
-	for _, seriesID := range s.bitmap.ToArray() {
-		a = append(a, uint64(seriesID))
-	}
-	return a
-}
-
-type SeriesIDSetIterable interface {
-	HasNext() bool
-	Next() uint32
-}

tsdb/series_set_test.go

@@ -1,778 +0,0 @@
-package tsdb
-
-import (
-	"bytes"
-	"fmt"
-	"math"
-	"math/rand"
-	"runtime"
-	"sync"
-	"testing"
-)
-
-func TestSeriesIDSet_NewSeriesIDSetNegate(t *testing.T) {
-	examples := [][3][]uint64{
-		[3][]uint64{
-			{1, 10, 20, 30},
-			{10, 12, 13, 14, 20},
-			{1, 30},
-		},
-		[3][]uint64{
-			{},
-			{10},
-			{},
-		},
-		[3][]uint64{
-			{1, 10, 20, 30},
-			{1, 10, 20, 30},
-			{},
-		},
-		[3][]uint64{
-			{1, 10},
-			{1, 10, 100},
-			{},
-		},
-		[3][]uint64{
-			{1, 10},
-			{},
-			{1, 10},
-		},
-	}
-
-	for i, example := range examples {
-		t.Run(fmt.Sprint(i), func(t *testing.T) {
-			// Build sets.
-			a, b := NewSeriesIDSet(), NewSeriesIDSet()
-			for _, v := range example[0] {
-				a.Add(NewSeriesID(v))
-			}
-			for _, v := range example[1] {
-				b.Add(NewSeriesID(v))
-			}
-
-			expected := NewSeriesIDSet()
-			for _, v := range example[2] {
-				expected.Add(NewSeriesID(v))
-			}
-
-			got := NewSeriesIDSetNegate(a, b)
-			if got.String() != expected.String() {
-				t.Fatalf("got %s, expected %s", got.String(), expected.String())
-			}
-		})
-	}
-}
-
-func TestSeriesIDSet_RemoveSet(t *testing.T) {
-	examples := [][3][]uint64{
-		[3][]uint64{
-			{1, 10, 20, 30},
-			{10, 12, 13, 14, 20},
-			{1, 30},
-		},
-		[3][]uint64{
-			{},
-			{10},
-			{},
-		},
-		[3][]uint64{
-			{1, 10, 20, 30},
-			{1, 10, 20, 30},
-			{},
-		},
-		[3][]uint64{
-			{1, 10},
-			{1, 10, 100},
-			{},
-		},
-		[3][]uint64{
-			{1, 10},
-			{},
-			{1, 10},
-		},
-	}
-
-	for i, example := range examples {
-		t.Run(fmt.Sprint(i), func(t *testing.T) {
-			// Build sets.
-			a, b := NewSeriesIDSet(), NewSeriesIDSet()
-			for _, v := range example[0] {
-				a.Add(NewSeriesID(v))
-			}
-			for _, v := range example[1] {
-				b.Add(NewSeriesID(v))
-			}
-
-			expected := NewSeriesIDSet()
-			for _, v := range example[2] {
-				expected.Add(NewSeriesID(v))
-			}
-
-			a.RemoveSet(b)
-			if a.String() != expected.String() {
-				t.Fatalf("got %s, expected %s", a.String(), expected.String())
-			}
-		})
-	}
-}
-
-// Ensure that cloning is race-free.
-func TestSeriesIDSet_Clone_Race(t *testing.T) {
-	main := NewSeriesIDSet()
-	total := NewSeriesIDSet()
-	for i := uint64(0); i < 1024; i++ {
-		id := NewSeriesID(i)
-		main.AddNoLock(id)
-		total.AddNoLock(id)
-	}
-
-	// One test with a closure around the main SeriesIDSet,
-	// so that we can run a subtest with and without COW.
-	test := func(t *testing.T) {
-		n := 10 * (runtime.NumCPU() + 1)
-		clones := make([]*SeriesIDSet, n)
-		var wg sync.WaitGroup
-		wg.Add(n)
-		for i := 1; i <= n; i++ {
-			go func(i int) {
-				defer wg.Done()
-				clones[i-1] = main.Clone()
-
-				for j := 0; j < 1000; j++ {
-					id := NewSeriesID(uint64(j + (100000 * i)))
-					total.Add(id)
-					clones[i-1].AddNoLock(id)
-				}
-			}(i)
-		}
-
-		wg.Wait()
-		for _, o := range clones {
-			if got, exp := o.Cardinality(), uint64(2024); got != exp {
-				t.Errorf("got cardinality %d, expected %d", got, exp)
-			}
-		}
-
-		// The original set should be unaffected
-		if got, exp := main.Cardinality(), uint64(1024); got != exp {
-			t.Errorf("got cardinality %d, expected %d", got, exp)
-		}
-
-		// Merging the clones should result in only 1024 shared values.
-		union := NewSeriesIDSet()
-		for _, o := range clones {
-			o.ForEachNoLock(func(id SeriesID) {
-				union.AddNoLock(id)
-			})
-		}
-
-		if !union.Equals(total) {
-			t.Fatal("union not equal to total")
-		}
-	}
-	t.Run("clone", test)
-}
-
-var resultBool bool
-
-// Contains should be typically a constant time lookup. Example results on a laptop:
-//
-// BenchmarkSeriesIDSet_Contains/1-4 			20000000	        68.5 ns/op	       0 B/op	       0 allocs/op
-// BenchmarkSeriesIDSet_Contains/2-4 			20000000	        70.8 ns/op	       0 B/op	       0 allocs/op
-// BenchmarkSeriesIDSet_Contains/10-4         	20000000	        70.3 ns/op	       0 B/op	       0 allocs/op
-// BenchmarkSeriesIDSet_Contains/100-4        	20000000	        71.3 ns/op	       0 B/op	       0 allocs/op
-// BenchmarkSeriesIDSet_Contains/1000-4       	20000000	        80.5 ns/op	       0 B/op	       0 allocs/op
-// BenchmarkSeriesIDSet_Contains/10000-4      	20000000	        67.3 ns/op	       0 B/op	       0 allocs/op
-// BenchmarkSeriesIDSet_Contains/100000-4     	20000000	        73.1 ns/op	       0 B/op	       0 allocs/op
-// BenchmarkSeriesIDSet_Contains/1000000-4    	20000000	        77.3 ns/op	       0 B/op	       0 allocs/op
-// BenchmarkSeriesIDSet_Contains/10000000-4   	20000000	        75.3 ns/op	       0 B/op	       0 allocs/op
-func BenchmarkSeriesIDSet_Contains(b *testing.B) {
-	cardinalities := []uint64{1, 2, 10, 100, 1000, 10000, 100000, 1000000, 10000000}
-
-	for _, cardinality := range cardinalities {
-		// Setup...
-		set := NewSeriesIDSet()
-		for i := uint64(0); i < cardinality; i++ {
-			set.Add(NewSeriesID(i))
-		}
-
-		lookup := cardinality / 2
-		b.Run(fmt.Sprint(cardinality), func(b *testing.B) {
-			for i := 0; i < b.N; i++ {
-				resultBool = set.Contains(NewSeriesID(lookup))
-			}
-		})
-	}
-}
-
-var set *SeriesIDSet
-
-// Adding to a larger bitset shouldn't be significantly more expensive than adding
-// to a smaller one. This benchmark adds a value to different cardinality sets.
-//
-// Example results from a laptop:
-// BenchmarkSeriesIDSet_Add/1-4 	 		1000000	      1053 ns/op	      48 B/op	       2 allocs/op
-// BenchmarkSeriesIDSet_Add/2-4 	 		5000000	       303 ns/op	       0 B/op	       0 allocs/op
-// BenchmarkSeriesIDSet_Add/10-4         	5000000	       348 ns/op	       0 B/op	       0 allocs/op
-// BenchmarkSeriesIDSet_Add/100-4        	5000000	       373 ns/op	       0 B/op	       0 allocs/op
-// BenchmarkSeriesIDSet_Add/1000-4       	5000000	       342 ns/op	       0 B/op	       0 allocs/op
-//
-//
-func BenchmarkSeriesIDSet_AddMore(b *testing.B) {
-	cardinalities := []uint64{1, 2, 10, 100, 1000, 10000, 100000, 1000000, 10000000}
-
-	for _, cardinality := range cardinalities {
-		// Setup...
-		set = NewSeriesIDSet()
-		for i := uint64(0); i < cardinality-1; i++ {
-			set.Add(NewSeriesID(i))
-		}
-
-		b.Run(fmt.Sprint(cardinality), func(b *testing.B) {
-			for i := 0; i < b.N; i++ {
-				// Add next value
-				set.Add(NewSeriesID(cardinality))
-
-				b.StopTimer()
-				set.Remove(NewSeriesID(cardinality))
-				b.StartTimer()
-			}
-		})
-	}
-}
-
-// Add benchmarks the cost of adding the same element to a set versus the
-// cost of checking if it exists before adding it.
-//
-// Typical benchmarks from a laptop:
-//
-// BenchmarkSeriesIDSet_Add/cardinality_1000000_add/same-8    							20000000	        64.8 ns/op	       0 B/op	       0 allocs/op
-// BenchmarkSeriesIDSet_Add/cardinality_1000000_add/random-8  	 						 2000000	       704 	 ns/op	       5 B/op	       0 allocs/op
-// BenchmarkSeriesIDSet_Add/cardinality_1000000_add/same_no_lock-8         				50000000	        40.3 ns/op	       0 B/op	       0 allocs/op
-// BenchmarkSeriesIDSet_Add/cardinality_1000000_add/random_no_lock-8       				 2000000	       644   ns/op	       5 B/op	       0 allocs/op
-// BenchmarkSeriesIDSet_Add/cardinality_1000000_check_add/same_no_lock-8   				50000000	        34.0 ns/op	       0 B/op	       0 allocs/op
-// BenchmarkSeriesIDSet_Add/cardinality_1000000_check_add/random_no_lock-8 	 		   	 2000000	       860   ns/op	      14 B/op	       0 allocs/op
-// BenchmarkSeriesIDSet_Add/cardinality_1000000_check_add/same_global_lock-8         	30000000	        49.8 ns/op	       0 B/op	       0 allocs/op
-// BenchmarkSeriesIDSet_Add/cardinality_1000000_check_add/random_global_lock-8       	 2000000	       914   ns/op	       0 B/op	       0 allocs/op
-// BenchmarkSeriesIDSet_Add/cardinality_1000000_check_add/same_multi_lock-8          	30000000	        39.7 ns/op	       0 B/op	       0 allocs/op
-// BenchmarkSeriesIDSet_Add/cardinality_1000000_check_add/random_multi_lock-8        	 1000000	      1002   ns/op	       0 B/op	       0 allocs/op
-//
-func BenchmarkSeriesIDSet_Add(b *testing.B) {
-	// Setup...
-	set = NewSeriesIDSet()
-	for i := uint64(0); i < 1000000; i++ {
-		set.Add(NewSeriesID(i))
-	}
-	lookup := NewSeriesID(300032)
-
-	// Add the same value over and over.
-	b.Run("cardinality_1000000_add", func(b *testing.B) {
-		b.Run("same", func(b *testing.B) {
-			for i := 0; i < b.N; i++ {
-				set.Add(lookup)
-			}
-		})
-
-		b.Run("random", func(b *testing.B) {
-			for i := 0; i < b.N; i++ {
-				b.StopTimer()
-				x := NewSeriesID(uint64(rand.Intn(math.MaxInt32)))
-				b.StartTimer()
-				set.Add(x)
-			}
-		})
-
-		b.Run("same no lock", func(b *testing.B) {
-			for i := 0; i < b.N; i++ {
-				set.AddNoLock(lookup)
-			}
-		})
-
-		b.Run("random no lock", func(b *testing.B) {
-			for i := 0; i < b.N; i++ {
-				b.StopTimer()
-				x := NewSeriesID(uint64(rand.Intn(math.MaxInt32)))
-				b.StartTimer()
-				set.AddNoLock(x)
-			}
-		})
-	})
-
-	// Add the same value over and over with no lock
-	b.Run("cardinality_1000000_check_add", func(b *testing.B) {
-		b.Run("same no lock", func(b *testing.B) {
-			for i := 0; i < b.N; i++ {
-				if !set.ContainsNoLock(lookup) {
-					set.AddNoLock(lookup)
-				}
-			}
-		})
-
-		b.Run("random no lock", func(b *testing.B) {
-			for i := 0; i < b.N; i++ {
-				b.StopTimer()
-				x := NewSeriesID(uint64(rand.Intn(math.MaxInt32)))
-				b.StartTimer()
-				if !set.ContainsNoLock(x) {
-					set.AddNoLock(x)
-				}
-			}
-		})
-
-		b.Run("same global lock", func(b *testing.B) {
-			for i := 0; i < b.N; i++ {
-				set.Lock()
-				if !set.ContainsNoLock(lookup) {
-					set.AddNoLock(lookup)
-				}
-				set.Unlock()
-			}
-		})
-
-		b.Run("random global lock", func(b *testing.B) {
-			for i := 0; i < b.N; i++ {
-				b.StopTimer()
-				x := NewSeriesID(uint64(rand.Intn(math.MaxInt32)))
-				b.StartTimer()
-				set.Lock()
-				if !set.ContainsNoLock(x) {
-					set.AddNoLock(x)
-				}
-				set.Unlock()
-			}
-		})
-
-		b.Run("same multi lock", func(b *testing.B) {
-			for i := 0; i < b.N; i++ {
-				if !set.Contains(lookup) {
-					set.Add(lookup)
-				}
-			}
-		})
-
-		b.Run("random multi lock", func(b *testing.B) {
-			for i := 0; i < b.N; i++ {
-				b.StopTimer()
-				x := NewSeriesID(uint64(rand.Intn(math.MaxInt32)))
-				b.StartTimer()
-				if !set.Contains(x) {
-					set.Add(x)
-				}
-			}
-		})
-	})
-}
-
-var ssResult *SeriesIDSet
-
-// Benchmark various ways of creating a copy of a bitmap. Note, Clone_COW will result
-// in a bitmap where future modifications will involve copies.
-//
-// Typical results from an i7 laptop.
-// BenchmarkSeriesIDSet_Clone/cardinality_1000/re-use/Clone-8         	   			   30000	     44171 ns/op	   47200 B/op	    1737 allocs/op
-// BenchmarkSeriesIDSet_Clone/cardinality_1000/re-use/Merge-8         	  			  100000	     17877 ns/op	   39008 B/op	      30 allocs/op
-// BenchmarkSeriesIDSet_Clone/cardinality_1000/re-use/MergeInPlace-8  	  			  200000	      7367 ns/op	       0 B/op	       0 allocs/op
-// BenchmarkSeriesIDSet_Clone/cardinality_1000/re-use/Add-8           	   			   10000	    137460 ns/op	   62336 B/op	    2596 allocs/op
-// BenchmarkSeriesIDSet_Clone/cardinality_1000/re-use/WriteTo-8       	   			   30000	     52896 ns/op	   35872 B/op	     866 allocs/op
-// BenchmarkSeriesIDSet_Clone/cardinality_1000/don't_re-use/Clone-8   	   			   30000	     41940 ns/op	   47200 B/op	    1737 allocs/op
-// BenchmarkSeriesIDSet_Clone/cardinality_1000/don't_re-use/Merge-8             	  100000	     17624 ns/op	   39008 B/op	      30 allocs/op
-// BenchmarkSeriesIDSet_Clone/cardinality_1000/don't_re-use/MergeInPlace-8      	  100000	     17320 ns/op	   38880 B/op	      28 allocs/op
-// BenchmarkSeriesIDSet_Clone/cardinality_1000/don't_re-use/Add-8               	   10000	    167544 ns/op	  101216 B/op	    2624 allocs/op
-// BenchmarkSeriesIDSet_Clone/cardinality_1000/don't_re-use/WriteTo-8           	   20000	     66976 ns/op	   52897 B/op	     869 allocs/op
-// BenchmarkSeriesIDSet_Clone/cardinality_10000/re-use/Clone-8                  	   10000	    179933 ns/op	  177072 B/op	    5895 allocs/op
-// BenchmarkSeriesIDSet_Clone/cardinality_10000/re-use/Merge-8                  	   20000	     77574 ns/op	  210656 B/op	      42 allocs/op
-// BenchmarkSeriesIDSet_Clone/cardinality_10000/re-use/MergeInPlace-8           	  100000	     23645 ns/op	       0 B/op	       0 allocs/op
-// BenchmarkSeriesIDSet_Clone/cardinality_10000/re-use/Add-8                    	    2000	    689254 ns/op	  224161 B/op	    9572 allocs/op
-// BenchmarkSeriesIDSet_Clone/cardinality_10000/re-use/WriteTo-8                	   10000	    199052 ns/op	  118791 B/op	    2945 allocs/op
-// BenchmarkSeriesIDSet_Clone/cardinality_10000/don't_re-use/Clone-8            	   10000	    183137 ns/op	  177073 B/op	    5895 allocs/op
-// BenchmarkSeriesIDSet_Clone/cardinality_10000/don't_re-use/Merge-8            	   20000	     77502 ns/op	  210656 B/op	      42 allocs/op
-// BenchmarkSeriesIDSet_Clone/cardinality_10000/don't_re-use/MergeInPlace-8     	   20000	     72610 ns/op	  210528 B/op	      40 allocs/op
-// BenchmarkSeriesIDSet_Clone/cardinality_10000/don't_re-use/Add-8              	    2000	    724789 ns/op	  434691 B/op	    9612 allocs/op
-// BenchmarkSeriesIDSet_Clone/cardinality_10000/don't_re-use/WriteTo-8          	   10000	    215734 ns/op	  177159 B/op	    2948 allocs/op
-// BenchmarkSeriesIDSet_Clone/cardinality_100000/re-use/Clone-8                 	    5000	    244971 ns/op	  377648 B/op	    6111 allocs/op
-// BenchmarkSeriesIDSet_Clone/cardinality_100000/re-use/Merge-8                 	   20000	     90580 ns/op	  210656 B/op	      42 allocs/op
-// BenchmarkSeriesIDSet_Clone/cardinality_100000/re-use/MergeInPlace-8          	   50000	     24697 ns/op	       0 B/op	       0 allocs/op
-// BenchmarkSeriesIDSet_Clone/cardinality_100000/re-use/Add-8                   	     500	   3274456 ns/op	  758996 B/op	   19853 allocs/op
-// BenchmarkSeriesIDSet_Clone/cardinality_100000/re-use/WriteTo-8               	    5000	    248791 ns/op	  122392 B/op	    3053 allocs/op
-// BenchmarkSeriesIDSet_Clone/cardinality_100000/don't_re-use/Clone-8           	    5000	    269152 ns/op	  377648 B/op	    6111 allocs/op
-// BenchmarkSeriesIDSet_Clone/cardinality_100000/don't_re-use/Merge-8           	   20000	     85948 ns/op	  210657 B/op	      42 allocs/op
-// BenchmarkSeriesIDSet_Clone/cardinality_100000/don't_re-use/MergeInPlace-8    	   20000	     78142 ns/op	  210528 B/op	      40 allocs/op
-// BenchmarkSeriesIDSet_Clone/cardinality_100000/don't_re-use/Add-8             	     500	   3123753 ns/op	  969529 B/op	   19893 allocs/op
-// BenchmarkSeriesIDSet_Clone/cardinality_100000/don't_re-use/WriteTo-8         	   10000	    230657 ns/op	  180684 B/op	    3056 allocs/op
-// BenchmarkSeriesIDSet_Clone/cardinality_1000000/re-use/Clone-8                	    3000	    551781 ns/op	 2245424 B/op	    6111 allocs/op
-// BenchmarkSeriesIDSet_Clone/cardinality_1000000/re-use/Merge-8                	   20000	     92104 ns/op	  210656 B/op	      42 allocs/op
-// BenchmarkSeriesIDSet_Clone/cardinality_1000000/re-use/MergeInPlace-8         	   50000	     27408 ns/op	       0 B/op	       0 allocs/op
-// BenchmarkSeriesIDSet_Clone/cardinality_1000000/re-use/Add-8                  	     100	  22573498 ns/op	 6420446 B/op	   30520 allocs/op
-// BenchmarkSeriesIDSet_Clone/cardinality_1000000/re-use/WriteTo-8              	    5000	    284901 ns/op	  123522 B/op	    3053 allocs/op
-// BenchmarkSeriesIDSet_Clone/cardinality_1000000/don't_re-use/Clone-8          	    3000	    679284 ns/op	 2245424 B/op	    6111 allocs/op
-// BenchmarkSeriesIDSet_Clone/cardinality_1000000/don't_re-use/Merge-8          	   20000	     68965 ns/op	  210656 B/op	      42 allocs/op
-// BenchmarkSeriesIDSet_Clone/cardinality_1000000/don't_re-use/MergeInPlace-8   	   20000	     64236 ns/op	  210528 B/op	      40 allocs/op
-// BenchmarkSeriesIDSet_Clone/cardinality_1000000/don't_re-use/Add-8            	     100	  21960668 ns/op	 6630979 B/op	   30560 allocs/op
-// BenchmarkSeriesIDSet_Clone/cardinality_1000000/don't_re-use/WriteTo-8        	    5000	    298276 ns/op	  181890 B/op	    3056 allocs/op
-
-func BenchmarkSeriesIDSet_Clone(b *testing.B) {
-	toAddCardinalities := []int{1e3, 1e4, 1e5, 1e6}
-
-	runBenchmarks := func(b *testing.B, other *SeriesIDSet, init func() *SeriesIDSet) {
-		b.Run("Clone", func(b *testing.B) {
-			for i := 0; i < b.N; i++ {
-				ssResult = other.Clone()
-			}
-		})
-
-		b.Run("Merge", func(b *testing.B) {
-			ssResult = init()
-			for i := 0; i < b.N; i++ {
-				ssResult.Merge(other)
-				b.StopTimer()
-				ssResult = init()
-				b.StartTimer()
-			}
-		})
-
-		b.Run("MergeInPlace", func(b *testing.B) {
-			ssResult = init()
-			for i := 0; i < b.N; i++ {
-				ssResult.MergeInPlace(other)
-				b.StopTimer()
-				ssResult = init()
-				b.StartTimer()
-			}
-		})
-
-		b.Run("Add", func(b *testing.B) {
-			ssResult = init()
-			for i := 0; i < b.N; i++ {
-				itr := other.Iterator()
-				ssResult.Lock()
-				for itr.HasNext() {
-					ssResult.AddNoLock(NewSeriesID(uint64(itr.Next())))
-				}
-				ssResult.Unlock()
-				b.StopTimer()
-				ssResult = init()
-				b.StartTimer()
-			}
-		})
-
-		b.Run("WriteTo", func(b *testing.B) {
-			var buf bytes.Buffer
-			ssResult = init()
-			for i := 0; i < b.N; i++ {
-				other.WriteTo(&buf)
-				ssResult.UnmarshalBinaryUnsafe(buf.Bytes())
-				b.StopTimer()
-				ssResult = init()
-				buf.Reset()
-				b.StartTimer()
-			}
-		})
-	}
-
-	for _, toAddCardinality := range toAddCardinalities {
-		b.Run(fmt.Sprintf("cardinality %d", toAddCardinality), func(b *testing.B) {
-			ids := make([]SeriesID, 0, toAddCardinality)
-			for i := 0; i < toAddCardinality; i++ {
-				ids = append(ids, NewSeriesID(uint64(rand.Intn(200000000))))
-			}
-			other := NewSeriesIDSet(ids...)
-
-			b.Run("re-use", func(b *testing.B) {
-				base := NewSeriesIDSet()
-				runBenchmarks(b, other, func() *SeriesIDSet {
-					base.Clear()
-					return base
-				})
-			})
-
-			b.Run("don't re-use", func(b *testing.B) {
-				runBenchmarks(b, other, func() *SeriesIDSet {
-					return NewSeriesIDSet()
-				})
-			})
-		})
-	}
-}
-func BenchmarkSeriesIDSet_AddMany(b *testing.B) {
-	cardinalities := []int{1, 1e3, 1e4, 1e5, 1e6}
-	toAddCardinalities := []int{1e3, 1e4, 1e5}
-
-	for _, cardinality := range cardinalities {
-		ids := make([]SeriesID, 0, cardinality)
-		for i := 0; i < cardinality; i++ {
-			ids = append(ids, NewSeriesID(uint64(rand.Intn(200000000))))
-		}
-
-		// Setup...
-		set = NewSeriesIDSet(ids...)
-
-		// Check if the value exists before adding it under two locks.
-		b.Run(fmt.Sprintf("cardinality %d", cardinality), func(b *testing.B) {
-			for _, toAddCardinality := range toAddCardinalities {
-				ids := make([]SeriesID, 0, toAddCardinality)
-				for i := 0; i < toAddCardinality; i++ {
-					ids = append(ids, NewSeriesID(uint64(rand.Intn(200000000))))
-				}
-
-				b.Run(fmt.Sprintf("adding %d", toAddCardinality), func(b *testing.B) {
-					b.Run("AddNoLock", func(b *testing.B) {
-						clone := set.Clone()
-						for i := 0; i < b.N; i++ {
-							for _, id := range ids {
-								clone.AddNoLock(id)
-							}
-
-							b.StopTimer()
-							clone = set.Clone()
-							b.StartTimer()
-						}
-					})
-
-					b.Run("AddMany", func(b *testing.B) {
-						clone := set.Clone()
-						for i := 0; i < b.N; i++ {
-							clone.AddMany(ids...)
-							b.StopTimer()
-							clone = set.Clone()
-							b.StartTimer()
-						}
-					})
-
-					// Merge will involve a new bitmap being allocated.
-					b.Run("Merge", func(b *testing.B) {
-						clone := set.Clone()
-						for i := 0; i < b.N; i++ {
-							other := NewSeriesIDSet(ids...)
-							clone.Merge(other)
-
-							b.StopTimer()
-							clone = set.Clone()
-							b.StartTimer()
-						}
-					})
-
-					b.Run("MergeInPlace", func(b *testing.B) {
-						clone := set.Clone()
-						for i := 0; i < b.N; i++ {
-							other := NewSeriesIDSet(ids...)
-							clone.MergeInPlace(other)
-
-							b.StopTimer()
-							clone = set.Clone()
-							b.StartTimer()
-						}
-					})
-				})
-			}
-		})
-	}
-}
-
-// Remove benchmarks the cost of removing the same element in a set versus the
-// cost of checking if it exists before removing it.
-//
-// Typical benchmarks from a laptop:
-//
-// BenchmarkSeriesIDSet_Remove/cardinality_1000000_remove_same-4         		20000000	        99.1 ns/op	       0 B/op	       0 allocs/op
-// BenchmarkSeriesIDSet_Remove/cardinality_1000000_check_remove_global_lock-4   20000000	        57.7 ns/op	       0 B/op	       0 allocs/op
-// BenchmarkSeriesIDSet_Remove/cardinality_1000000_check_remove_multi_lock-4    20000000	        80.1 ns/op	       0 B/op	       0 allocs/op
-//
-func BenchmarkSeriesIDSet_Remove(b *testing.B) {
-	// Setup...
-	set = NewSeriesIDSet()
-	for i := uint64(0); i < 1000000; i++ {
-		set.Add(NewSeriesID(i))
-	}
-	lookup := uint64(300032)
-
-	// Remove the same value over and over.
-	b.Run("cardinality_1000000_remove_same", func(b *testing.B) {
-		for i := 0; i < b.N; i++ {
-			set.Remove(NewSeriesID(lookup))
-		}
-	})
-
-	// Check if the value exists before adding it. Subsequent repeats of the code
-	// will result in contains checks.
-	b.Run("cardinality_1000000_check_remove_global_lock", func(b *testing.B) {
-		for i := 0; i < b.N; i++ {
-			set.Lock()
-			if set.ContainsNoLock(NewSeriesID(lookup)) {
-				set.RemoveNoLock(NewSeriesID(lookup))
-			}
-			set.Unlock()
-		}
-	})
-
-	// Check if the value exists before adding it under two locks.
-	b.Run("cardinality_1000000_check_remove_multi_lock", func(b *testing.B) {
-		for i := 0; i < b.N; i++ {
-			if set.Contains(NewSeriesID(lookup)) {
-				set.Remove(NewSeriesID(lookup))
-			}
-		}
-	})
-}
-
-// BenchmarkSeriesIDSet_MassRemove benchmarks the cost of removing a large set of values.
-func BenchmarkSeriesIDSet_MassRemove(b *testing.B) {
-	var size = uint64(1000000)
-	// Setup...
-	set = NewSeriesIDSet()
-	for i := uint64(0); i < size; i++ {
-		set.Add(NewSeriesID(i))
-	}
-
-	// Remove one at a time
-	b.Run("cardinality_1000000_remove_each", func(b *testing.B) {
-		clone := set.Clone()
-		for i := 0; i < b.N; i++ {
-			for j := uint64(0); j < size/2; j++ {
-				clone.RemoveNoLock(NewSeriesID(j))
-			}
-
-			b.StopTimer()
-			clone = set.Clone()
-			b.StartTimer()
-		}
-	})
-
-	// This is the case where a target series id set exists.
-	b.Run("cardinality_1000000_remove_set_exists", func(b *testing.B) {
-		clone := set.Clone()
-		other := NewSeriesIDSet()
-		for j := uint64(0); j < size/2; j++ {
-			other.AddNoLock(NewSeriesID(j))
-		}
-
-		for i := 0; i < b.N; i++ {
-			clone.RemoveSet(other)
-			b.StopTimer()
-			clone = set.Clone()
-			b.StartTimer()
-		}
-	})
-
-	// Make a target series id set and negate it
-	b.Run("cardinality_1000000_remove_set", func(b *testing.B) {
-		clone := set.Clone()
-		for i := 0; i < b.N; i++ {
-			other := NewSeriesIDSet()
-			for j := uint64(0); j < size/2; j++ {
-				other.AddNoLock(NewSeriesID(j))
-			}
-
-			clone.RemoveSet(other)
-			b.StopTimer()
-			clone = set.Clone()
-			b.StartTimer()
-		}
-	})
-
-	// This is the case where a new result set is created.
-	b.Run("cardinality_1000000_remove_set_new", func(b *testing.B) {
-		clone := set.Clone()
-		other := NewSeriesIDSet()
-		for j := uint64(0); j < size/2; j++ {
-			other.AddNoLock(NewSeriesID(j))
-		}
-
-		for i := 0; i < b.N; i++ {
-			_ = NewSeriesIDSetNegate(clone, other)
-			b.StopTimer()
-			clone = set.Clone()
-			b.StartTimer()
-		}
-	})
-}
-
-// Typical benchmarks for a laptop:
-//
-// BenchmarkSeriesIDSet_Merge_Duplicates/cardinality_1/shards_1-4         	  200000	      8095 ns/op	   16656 B/op	      11 allocs/op
-// BenchmarkSeriesIDSet_Merge_Duplicates/cardinality_1/shards_10-4        	  200000	     11755 ns/op	   18032 B/op	      47 allocs/op
-// BenchmarkSeriesIDSet_Merge_Duplicates/cardinality_1/shards_100-4       	   50000	     41632 ns/op	   31794 B/op	     407 allocs/op
-// BenchmarkSeriesIDSet_Merge_Duplicates/cardinality_10000/shards_1-4     	  200000	      6022 ns/op	    8384 B/op	       7 allocs/op
-// BenchmarkSeriesIDSet_Merge_Duplicates/cardinality_10000/shards_10-4    	  100000	     19674 ns/op	    9760 B/op	      43 allocs/op
-// BenchmarkSeriesIDSet_Merge_Duplicates/cardinality_10000/shards_100-4   	   10000	    152865 ns/op	   23522 B/op	     403 allocs/op
-// BenchmarkSeriesIDSet_Merge_Duplicates/cardinality_1000000/shards_1-4   	  200000	      8252 ns/op	    9712 B/op	      44 allocs/op
-// BenchmarkSeriesIDSet_Merge_Duplicates/cardinality_1000000/shards_10-4  	   50000	     29566 ns/op	   15984 B/op	     143 allocs/op
-// BenchmarkSeriesIDSet_Merge_Duplicates/cardinality_1000000/shards_100-4 	   10000	    237672 ns/op	   78710 B/op	    1133 allocs/op
-// BenchmarkSeriesIDSet_Merge_Duplicates/cardinality_10000000/shards_1-4  	  100000	     21559 ns/op	   25968 B/op	     330 allocs/op
-// BenchmarkSeriesIDSet_Merge_Duplicates/cardinality_10000000/shards_10-4 	   20000	    102326 ns/op	  114325 B/op	     537 allocs/op
-// BenchmarkSeriesIDSet_Merge_Duplicates/cardinality_10000000/shards_100-4      2000	   1042697 ns/op	  997909 B/op	    2608 allocs/op
-func BenchmarkSeriesIDSet_Merge_Duplicates(b *testing.B) {
-	cardinalities := []int{1, 10000, 1000000, 10000000}
-	shards := []int{1, 10, 100}
-
-	for _, cardinality := range cardinalities {
-		set = NewSeriesIDSet()
-		for i := 0; i < cardinality; i++ {
-			set.Add(NewSeriesID(uint64(i)))
-		}
-
-		for _, shard := range shards {
-			others := make([]*SeriesIDSet, 0, shard)
-			for s := 0; s < shard; s++ {
-				others = append(others, &SeriesIDSet{bitmap: set.bitmap.Clone()})
-			}
-
-			b.Run(fmt.Sprintf("cardinality_%d/shards_%d", cardinality, shard), func(b *testing.B) {
-				base := &SeriesIDSet{bitmap: set.bitmap.Clone()}
-				for i := 0; i < b.N; i++ {
-					base.Merge(others...)
-					b.StopTimer()
-					base.bitmap = set.bitmap.Clone()
-					b.StartTimer()
-				}
-			})
-
-		}
-	}
-}
-
-// Typical benchmarks for a laptop:
-//
-// BenchmarkSeriesIDSet_Merge_Unique/cardinality_1/shards_1-4         	  200000	      7841 ns/op	   16656 B/op	      11 allocs/op
-// BenchmarkSeriesIDSet_Merge_Unique/cardinality_1/shards_10-4        	  200000	     13093 ns/op	   18048 B/op	      47 allocs/op
-// BenchmarkSeriesIDSet_Merge_Unique/cardinality_1/shards_100-4       	   30000	     57399 ns/op	   31985 B/op	     407 allocs/op
-// BenchmarkSeriesIDSet_Merge_Unique/cardinality_10000/shards_1-4     	  200000	      7740 ns/op	    8384 B/op	       7 allocs/op
-// BenchmarkSeriesIDSet_Merge_Unique/cardinality_10000/shards_10-4    	   50000	     37116 ns/op	   18208 B/op	      52 allocs/op
-// BenchmarkSeriesIDSet_Merge_Unique/cardinality_10000/shards_100-4   	    5000	    409487 ns/op	  210563 B/op	     955 allocs/op
-// BenchmarkSeriesIDSet_Merge_Unique/cardinality_1000000/shards_1-4   	  100000	     19289 ns/op	   19328 B/op	      79 allocs/op
-// BenchmarkSeriesIDSet_Merge_Unique/cardinality_1000000/shards_10-4  	   10000	    129048 ns/op	  159716 B/op	     556 allocs/op
-// BenchmarkSeriesIDSet_Merge_Unique/cardinality_1000000/shards_100-4 	     500	   3482907 ns/op	 5428116 B/op	    6174 allocs/op
-// BenchmarkSeriesIDSet_Merge_Unique/cardinality_10000000/shards_1-4  	   30000	     43734 ns/op	   51872 B/op	     641 allocs/op
-// BenchmarkSeriesIDSet_Merge_Unique/cardinality_10000000/shards_10-4 	    3000	    514412 ns/op	  748678 B/op	    3687 allocs/op
-// BenchmarkSeriesIDSet_Merge_Unique/cardinality_10000000/shards_100-4         	      30	  61891687 ns/op	69626539 B/op	   36038 allocs/op
-func BenchmarkSeriesIDSet_Merge_Unique(b *testing.B) {
-	cardinalities := []int{1, 10000, 1000000, 10000000}
-	shards := []int{1, 10, 100}
-
-	for _, cardinality := range cardinalities {
-		set = NewSeriesIDSet()
-		for i := 0; i < cardinality; i++ {
-			set.Add(NewSeriesID(uint64(i)))
-		}
-
-		for _, shard := range shards {
-			others := make([]*SeriesIDSet, 0, shard)
-			for s := 1; s <= shard; s++ {
-				other := NewSeriesIDSet()
-				for i := 0; i < cardinality; i++ {
-					other.Add(NewSeriesID(uint64(i + (s * cardinality))))
-				}
-				others = append(others, other)
-			}
-
-			b.Run(fmt.Sprintf("cardinality_%d/shards_%d", cardinality, shard), func(b *testing.B) {
-				base := &SeriesIDSet{bitmap: set.bitmap.Clone()}
-				for i := 0; i < b.N; i++ {
-					base.Merge(others...)
-					b.StopTimer()
-					base.bitmap = set.bitmap.Clone()
-					b.StartTimer()
-				}
-			})
-		}
-	}
-}

tsdb/seriesfile/config.go

@@ -1,21 +0,0 @@
-package seriesfile
-
-const (
-	// DefaultLargeSeriesWriteThreshold is the number of series per write
-	// that requires the series index be pregrown before insert.
-	DefaultLargeSeriesWriteThreshold = 10000
-)
-
-// Config contains all of the configuration related to tsdb.
-type Config struct {
-	// LargeSeriesWriteThreshold is the threshold before a write requires
-	// preallocation to improve throughput. Currently used in the series file.
-	LargeSeriesWriteThreshold int `toml:"large-series-write-threshold"`
-}
-
-// NewConfig return a new instance of config with default settings.
-func NewConfig() Config {
-	return Config{
-		LargeSeriesWriteThreshold: DefaultLargeSeriesWriteThreshold,
-	}
-}

tsdb/seriesfile/metrics.go

@@ -1,126 +0,0 @@
-package seriesfile
-
-import (
-	"sort"
-	"sync"
-
-	"github.com/influxdata/influxdb/v2/pkg/rhh"
-	"github.com/prometheus/client_golang/prometheus"
-)
-
-// The following package variables act as singletons, to be shared by all
-// storage.Engine instantiations. This allows multiple Series Files to be
-// monitored within the same process.
-var (
-	sms *seriesFileMetrics // main metrics
-	ims *rhh.Metrics       // hashmap specific metrics
-	mmu sync.RWMutex
-)
-
-// PrometheusCollectors returns all the metrics associated with the tsdb package.
-func PrometheusCollectors() []prometheus.Collector {
-	mmu.RLock()
-	defer mmu.RUnlock()
-
-	var collectors []prometheus.Collector
-	if sms != nil {
-		collectors = append(collectors, sms.PrometheusCollectors()...)
-	}
-
-	if ims != nil {
-		collectors = append(collectors, ims.PrometheusCollectors()...)
-	}
-	return collectors
-}
-
-// namespace is the leading part of all published metrics for the Storage service.
-const namespace = "storage"
-
-const seriesFileSubsystem = "series_file" // sub-system associated with metrics for the Series File.
-
-type seriesFileMetrics struct {
-	SeriesCreated *prometheus.CounterVec // Number of series created in Series File.
-	Series        *prometheus.GaugeVec   // Number of series.
-	DiskSize      *prometheus.GaugeVec   // Size occupied on disk.
-	Segments      *prometheus.GaugeVec   // Number of segment files.
-
-	CompactionsActive  *prometheus.GaugeVec     // Number of active compactions.
-	CompactionDuration *prometheus.HistogramVec // Duration of compactions.
-	// The following metrics include a ``"status" = {ok, error}` label
-	Compactions *prometheus.CounterVec // Total number of compactions.
-}
-
-// newSeriesFileMetrics initialises the prometheus metrics for tracking the Series File.
-func newSeriesFileMetrics(labels prometheus.Labels) *seriesFileMetrics {
-	names := []string{"series_file_partition"} // All metrics have this label.
-	for k := range labels {
-		names = append(names, k)
-	}
-	sort.Strings(names)
-
-	totalCompactions := append(append([]string(nil), names...), "status")
-	sort.Strings(totalCompactions)
-
-	durationCompaction := append(append([]string(nil), names...), "component")
-	sort.Strings(durationCompaction)
-
-	return &seriesFileMetrics{
-		SeriesCreated: prometheus.NewCounterVec(prometheus.CounterOpts{
-			Namespace: namespace,
-			Subsystem: seriesFileSubsystem,
-			Name:      "series_created",
-			Help:      "Number of series created in Series File.",
-		}, names),
-		Series: prometheus.NewGaugeVec(prometheus.GaugeOpts{
-			Namespace: namespace,
-			Subsystem: seriesFileSubsystem,
-			Name:      "series_total",
-			Help:      "Number of series in Series File.",
-		}, names),
-		DiskSize: prometheus.NewGaugeVec(prometheus.GaugeOpts{
-			Namespace: namespace,
-			Subsystem: seriesFileSubsystem,
-			Name:      "disk_bytes",
-			Help:      "Number of bytes Series File is using on disk.",
-		}, names),
-		Segments: prometheus.NewGaugeVec(prometheus.GaugeOpts{
-			Namespace: namespace,
-			Subsystem: seriesFileSubsystem,
-			Name:      "segments_total",
-			Help:      "Number of segment files in Series File.",
-		}, names),
-		CompactionsActive: prometheus.NewGaugeVec(prometheus.GaugeOpts{
-			Namespace: namespace,
-			Subsystem: seriesFileSubsystem,
-			Name:      "index_compactions_active",
-			Help:      "Number of active index compactions.",
-		}, durationCompaction),
-		CompactionDuration: prometheus.NewHistogramVec(prometheus.HistogramOpts{
-			Namespace: namespace,
-			Subsystem: seriesFileSubsystem,
-			Name:      "index_compactions_duration_seconds",
-			Help:      "Time taken for a successful compaction of index.",
-			// 30 buckets spaced exponentially between 5s and ~53 minutes.
-			Buckets: prometheus.ExponentialBuckets(5.0, 1.25, 30),
-		}, durationCompaction),
-		Compactions: prometheus.NewCounterVec(prometheus.CounterOpts{
-			Namespace: namespace,
-			Subsystem: seriesFileSubsystem,
-			Name:      "compactions_total",
-			Help:      "Number of compactions.",
-		}, totalCompactions),
-	}
-}
-
-// PrometheusCollectors satisfies the prom.PrometheusCollector interface.
-func (m *seriesFileMetrics) PrometheusCollectors() []prometheus.Collector {
-	return []prometheus.Collector{
-		m.SeriesCreated,
-		m.Series,
-		m.DiskSize,
-		m.Segments,
-		m.CompactionsActive,
-		m.CompactionDuration,
-		m.Compactions,
-	}
-}

tsdb/seriesfile/metrics_test.go

@@ -1,180 +0,0 @@
-package seriesfile
-
-import (
-	"context"
-	"io/ioutil"
-	"os"
-	"testing"
-	"time"
-
-	"github.com/influxdata/influxdb/v2/kit/prom/promtest"
-	"github.com/influxdata/influxdb/v2/models"
-	"github.com/influxdata/influxdb/v2/tsdb"
-	"github.com/prometheus/client_golang/prometheus"
-	dto "github.com/prometheus/client_model/go"
-)
-
-func TestMetrics_SeriesPartition(t *testing.T) {
-	// metrics to be shared by multiple file stores.
-	metrics := newSeriesFileMetrics(prometheus.Labels{"engine_id": "", "node_id": ""})
-
-	t1 := newSeriesPartitionTracker(metrics, prometheus.Labels{"series_file_partition": "0", "engine_id": "0", "node_id": "0"})
-	t2 := newSeriesPartitionTracker(metrics, prometheus.Labels{"series_file_partition": "0", "engine_id": "1", "node_id": "0"})
-
-	reg := prometheus.NewRegistry()
-	reg.MustRegister(metrics.PrometheusCollectors()...)
-
-	base := namespace + "_" + seriesFileSubsystem + "_"
-
-	// All the metric names
-	gauges := []string{
-		base + "series_total",
-		base + "disk_bytes",
-		base + "segments_total",
-		base + "index_compactions_active",
-	}
-
-	counters := []string{
-		base + "series_created",
-		base + "compactions_total",
-	}
-
-	histograms := []string{
-		base + "index_compactions_duration_seconds",
-	}
-
-	// Generate some measurements.
-	for i, tracker := range []*seriesPartitionTracker{t1, t2} {
-		tracker.SetSeries(uint64(i + len(gauges[0])))
-		tracker.SetDiskSize(uint64(i + len(gauges[1])))
-		tracker.SetSegments(uint64(i + len(gauges[2])))
-
-		labels := tracker.Labels()
-		labels["component"] = "index"
-		tracker.metrics.CompactionsActive.With(labels).Add(float64(i + len(gauges[3])))
-
-		tracker.AddSeriesCreated(uint64(i + len(counters[0])))
-		labels = tracker.Labels()
-		labels["status"] = "ok"
-		tracker.metrics.Compactions.With(labels).Add(float64(i + len(counters[1])))
-
-		labels = tracker.Labels()
-		labels["component"] = "index"
-		tracker.metrics.CompactionDuration.With(labels).Observe(float64(i + len(histograms[0])))
-	}
-
-	// Test that all the correct metrics are present.
-	mfs, err := reg.Gather()
-	if err != nil {
-		t.Fatal(err)
-	}
-
-	// The label variants for the two caches.
-	labelVariants := []prometheus.Labels{
-		prometheus.Labels{"engine_id": "0", "node_id": "0"},
-		prometheus.Labels{"engine_id": "1", "node_id": "0"},
-	}
-
-	for i, labels := range labelVariants {
-		labels["series_file_partition"] = "0"
-		var metric *dto.Metric
-
-		for _, name := range gauges {
-			exp := float64(i + len(name))
-
-			if name == base+"index_compactions_active" {
-				// Make a copy since we need to add a label
-				l := make(prometheus.Labels, len(labels))
-				for k, v := range labels {
-					l[k] = v
-				}
-				l["component"] = "index"
-				metric = promtest.MustFindMetric(t, mfs, name, l)
-			} else {
-				metric = promtest.MustFindMetric(t, mfs, name, labels)
-			}
-
-			if got := metric.GetGauge().GetValue(); got != exp {
-				t.Errorf("[%s %d] got %v, expected %v", name, i, got, exp)
-			}
-		}
-
-		for _, name := range counters {
-			exp := float64(i + len(name))
-
-			if name == base+"compactions_total" {
-				// Make a copy since we need to add a label
-				l := make(prometheus.Labels, len(labels))
-				for k, v := range labels {
-					l[k] = v
-				}
-				l["status"] = "ok"
-
-				metric = promtest.MustFindMetric(t, mfs, name, l)
-			} else {
-				metric = promtest.MustFindMetric(t, mfs, name, labels)
-			}
-
-			if got := metric.GetCounter().GetValue(); got != exp {
-				t.Errorf("[%s %d] got %v, expected %v", name, i, got, exp)
-			}
-		}
-
-		for _, name := range histograms {
-			// Make a copy since we need to add a label
-			l := make(prometheus.Labels, len(labels))
-			for k, v := range labels {
-				l[k] = v
-			}
-			l["component"] = "index"
-
-			exp := float64(i + len(name))
-			metric := promtest.MustFindMetric(t, mfs, name, l)
-			if got := metric.GetHistogram().GetSampleSum(); got != exp {
-				t.Errorf("[%s %d] got %v, expected %v", name, i, got, exp)
-			}
-		}
-	}
-}
-
-// This test ensures that disabling metrics works even if a series file has been created before.
-func TestMetrics_Disabled(t *testing.T) {
-	// This test messes with global state. Gotta fix it up otherwise other tests panic. I really
-	// am beginning to wonder about our metrics.
-	defer func() {
-		mmu.Lock()
-		sms = nil
-		ims = nil
-		mmu.Unlock()
-	}()
-
-	path, err := ioutil.TempDir("", "sfile-metrics-")
-	if err != nil {
-		t.Fatal(err)
-	}
-	defer os.RemoveAll(path)
-
-	// Step 1. make a series file with metrics and some labels
-	sfile := NewSeriesFile(path)
-	sfile.SetDefaultMetricLabels(prometheus.Labels{"foo": "bar"})
-	if err := sfile.Open(context.Background()); err != nil {
-		t.Fatal(err)
-	}
-	if err := sfile.Close(); err != nil {
-		t.Fatal(err)
-	}
-
-	// Step 2. open the series file again, but disable metrics
-	sfile = NewSeriesFile(path)
-	sfile.DisableMetrics()
-	if err := sfile.Open(context.Background()); err != nil {
-		t.Fatal(err)
-	}
-	defer sfile.Close()
-
-	// Step 3. add a series
-	points := []models.Point{models.MustNewPoint("a", models.Tags{}, models.Fields{"f": 1.0}, time.Now())}
-	if err := sfile.CreateSeriesListIfNotExists(tsdb.NewSeriesCollection(points)); err != nil {
-		t.Fatal(err)
-	}
-}

tsdb/seriesfile/series_file.go

@@ -1,628 +0,0 @@
-package seriesfile
-
-import (
-	"bytes"
-	"context"
-	"encoding/binary"
-	"errors"
-	"fmt"
-	"os"
-	"path/filepath"
-	"sort"
-	"sync"
-
-	"github.com/cespare/xxhash"
-	"github.com/influxdata/influxdb/v2/kit/tracing"
-	"github.com/influxdata/influxdb/v2/logger"
-	"github.com/influxdata/influxdb/v2/models"
-	"github.com/influxdata/influxdb/v2/pkg/binaryutil"
-	"github.com/influxdata/influxdb/v2/pkg/lifecycle"
-	"github.com/influxdata/influxdb/v2/pkg/mincore"
-	"github.com/influxdata/influxdb/v2/pkg/rhh"
-	"github.com/influxdata/influxdb/v2/tsdb"
-	"github.com/prometheus/client_golang/prometheus"
-	"go.uber.org/multierr"
-	"go.uber.org/zap"
-	"golang.org/x/sync/errgroup"
-	"golang.org/x/time/rate"
-)
-
-var (
-	ErrSeriesFileClosed         = errors.New("tsdb: series file closed")
-	ErrInvalidSeriesPartitionID = errors.New("tsdb: invalid series partition id")
-)
-
-const (
-	// SeriesFilePartitionN is the number of partitions a series file is split into.
-	SeriesFilePartitionN = 8
-)
-
-// SeriesFile represents the section of the index that holds series data.
-type SeriesFile struct {
-	mu  sync.Mutex // protects concurrent open and close
-	res lifecycle.Resource
-
-	path       string
-	partitions []*SeriesPartition
-
-	// N.B we have many partitions, but they must share the same metrics, so the
-	// metrics are managed in a single shared package variable and
-	// each partition decorates the same metric measurements with different
-	// partition id label values.
-	defaultMetricLabels prometheus.Labels
-	metricsEnabled      bool
-
-	pageFaultLimiter *rate.Limiter // Limits page faults by the series file
-
-	LargeWriteThreshold int
-
-	Logger *zap.Logger
-}
-
-// NewSeriesFile returns a new instance of SeriesFile.
-func NewSeriesFile(path string) *SeriesFile {
-	return &SeriesFile{
-		path:           path,
-		metricsEnabled: true,
-		Logger:         zap.NewNop(),
-
-		LargeWriteThreshold: DefaultLargeSeriesWriteThreshold,
-	}
-}
-
-// WithLogger sets the logger on the SeriesFile and all underlying partitions. It must be called before Open.
-func (f *SeriesFile) WithLogger(log *zap.Logger) {
-	f.Logger = log.With(zap.String("service", "series-file"))
-}
-
-// SetDefaultMetricLabels sets the default labels for metrics on the Series File.
-// It must be called before the SeriesFile is opened.
-func (f *SeriesFile) SetDefaultMetricLabels(labels prometheus.Labels) {
-	f.defaultMetricLabels = make(prometheus.Labels, len(labels))
-	for k, v := range labels {
-		f.defaultMetricLabels[k] = v
-	}
-}
-
-// DisableMetrics ensures that activity is not collected via the prometheus metrics.
-// DisableMetrics must be called before Open.
-func (f *SeriesFile) DisableMetrics() {
-	f.metricsEnabled = false
-}
-
-// WithPageFaultLimiter sets a limiter to restrict the number of page faults.
-func (f *SeriesFile) WithPageFaultLimiter(limiter *rate.Limiter) {
-	f.pageFaultLimiter = limiter
-}
-
-// Open memory maps the data file at the file's path.
-func (f *SeriesFile) Open(ctx context.Context) error {
-	f.mu.Lock()
-	defer f.mu.Unlock()
-
-	if f.res.Opened() {
-		return errors.New("series file already opened")
-	}
-
-	span, ctx := tracing.StartSpanFromContext(ctx)
-	defer span.Finish()
-
-	_, logEnd := logger.NewOperation(ctx, f.Logger, "Opening Series File", "series_file_open", zap.String("path", f.path))
-	defer logEnd()
-
-	// Create path if it doesn't exist.
-	if err := os.MkdirAll(filepath.Join(f.path), 0777); err != nil {
-		return err
-	}
-
-	// Initialise metrics for trackers.
-	mmu.Lock()
-	if sms == nil && f.metricsEnabled {
-		sms = newSeriesFileMetrics(f.defaultMetricLabels)
-	}
-	if ims == nil && f.metricsEnabled {
-		// Make a copy of the default labels so that another label can be provided.
-		labels := make(prometheus.Labels, len(f.defaultMetricLabels))
-		for k, v := range f.defaultMetricLabels {
-			labels[k] = v
-		}
-		labels["series_file_partition"] = "" // All partitions have this label.
-		ims = rhh.NewMetrics(namespace, seriesFileSubsystem+"_index", labels)
-	}
-	mmu.Unlock()
-
-	// Open partitions.
-	f.partitions = make([]*SeriesPartition, 0, SeriesFilePartitionN)
-	for i := 0; i < SeriesFilePartitionN; i++ {
-		// TODO(edd): These partition initialisation should be moved up to NewSeriesFile.
-		p := NewSeriesPartition(i, f.SeriesPartitionPath(i))
-		p.LargeWriteThreshold = f.LargeWriteThreshold
-		p.Logger = f.Logger.With(zap.Int("partition", p.ID()))
-		p.pageFaultLimiter = f.pageFaultLimiter
-
-		// For each series file index, rhh trackers are used to track the RHH Hashmap.
-		// Each of the trackers needs to be given slightly different default
-		// labels to ensure the correct partition_ids are set as labels.
-		labels := make(prometheus.Labels, len(f.defaultMetricLabels))
-		for k, v := range f.defaultMetricLabels {
-			labels[k] = v
-		}
-		labels["series_file_partition"] = fmt.Sprint(p.ID())
-
-		p.index.rhhMetrics = ims
-		p.index.rhhLabels = labels
-		p.index.rhhMetricsEnabled = f.metricsEnabled
-
-		// Set the metric trackers on the partition with any injected default labels.
-		p.tracker = newSeriesPartitionTracker(sms, labels)
-		p.tracker.enabled = f.metricsEnabled
-
-		if err := p.Open(); err != nil {
-			f.Logger.Error("Unable to open series file",
-				zap.String("path", f.path),
-				zap.Int("partition", p.ID()),
-				zap.Error(err))
-			f.closeNoLock()
-			return err
-		}
-		f.partitions = append(f.partitions, p)
-	}
-
-	// The resource is now open.
-	f.res.Open()
-
-	return nil
-}
-
-func (f *SeriesFile) closeNoLock() (err error) {
-	// Close the resource and wait for any outstanding references.
-	f.res.Close()
-
-	var errs []error
-	for _, p := range f.partitions {
-		errs = append(errs, p.Close())
-	}
-	return multierr.Combine(errs...)
-}
-
-// Close unmaps the data file.
-func (f *SeriesFile) Close() error {
-	f.mu.Lock()
-	defer f.mu.Unlock()
-	return f.closeNoLock()
-}
-
-// Path returns the path to the file.
-func (f *SeriesFile) Path() string { return f.path }
-
-// SeriesPartitionPath returns the path to a given partition.
-func (f *SeriesFile) SeriesPartitionPath(i int) string {
-	return filepath.Join(f.path, fmt.Sprintf("%02x", i))
-}
-
-// Partitions returns all partitions.
-func (f *SeriesFile) Partitions() []*SeriesPartition { return f.partitions }
-
-// Acquire ensures that the series file won't be closed until after the reference
-// has been released.
-func (f *SeriesFile) Acquire() (*lifecycle.Reference, error) {
-	return f.res.Acquire()
-}
-
-// EnableCompactions allows compactions to run.
-func (f *SeriesFile) EnableCompactions() {
-	for _, p := range f.partitions {
-		p.EnableCompactions()
-	}
-}
-
-// DisableCompactions prevents new compactions from running.
-func (f *SeriesFile) DisableCompactions() {
-	for _, p := range f.partitions {
-		p.DisableCompactions()
-	}
-}
-
-// FileSize returns the size of all partitions, in bytes.
-func (f *SeriesFile) FileSize() (n int64, err error) {
-	for _, p := range f.partitions {
-		v, err := p.FileSize()
-		n += v
-		if err != nil {
-			return n, err
-		}
-	}
-	return n, err
-}
-
-// CreateSeriesListIfNotExists creates a list of series in bulk if they don't exist. It overwrites
-// the collection's Keys and SeriesIDs fields. The collection's SeriesIDs slice will have IDs for
-// every name+tags, creating new series IDs as needed. If any SeriesID is zero, then a type
-// conflict has occurred for that series.
-func (f *SeriesFile) CreateSeriesListIfNotExists(collection *tsdb.SeriesCollection) error {
-	collection.SeriesKeys = GenerateSeriesKeys(collection.Names, collection.Tags)
-	collection.SeriesIDs = make([]tsdb.SeriesID, len(collection.SeriesKeys))
-	keyPartitionIDs := f.SeriesKeysPartitionIDs(collection.SeriesKeys)
-
-	var g errgroup.Group
-	for i := range f.partitions {
-		p := f.partitions[i]
-		g.Go(func() error {
-			return p.CreateSeriesListIfNotExists(collection, keyPartitionIDs)
-		})
-	}
-	if err := g.Wait(); err != nil {
-		return err
-	}
-
-	collection.ApplyConcurrentDrops()
-	return nil
-}
-
-// DeleteSeriesID flags a list of series as permanently deleted.
-// If a series is reintroduced later then it must create a new id.
-func (f *SeriesFile) DeleteSeriesIDs(ids []tsdb.SeriesID) error {
-	m := make(map[int][]tsdb.SeriesID)
-	for _, id := range ids {
-		partitionID := f.SeriesIDPartitionID(id)
-		m[partitionID] = append(m[partitionID], id)
-	}
-
-	var g errgroup.Group
-	for partitionID, partitionIDs := range m {
-		partitionID, partitionIDs := partitionID, partitionIDs
-		g.Go(func() error { return f.partitions[partitionID].DeleteSeriesIDs(partitionIDs) })
-	}
-	return g.Wait()
-}
-
-// IsDeleted returns true if the ID has been deleted before.
-func (f *SeriesFile) IsDeleted(id tsdb.SeriesID) bool {
-	p := f.SeriesIDPartition(id)
-	if p == nil {
-		return false
-	}
-	return p.IsDeleted(id)
-}
-
-// SeriesKey returns the series key for a given id.
-func (f *SeriesFile) SeriesKey(id tsdb.SeriesID) []byte {
-	if id.IsZero() {
-		return nil
-	}
-	p := f.SeriesIDPartition(id)
-	if p == nil {
-		return nil
-	}
-	return p.SeriesKey(id)
-}
-
-// SeriesKeyName returns the measurement name for a series id.
-func (f *SeriesFile) SeriesKeyName(id tsdb.SeriesID) []byte {
-	if id.IsZero() {
-		return nil
-	}
-	data := f.SeriesIDPartition(id).SeriesKey(id)
-	if data == nil {
-		return nil
-	}
-	_, data = ReadSeriesKeyLen(data)
-	name, _ := ReadSeriesKeyMeasurement(data)
-	return name
-}
-
-// SeriesKeys returns a list of series keys from a list of ids.
-func (f *SeriesFile) SeriesKeys(ids []tsdb.SeriesID) [][]byte {
-	keys := make([][]byte, len(ids))
-	for i := range ids {
-		keys[i] = f.SeriesKey(ids[i])
-	}
-	return keys
-}
-
-// Series returns the parsed series name and tags for an offset.
-func (f *SeriesFile) Series(id tsdb.SeriesID) ([]byte, models.Tags) {
-	key := f.SeriesKey(id)
-	if key == nil {
-		return nil, nil
-	}
-	return ParseSeriesKey(key)
-}
-
-// SeriesID returns the series id for the series.
-func (f *SeriesFile) SeriesID(name []byte, tags models.Tags, buf []byte) tsdb.SeriesID {
-	return f.SeriesIDTyped(name, tags, buf).SeriesID()
-}
-
-// SeriesIDTyped returns the typed series id for the series.
-func (f *SeriesFile) SeriesIDTyped(name []byte, tags models.Tags, buf []byte) tsdb.SeriesIDTyped {
-	key := AppendSeriesKey(buf[:0], name, tags)
-	return f.SeriesIDTypedBySeriesKey(key)
-}
-
-// SeriesIDTypedBySeriesKey returns the typed series id for the series.
-func (f *SeriesFile) SeriesIDTypedBySeriesKey(key []byte) tsdb.SeriesIDTyped {
-	keyPartition := f.SeriesKeyPartition(key)
-	if keyPartition == nil {
-		return tsdb.SeriesIDTyped{}
-	}
-	return keyPartition.FindIDTypedBySeriesKey(key)
-}
-
-// HasSeries return true if the series exists.
-func (f *SeriesFile) HasSeries(name []byte, tags models.Tags, buf []byte) bool {
-	return !f.SeriesID(name, tags, buf).IsZero()
-}
-
-// SeriesCount returns the number of series.
-func (f *SeriesFile) SeriesCount() uint64 {
-	var n uint64
-	for _, p := range f.partitions {
-		n += p.SeriesCount()
-	}
-	return n
-}
-
-// SeriesIDs returns a slice of series IDs in all partitions, sorted.
-// This may return a lot of data at once, so use sparingly.
-func (f *SeriesFile) SeriesIDs() []tsdb.SeriesID {
-	var ids []tsdb.SeriesID
-	for _, p := range f.partitions {
-		ids = p.AppendSeriesIDs(ids)
-	}
-	sort.Slice(ids, func(i, j int) bool { return ids[i].Less(ids[j]) })
-	return ids
-}
-
-func (f *SeriesFile) SeriesIDPartitionID(id tsdb.SeriesID) int {
-	return int((id.RawID() - 1) % SeriesFilePartitionN)
-}
-
-func (f *SeriesFile) SeriesIDPartition(id tsdb.SeriesID) *SeriesPartition {
-	partitionID := f.SeriesIDPartitionID(id)
-	if partitionID >= len(f.partitions) {
-		return nil
-	}
-	return f.partitions[partitionID]
-}
-
-func (f *SeriesFile) SeriesKeysPartitionIDs(keys [][]byte) []int {
-	partitionIDs := make([]int, len(keys))
-	for i := range keys {
-		partitionIDs[i] = f.SeriesKeyPartitionID(keys[i])
-	}
-	return partitionIDs
-}
-
-func (f *SeriesFile) SeriesKeyPartitionID(key []byte) int {
-	return int(xxhash.Sum64(key) % SeriesFilePartitionN)
-}
-
-func (f *SeriesFile) SeriesKeyPartition(key []byte) *SeriesPartition {
-	partitionID := f.SeriesKeyPartitionID(key)
-	if partitionID >= len(f.partitions) {
-		return nil
-	}
-	return f.partitions[partitionID]
-}
-
-// AppendSeriesKey serializes name and tags to a byte slice.
-// The total length is prepended as a uvarint.
-func AppendSeriesKey(dst []byte, name []byte, tags models.Tags) []byte {
-	buf := make([]byte, binary.MaxVarintLen64)
-	origLen := len(dst)
-
-	// The tag count is variable encoded, so we need to know ahead of time what
-	// the size of the tag count value will be.
-	tcBuf := make([]byte, binary.MaxVarintLen64)
-	tcSz := binary.PutUvarint(tcBuf, uint64(len(tags)))
-
-	// Size of name/tags. Does not include total length.
-	size := 0 + //
-		2 + // size of measurement
-		len(name) + // measurement
-		tcSz + // size of number of tags
-		(4 * len(tags)) + // length of each tag key and value
-		tags.Size() // size of tag keys/values
-
-	// Variable encode length.
-	totalSz := binary.PutUvarint(buf, uint64(size))
-
-	// If caller doesn't provide a buffer then pre-allocate an exact one.
-	if dst == nil {
-		dst = make([]byte, 0, size+totalSz)
-	}
-
-	// Append total length.
-	dst = append(dst, buf[:totalSz]...)
-
-	// Append name.
-	binary.BigEndian.PutUint16(buf, uint16(len(name)))
-	dst = append(dst, buf[:2]...)
-	dst = append(dst, name...)
-
-	// Append tag count.
-	dst = append(dst, tcBuf[:tcSz]...)
-
-	// Append tags.
-	for _, tag := range tags {
-		binary.BigEndian.PutUint16(buf, uint16(len(tag.Key)))
-		dst = append(dst, buf[:2]...)
-		dst = append(dst, tag.Key...)
-
-		binary.BigEndian.PutUint16(buf, uint16(len(tag.Value)))
-		dst = append(dst, buf[:2]...)
-		dst = append(dst, tag.Value...)
-	}
-
-	// Verify that the total length equals the encoded byte count.
-	if got, exp := len(dst)-origLen, size+totalSz; got != exp {
-		panic(fmt.Sprintf("series key encoding does not match calculated total length: actual=%d, exp=%d, key=%x", got, exp, dst))
-	}
-
-	return dst
-}
-
-// ReadSeriesKey returns the series key from the beginning of the buffer.
-func ReadSeriesKey(data []byte) (key, remainder []byte) {
-	sz, n := binary.Uvarint(data)
-	return data[:int(sz)+n], data[int(sz)+n:]
-}
-
-func ReadSeriesKeyLen(data []byte) (sz int, remainder []byte) {
-	sz64, i := binary.Uvarint(data)
-	return int(sz64), data[i:]
-}
-
-func ReadSeriesKeyMeasurement(data []byte) (name, remainder []byte) {
-	n, data := binary.BigEndian.Uint16(data), data[2:]
-	return data[:n], data[n:]
-}
-
-func ReadSeriesKeyTagN(data []byte) (n int, remainder []byte) {
-	n64, i := binary.Uvarint(data)
-	return int(n64), data[i:]
-}
-
-func ReadSeriesKeyTag(data []byte) (key, value, remainder []byte) {
-	n, data := binary.BigEndian.Uint16(data), data[2:]
-	key, data = data[:n], data[n:]
-
-	n, data = binary.BigEndian.Uint16(data), data[2:]
-	value, data = data[:n], data[n:]
-	return key, value, data
-}
-
-// ParseSeriesKey extracts the name & tags from a series key.
-func ParseSeriesKey(data []byte) (name []byte, tags models.Tags) {
-	return parseSeriesKey(data, nil)
-}
-
-// ParseSeriesKeyInto extracts the name and tags for data, parsing the tags into
-// dstTags, which is then returened.
-//
-// The returned dstTags may have a different length and capacity.
-func ParseSeriesKeyInto(data []byte, dstTags models.Tags) ([]byte, models.Tags) {
-	return parseSeriesKey(data, dstTags)
-}
-
-// parseSeriesKey extracts the name and tags from data, attempting to re-use the
-// provided tags value rather than allocating. The returned tags may have a
-// different length and capacity to those provided.
-func parseSeriesKey(data []byte, dst models.Tags) ([]byte, models.Tags) {
-	var name []byte
-	_, data = ReadSeriesKeyLen(data)
-	name, data = ReadSeriesKeyMeasurement(data)
-	tagN, data := ReadSeriesKeyTagN(data)
-
-	dst = dst[:cap(dst)] // Grow dst to use full capacity
-	if got, want := len(dst), tagN; got < want {
-		dst = append(dst, make(models.Tags, want-got)...)
-	} else if got > want {
-		dst = dst[:want]
-	}
-	dst = dst[:tagN]
-
-	for i := 0; i < tagN; i++ {
-		var key, value []byte
-		key, value, data = ReadSeriesKeyTag(data)
-		dst[i].Key, dst[i].Value = key, value
-	}
-
-	return name, dst
-}
-
-func CompareSeriesKeys(a, b []byte) int {
-	// Handle 'nil' keys.
-	if len(a) == 0 && len(b) == 0 {
-		return 0
-	} else if len(a) == 0 {
-		return -1
-	} else if len(b) == 0 {
-		return 1
-	}
-
-	// Read total size.
-	_, a = ReadSeriesKeyLen(a)
-	_, b = ReadSeriesKeyLen(b)
-
-	// Read names.
-	name0, a := ReadSeriesKeyMeasurement(a)
-	name1, b := ReadSeriesKeyMeasurement(b)
-
-	// Compare names, return if not equal.
-	if cmp := bytes.Compare(name0, name1); cmp != 0 {
-		return cmp
-	}
-
-	// Read tag counts.
-	tagN0, a := ReadSeriesKeyTagN(a)
-	tagN1, b := ReadSeriesKeyTagN(b)
-
-	// Compare each tag in order.
-	for i := 0; ; i++ {
-		// Check for EOF.
-		if i == tagN0 && i == tagN1 {
-			return 0
-		} else if i == tagN0 {
-			return -1
-		} else if i == tagN1 {
-			return 1
-		}
-
-		// Read keys.
-		var key0, key1, value0, value1 []byte
-		key0, value0, a = ReadSeriesKeyTag(a)
-		key1, value1, b = ReadSeriesKeyTag(b)
-
-		// Compare keys & values.
-		if cmp := bytes.Compare(key0, key1); cmp != 0 {
-			return cmp
-		} else if cmp := bytes.Compare(value0, value1); cmp != 0 {
-			return cmp
-		}
-	}
-}
-
-// GenerateSeriesKeys generates series keys for a list of names & tags using
-// a single large memory block.
-func GenerateSeriesKeys(names [][]byte, tagsSlice []models.Tags) [][]byte {
-	buf := make([]byte, 0, SeriesKeysSize(names, tagsSlice))
-	keys := make([][]byte, len(names))
-	for i := range names {
-		offset := len(buf)
-		buf = AppendSeriesKey(buf, names[i], tagsSlice[i])
-		keys[i] = buf[offset:]
-	}
-	return keys
-}
-
-// SeriesKeysSize returns the number of bytes required to encode a list of name/tags.
-func SeriesKeysSize(names [][]byte, tagsSlice []models.Tags) int {
-	var n int
-	for i := range names {
-		n += SeriesKeySize(names[i], tagsSlice[i])
-	}
-	return n
-}
-
-// SeriesKeySize returns the number of bytes required to encode a series key.
-func SeriesKeySize(name []byte, tags models.Tags) int {
-	var n int
-	n += 2 + len(name)
-	n += binaryutil.UvarintSize(uint64(len(tags)))
-	for _, tag := range tags {
-		n += 2 + len(tag.Key)
-		n += 2 + len(tag.Value)
-	}
-	n += binaryutil.UvarintSize(uint64(n))
-	return n
-}
-
-// wait rate limits page faults to the underlying data. Skipped if limiter is not set.
-func wait(limiter *mincore.Limiter, b []byte) error {
-	if limiter == nil {
-		return nil
-	}
-	return limiter.WaitRange(context.Background(), b)
-}

tsdb/seriesfile/series_file_test.go

@@ -1,494 +0,0 @@
-package seriesfile_test
-
-import (
-	"bytes"
-	"context"
-	"fmt"
-	"io/ioutil"
-	"os"
-	"path"
-	"testing"
-
-	"github.com/influxdata/influxdb/v2/logger"
-	"github.com/influxdata/influxdb/v2/models"
-	"github.com/influxdata/influxdb/v2/tsdb"
-	"github.com/influxdata/influxdb/v2/tsdb/seriesfile"
-	"golang.org/x/sync/errgroup"
-)
-
-func TestParseSeriesKeyInto(t *testing.T) {
-	name := []byte("cpu")
-	tags := models.NewTags(map[string]string{"region": "east", "server": "a"})
-	key := seriesfile.AppendSeriesKey(nil, name, tags)
-
-	dst := make(models.Tags, 0)
-	gotName, gotTags := seriesfile.ParseSeriesKeyInto(key, dst)
-
-	if !bytes.Equal(gotName, name) {
-		t.Fatalf("got %q, expected %q", gotName, name)
-	}
-
-	if got, exp := len(gotTags), 2; got != exp {
-		t.Fatalf("got tags length %d, expected %d", got, exp)
-	} else if got, exp := gotTags, tags; !got.Equal(exp) {
-		t.Fatalf("got tags %v, expected %v", got, exp)
-	}
-
-	dst = make(models.Tags, 0, 5)
-	_, gotTags = seriesfile.ParseSeriesKeyInto(key, dst)
-	if got, exp := len(gotTags), 2; got != exp {
-		t.Fatalf("got tags length %d, expected %d", got, exp)
-	} else if got, exp := cap(gotTags), 5; got != exp {
-		t.Fatalf("got tags capacity %d, expected %d", got, exp)
-	} else if got, exp := gotTags, tags; !got.Equal(exp) {
-		t.Fatalf("got tags %v, expected %v", got, exp)
-	}
-
-	dst = make(models.Tags, 1)
-	_, gotTags = seriesfile.ParseSeriesKeyInto(key, dst)
-	if got, exp := len(gotTags), 2; got != exp {
-		t.Fatalf("got tags length %d, expected %d", got, exp)
-	} else if got, exp := gotTags, tags; !got.Equal(exp) {
-		t.Fatalf("got tags %v, expected %v", got, exp)
-	}
-}
-
-// Ensure that broken series files are closed
-func TestSeriesFile_Open_WhenFileCorrupt_ShouldReturnErr(t *testing.T) {
-	f := NewBrokenSeriesFile([]byte{0, 0, 0, 0, 0})
-	defer f.Close()
-	f.Logger = logger.New(os.Stdout)
-
-	err := f.Open(context.Background())
-	if err == nil {
-		t.Fatalf("should report error")
-	}
-}
-
-// Ensure series file contains the correct set of series.
-func TestSeriesFile_Series(t *testing.T) {
-	sfile := MustOpenSeriesFile()
-	defer sfile.Close()
-
-	series := []Series{
-		{Name: []byte("cpu"), Tags: models.NewTags(map[string]string{"region": "east"}), Type: models.Integer},
-		{Name: []byte("cpu"), Tags: models.NewTags(map[string]string{"region": "west"}), Type: models.Integer},
-		{Name: []byte("mem"), Tags: models.NewTags(map[string]string{"region": "east"}), Type: models.Integer},
-	}
-	for _, s := range series {
-		collection := &tsdb.SeriesCollection{
-			Names: [][]byte{[]byte(s.Name)},
-			Tags:  []models.Tags{s.Tags},
-			Types: []models.FieldType{s.Type},
-		}
-		if err := sfile.CreateSeriesListIfNotExists(collection); err != nil {
-			t.Fatal(err)
-		}
-	}
-
-	// Verify total number of series is correct.
-	if n := sfile.SeriesCount(); n != 3 {
-		t.Fatalf("unexpected series count: %d", n)
-	}
-
-	// Verify all series exist.
-	for i, s := range series {
-		if seriesID := sfile.SeriesID(s.Name, s.Tags, nil); seriesID.IsZero() {
-			t.Fatalf("series does not exist: i=%d", i)
-		}
-	}
-
-	// Verify non-existent series doesn't exist.
-	if sfile.HasSeries([]byte("foo"), models.NewTags(map[string]string{"region": "north"}), nil) {
-		t.Fatal("series should not exist")
-	}
-}
-
-// Ensure series file can be compacted.
-func TestSeriesFileCompactor(t *testing.T) {
-	sfile := MustOpenSeriesFile()
-	defer sfile.Close()
-
-	// Disable automatic compactions.
-	for _, p := range sfile.Partitions() {
-		p.CompactThreshold = 0
-	}
-
-	collection := new(tsdb.SeriesCollection)
-	for i := 0; i < 10000; i++ {
-		collection.Names = append(collection.Names, []byte(fmt.Sprintf("m%d", i)))
-		collection.Tags = append(collection.Tags, models.NewTags(map[string]string{"foo": "bar"}))
-		collection.Types = append(collection.Types, models.Integer)
-	}
-	if err := sfile.CreateSeriesListIfNotExists(collection); err != nil {
-		t.Fatal(err)
-	}
-	if err := collection.PartialWriteError(); err != nil {
-		t.Fatal(err)
-	}
-
-	// Verify total number of series is correct.
-	if n := sfile.SeriesCount(); n != uint64(len(collection.Names)) {
-		t.Fatalf("unexpected series count: %d", n)
-	}
-
-	// Compact in-place for each partition.
-	for _, p := range sfile.Partitions() {
-		compactor := seriesfile.NewSeriesPartitionCompactor()
-		if _, err := compactor.Compact(p); err != nil {
-			t.Fatal(err)
-		}
-	}
-
-	// Verify all series exist.
-	for iter := collection.Iterator(); iter.Next(); {
-		if seriesID := sfile.SeriesID(iter.Name(), iter.Tags(), nil); seriesID.IsZero() {
-			t.Fatalf("series does not exist: %s,%s", iter.Name(), iter.Tags().String())
-		}
-	}
-
-	// Verify total number of series is correct.
-	if got, exp := sfile.SeriesCount(), uint64(len(collection.Names)); got != exp {
-		t.Fatalf("SeriesCount()=%d, expected %d (after compaction)", got, exp)
-	}
-}
-
-// Ensures that types are tracked and checked by the series file.
-func TestSeriesFile_Type(t *testing.T) {
-	sfile := MustOpenSeriesFile()
-	defer sfile.Close()
-
-	// Add the series with some types.
-	collection := &tsdb.SeriesCollection{
-		Names: [][]byte{[]byte("a"), []byte("b"), []byte("c")},
-		Tags:  []models.Tags{{}, {}, {}},
-		Types: []models.FieldType{models.Integer, models.Float, models.Boolean},
-	}
-	if err := sfile.CreateSeriesListIfNotExists(collection); err != nil {
-		t.Fatal(err)
-	}
-
-	// Attempt to add the series again but with different types.
-	collection = &tsdb.SeriesCollection{
-		Names: [][]byte{[]byte("a"), []byte("b"), []byte("c"), []byte("d")},
-		Tags:  []models.Tags{{}, {}, {}, {}},
-		Types: []models.FieldType{models.String, models.String, models.String, models.String},
-	}
-	if err := sfile.CreateSeriesListIfNotExists(collection); err != nil {
-		t.Fatal(err)
-	}
-
-	// All of the series except d should be dropped.
-	if err := collection.PartialWriteError(); err == nil {
-		t.Fatal("expected partial write error")
-	}
-	if collection.Length() != 1 {
-		t.Fatal("expected one series to remain in collection")
-	}
-	if got := string(collection.Names[0]); got != "d" {
-		t.Fatal("got invalid name on remaining series:", got)
-	}
-}
-
-// Ensure series file deletions persist across compactions.
-func TestSeriesFile_DeleteSeriesID(t *testing.T) {
-	sfile := MustOpenSeriesFile()
-	defer sfile.Close()
-
-	if err := sfile.CreateSeriesListIfNotExists(&tsdb.SeriesCollection{
-		Names: [][]byte{[]byte("m1")},
-		Tags:  []models.Tags{{}},
-		Types: []models.FieldType{models.String},
-	}); err != nil {
-		t.Fatal(err)
-	} else if err := sfile.CreateSeriesListIfNotExists(&tsdb.SeriesCollection{
-		Names: [][]byte{[]byte("m2")},
-		Tags:  []models.Tags{{}},
-		Types: []models.FieldType{models.String},
-	}); err != nil {
-		t.Fatal(err)
-	} else if err := sfile.ForceCompact(); err != nil {
-		t.Fatal(err)
-	}
-	id := sfile.SeriesID([]byte("m1"), nil, nil)
-
-	// Verify total number of series is correct.
-	if got, exp := sfile.SeriesCount(), uint64(2); got != exp {
-		t.Fatalf("SeriesCount()=%d, expected %d (before deleted)", got, exp)
-	}
-
-	// Delete and ensure deletion.
-	if err := sfile.DeleteSeriesIDs([]tsdb.SeriesID{id}); err != nil {
-		t.Fatal(err)
-	} else if !sfile.IsDeleted(id) {
-		t.Fatal("expected deletion before compaction")
-	}
-
-	// Verify total number of series is correct.
-	if got, exp := sfile.SeriesCount(), uint64(1); got != exp {
-		t.Fatalf("SeriesCount()=%d, expected %d (before compaction)", got, exp)
-	}
-
-	if err := sfile.ForceCompact(); err != nil {
-		t.Fatal(err)
-	} else if !sfile.IsDeleted(id) {
-		t.Fatal("expected deletion after compaction")
-	} else if got, exp := sfile.SeriesCount(), uint64(1); got != exp {
-		t.Fatalf("SeriesCount()=%d, expected %d (after compaction)", got, exp)
-	}
-
-	if err := sfile.Reopen(); err != nil {
-		t.Fatal(err)
-	} else if !sfile.IsDeleted(id) {
-		t.Fatal("expected deletion after reopen")
-	} else if got, exp := sfile.SeriesCount(), uint64(1); got != exp {
-		t.Fatalf("SeriesCount()=%d, expected %d (after reopen)", got, exp)
-	}
-
-	// Recreate series with new ID.
-	if err := sfile.CreateSeriesListIfNotExists(&tsdb.SeriesCollection{
-		Names: [][]byte{[]byte("m1")},
-		Tags:  []models.Tags{{}},
-		Types: []models.FieldType{models.String},
-	}); err != nil {
-		t.Fatal(err)
-	} else if got, exp := sfile.SeriesCount(), uint64(2); got != exp {
-		t.Fatalf("SeriesCount()=%d, expected %d (after recreate)", got, exp)
-	}
-
-	if err := sfile.ForceCompact(); err != nil {
-		t.Fatal(err)
-	} else if !sfile.IsDeleted(id) {
-		t.Fatal("expected deletion after compaction")
-	} else if got, exp := sfile.SeriesCount(), uint64(2); got != exp {
-		t.Fatalf("SeriesCount()=%d, expected %d (after recreate & compaction)", got, exp)
-	}
-
-	if err := sfile.Reopen(); err != nil {
-		t.Fatal(err)
-	} else if !sfile.IsDeleted(id) {
-		t.Fatal("expected deletion after reopen")
-	} else if got, exp := sfile.SeriesCount(), uint64(2); got != exp {
-		t.Fatalf("SeriesCount()=%d, expected %d (after recreate & compaction)", got, exp)
-	}
-}
-
-func TestSeriesFile_Compaction(t *testing.T) {
-	const n = 1000
-
-	sfile := MustOpenSeriesFile()
-	defer sfile.Close()
-
-	// Generate a bunch of keys.
-	var collection tsdb.SeriesCollection
-	for i := 0; i < n; i++ {
-		collection.Names = append(collection.Names, []byte("cpu"))
-		collection.Tags = append(collection.Tags, models.NewTags(map[string]string{"region": fmt.Sprintf("r%d", i)}))
-		collection.Types = append(collection.Types, models.Integer)
-	}
-
-	// Add all to the series file.
-	err := sfile.CreateSeriesListIfNotExists(&collection)
-	if err != nil {
-		t.Fatal(err)
-	}
-
-	// Delete a subset of keys.
-	for i := 0; i < n; i++ {
-		if i%10 != 0 {
-			continue
-		}
-
-		if id := sfile.SeriesID(collection.Names[i], collection.Tags[i], nil); id.IsZero() {
-			t.Fatal("expected series id")
-		} else if err := sfile.DeleteSeriesIDs([]tsdb.SeriesID{id}); err != nil {
-			t.Fatal(err)
-		}
-	}
-
-	// Compute total size of all series data.
-	origSize, err := sfile.FileSize()
-	if err != nil {
-		t.Fatal(err)
-	}
-
-	// Compact all segments.
-	var paths []string
-	for _, p := range sfile.Partitions() {
-		for _, ss := range p.Segments() {
-			if err := ss.CompactToPath(ss.Path()+".tmp", p.Index()); err != nil {
-				t.Fatal(err)
-			}
-			paths = append(paths, ss.Path())
-		}
-	}
-
-	// Close index.
-	if err := sfile.SeriesFile.Close(); err != nil {
-		t.Fatal(err)
-	}
-
-	// Overwrite files.
-	for _, path := range paths {
-		if err := os.Rename(path+".tmp", path); err != nil {
-			t.Fatal(err)
-		}
-	}
-
-	// Reopen index.
-	sfile.SeriesFile = seriesfile.NewSeriesFile(sfile.SeriesFile.Path())
-	if err := sfile.SeriesFile.Open(context.Background()); err != nil {
-		t.Fatal(err)
-	}
-
-	// Ensure series status is correct.
-	for i := 0; i < n; i++ {
-		if id := sfile.SeriesID(collection.Names[i], collection.Tags[i], nil); id.IsZero() {
-			continue
-		} else if got, want := sfile.IsDeleted(id), (i%10) == 0; got != want {
-			t.Fatalf("IsDeleted(%d)=%v, want %v", id, got, want)
-		}
-	}
-
-	// Verify new size is smaller.
-	newSize, err := sfile.FileSize()
-	if err != nil {
-		t.Fatal(err)
-	} else if newSize >= origSize {
-		t.Fatalf("expected new size (%d) to be smaller than original size (%d)", newSize, origSize)
-	}
-
-	t.Logf("original size: %d, new size: %d", origSize, newSize)
-}
-
-var cachedCompactionSeriesFile *SeriesFile
-
-func BenchmarkSeriesFile_Compaction(b *testing.B) {
-	const n = 1000000
-
-	if cachedCompactionSeriesFile == nil {
-		sfile := MustOpenSeriesFile()
-
-		// Generate a bunch of keys.
-		ids := make([]tsdb.SeriesID, n)
-		for i := 0; i < n; i++ {
-			collection := &tsdb.SeriesCollection{
-				Names: [][]byte{[]byte("cpu")},
-				Tags:  []models.Tags{models.NewTags(map[string]string{"region": fmt.Sprintf("r%d", i)})},
-				Types: []models.FieldType{models.Integer},
-			}
-
-			if err := sfile.CreateSeriesListIfNotExists(collection); err != nil {
-				b.Fatal(err)
-			} else if ids[i] = sfile.SeriesID(collection.Names[0], collection.Tags[0], nil); ids[i].IsZero() {
-				b.Fatalf("expected series id: i=%d", i)
-			}
-		}
-
-		// Delete a subset of keys.
-		for i := 0; i < len(ids); i += 10 {
-			if err := sfile.DeleteSeriesIDs([]tsdb.SeriesID{ids[i]}); err != nil {
-				b.Fatal(err)
-			}
-		}
-
-		cachedCompactionSeriesFile = sfile
-	}
-
-	b.ResetTimer()
-
-	for i := 0; i < b.N; i++ {
-		// Compact all segments in parallel.
-		var g errgroup.Group
-		for _, p := range cachedCompactionSeriesFile.Partitions() {
-			for _, segment := range p.Segments() {
-				p, segment := p, segment
-				g.Go(func() error {
-					return segment.CompactToPath(segment.Path()+".tmp", p.Index())
-				})
-			}
-		}
-
-		if err := g.Wait(); err != nil {
-			b.Fatal(err)
-		}
-	}
-}
-
-// Series represents name/tagset pairs that are used in testing.
-type Series struct {
-	Name    []byte
-	Tags    models.Tags
-	Type    models.FieldType
-	Deleted bool
-}
-
-// SeriesFile is a test wrapper for tsdb.SeriesFile.
-type SeriesFile struct {
-	*seriesfile.SeriesFile
-}
-
-// NewSeriesFile returns a new instance of SeriesFile with a temporary file path.
-func NewSeriesFile() *SeriesFile {
-	dir, err := ioutil.TempDir("", "tsdb-series-file-")
-	if err != nil {
-		panic(err)
-	}
-	return &SeriesFile{SeriesFile: seriesfile.NewSeriesFile(dir)}
-}
-
-func NewBrokenSeriesFile(content []byte) *SeriesFile {
-	sFile := NewSeriesFile()
-	fPath := sFile.Path()
-	if err := sFile.Open(context.Background()); err != nil {
-		panic(err)
-	}
-	if err := sFile.SeriesFile.Close(); err != nil {
-		panic(err)
-	}
-
-	segPath := path.Join(fPath, "00", "0000")
-	if _, err := os.Stat(segPath); os.IsNotExist(err) {
-		panic(err)
-	}
-	err := ioutil.WriteFile(segPath, content, 0777)
-	if err != nil {
-		panic(err)
-	}
-	return sFile
-}
-
-// MustOpenSeriesFile returns a new, open instance of SeriesFile. Panic on error.
-func MustOpenSeriesFile() *SeriesFile {
-	f := NewSeriesFile()
-	f.Logger = logger.New(os.Stdout)
-	if err := f.Open(context.Background()); err != nil {
-		panic(err)
-	}
-	return f
-}
-
-// Close closes the log file and removes it from disk.
-func (f *SeriesFile) Close() error {
-	defer os.RemoveAll(f.Path())
-	return f.SeriesFile.Close()
-}
-
-// Reopen close & reopens the series file.
-func (f *SeriesFile) Reopen() error {
-	if err := f.SeriesFile.Close(); err != nil {
-		return err
-	}
-	f.SeriesFile = seriesfile.NewSeriesFile(f.SeriesFile.Path())
-	return f.SeriesFile.Open(context.Background())
-}
-
-// ForceCompact executes an immediate compaction across all partitions.
-func (f *SeriesFile) ForceCompact() error {
-	for _, p := range f.Partitions() {
-		if _, err := seriesfile.NewSeriesPartitionCompactor().Compact(p); err != nil {
-			return err
-		}
-	}
-	return nil
-}

tsdb/seriesfile/series_index.go

@@ -1,436 +0,0 @@
-package seriesfile
-
-import (
-	"bytes"
-	"encoding/binary"
-	"errors"
-	"io"
-	"os"
-
-	"github.com/influxdata/influxdb/v2/models"
-	"github.com/influxdata/influxdb/v2/pkg/mincore"
-	"github.com/influxdata/influxdb/v2/pkg/mmap"
-	"github.com/influxdata/influxdb/v2/pkg/rhh"
-	"github.com/influxdata/influxdb/v2/tsdb"
-	"github.com/prometheus/client_golang/prometheus"
-	"golang.org/x/time/rate"
-)
-
-const (
-	SeriesIndexVersion = 1
-	SeriesIndexMagic   = "SIDX"
-)
-
-const (
-	// SeriesIDSize is the size in bytes of a series key ID.
-	SeriesIDSize        = 8
-	SeriesOffsetSize    = 8
-	SeriesIndexElemSize = SeriesOffsetSize + SeriesIDSize
-
-	SeriesIndexLoadFactor = 90 // rhh load factor
-
-	SeriesIndexHeaderSize = 0 +
-		4 + 1 + // magic + version
-		8 + 8 + // max series + max offset
-		8 + 8 + // count + capacity
-		8 + 8 + // key/id map offset & size
-		8 + 8 + // id/offset map offset & size
-		0
-)
-
-var ErrInvalidSeriesIndex = errors.New("invalid series index")
-
-// SeriesIndex represents an index of key-to-id & id-to-offset mappings.
-type SeriesIndex struct {
-	path string
-
-	count    uint64
-	capacity int64
-	mask     int64
-
-	maxSeriesID tsdb.SeriesID
-	maxOffset   int64
-
-	// metrics stores a shard instance of some Prometheus metrics. metrics
-	// must be set before Open is called.
-	rhhMetrics        *rhh.Metrics
-	rhhLabels         prometheus.Labels
-	rhhMetricsEnabled bool
-
-	data         []byte // mmap data
-	keyIDData    []byte // key/id mmap data
-	idOffsetData []byte // id/offset mmap data
-
-	// In-memory data since rebuild.
-	keyIDMap    *rhh.HashMap
-	idOffsetMap map[tsdb.SeriesID]int64
-	tombstones  map[tsdb.SeriesID]struct{}
-
-	limiter *mincore.Limiter // Limits page faults by the partition
-}
-
-func NewSeriesIndex(path string) *SeriesIndex {
-	return &SeriesIndex{
-		path:              path,
-		rhhMetricsEnabled: true,
-	}
-}
-
-// Open memory-maps the index file.
-func (idx *SeriesIndex) Open() (err error) {
-	// Map data file, if it exists.
-	if err := func() error {
-		if _, err := os.Stat(idx.path); err != nil && !os.IsNotExist(err) {
-			return err
-		} else if err == nil {
-			if idx.data, err = mmap.Map(idx.path, 0); err != nil {
-				return err
-			}
-
-			hdr, err := ReadSeriesIndexHeader(idx.data)
-			if err != nil {
-				return err
-			}
-			idx.count, idx.capacity, idx.mask = hdr.Count, hdr.Capacity, hdr.Capacity-1
-			idx.maxSeriesID, idx.maxOffset = hdr.MaxSeriesID, hdr.MaxOffset
-
-			idx.keyIDData = idx.data[hdr.KeyIDMap.Offset : hdr.KeyIDMap.Offset+hdr.KeyIDMap.Size]
-			idx.idOffsetData = idx.data[hdr.IDOffsetMap.Offset : hdr.IDOffsetMap.Offset+hdr.IDOffsetMap.Size]
-		}
-		return nil
-	}(); err != nil {
-		idx.Close()
-		return err
-	}
-
-	options := rhh.DefaultOptions
-	options.Metrics = idx.rhhMetrics
-	options.Labels = idx.rhhLabels
-	options.MetricsEnabled = idx.rhhMetricsEnabled
-
-	idx.keyIDMap = rhh.NewHashMap(options)
-	idx.idOffsetMap = make(map[tsdb.SeriesID]int64)
-	idx.tombstones = make(map[tsdb.SeriesID]struct{})
-	return nil
-}
-
-// Close unmaps the index file.
-func (idx *SeriesIndex) Close() (err error) {
-	if idx.data != nil {
-		err = mmap.Unmap(idx.data)
-	}
-	idx.keyIDData = nil
-	idx.idOffsetData = nil
-
-	idx.keyIDMap = nil
-	idx.idOffsetMap = nil
-	idx.tombstones = nil
-	return err
-}
-
-// SetPageFaultLimiter sets the limiter used for rate limiting page faults.
-// Must be called after Open().
-func (idx *SeriesIndex) SetPageFaultLimiter(limiter *rate.Limiter) {
-	idx.limiter = mincore.NewLimiter(limiter, idx.data)
-}
-
-// Recover rebuilds the in-memory index for all new entries.
-func (idx *SeriesIndex) Recover(segments []*SeriesSegment) error {
-	// Allocate new in-memory maps.
-	options := rhh.DefaultOptions
-	options.Metrics = idx.rhhMetrics
-	options.Labels = idx.rhhLabels
-	options.MetricsEnabled = idx.rhhMetricsEnabled
-
-	idx.keyIDMap = rhh.NewHashMap(options)
-	idx.idOffsetMap = make(map[tsdb.SeriesID]int64)
-	idx.tombstones = make(map[tsdb.SeriesID]struct{})
-
-	// Process all entries since the maximum offset in the on-disk index.
-	minSegmentID, _ := SplitSeriesOffset(idx.maxOffset)
-	for _, segment := range segments {
-		if segment.ID() < minSegmentID {
-			continue
-		}
-
-		if err := segment.ForEachEntry(func(flag uint8, id tsdb.SeriesIDTyped, offset int64, key []byte) error {
-			if offset <= idx.maxOffset {
-				return nil
-			}
-			idx.execEntry(flag, id, offset, key)
-			return nil
-		}); err != nil {
-			return err
-		}
-	}
-	return nil
-}
-
-// GrowBy preallocates the in-memory hashmap to a larger size.
-func (idx *SeriesIndex) GrowBy(delta int) {
-	if delta < 0 {
-		return
-	}
-	idx.keyIDMap.Grow(((idx.keyIDMap.Len() + int64(delta)) * 100) / int64(idx.keyIDMap.LoadFactor()))
-}
-
-// Count returns the number of series in the index.
-func (idx *SeriesIndex) Count() uint64 {
-	n := int64(idx.OnDiskCount()+idx.InMemCount()) - int64(len(idx.tombstones))
-	if n < 0 {
-		n = 0
-	}
-	return uint64(n)
-}
-
-// OnDiskCount returns the number of series in the on-disk index.
-func (idx *SeriesIndex) OnDiskCount() uint64 { return idx.count }
-
-// InMemCount returns the number of series in the in-memory index.
-func (idx *SeriesIndex) InMemCount() uint64 { return uint64(len(idx.idOffsetMap)) }
-
-// OnDiskSize returns the on-disk size of the index in bytes.
-func (idx *SeriesIndex) OnDiskSize() uint64 { return uint64(len(idx.data)) }
-
-// InMemSize returns the heap size of the index in bytes. The returned value is
-// an estimation and does not include include all allocated memory.
-func (idx *SeriesIndex) InMemSize() uint64 {
-	n := len(idx.idOffsetMap)
-	return uint64(2*8*n) + uint64(len(idx.tombstones)*8)
-}
-
-func (idx *SeriesIndex) Insert(key []byte, id tsdb.SeriesIDTyped, offset int64) {
-	idx.execEntry(SeriesEntryInsertFlag, id, offset, key)
-}
-
-// Delete marks the series id as deleted.
-func (idx *SeriesIndex) Delete(id tsdb.SeriesID) {
-	// NOTE: WithType(0) kinda sucks here, but we know it will be masked off.
-	idx.execEntry(SeriesEntryTombstoneFlag, id.WithType(0), 0, nil)
-}
-
-// IsDeleted returns true if series id has been deleted.
-func (idx *SeriesIndex) IsDeleted(id tsdb.SeriesID) bool {
-	if _, ok := idx.tombstones[id]; ok {
-		return true
-	}
-	return idx.FindOffsetByID(id) == 0
-}
-
-func (idx *SeriesIndex) execEntry(flag uint8, id tsdb.SeriesIDTyped, offset int64, key []byte) {
-	untypedID := id.SeriesID()
-	switch flag {
-	case SeriesEntryInsertFlag:
-		idx.keyIDMap.PutQuiet(key, id)
-		idx.idOffsetMap[untypedID] = offset
-
-		if untypedID.Greater(idx.maxSeriesID) {
-			idx.maxSeriesID = untypedID
-		}
-		if offset > idx.maxOffset {
-			idx.maxOffset = offset
-		}
-
-	case SeriesEntryTombstoneFlag:
-		// Only add to tombstone if it exists on disk or in-memory.
-		// This affects counts if a tombstone exists but the ID doesn't exist.
-		if idx.FindOffsetByID(untypedID) != 0 {
-			idx.tombstones[untypedID] = struct{}{}
-		}
-
-	default:
-		panic("unreachable")
-	}
-}
-
-func (idx *SeriesIndex) FindIDBySeriesKey(segments []*SeriesSegment, key []byte) tsdb.SeriesIDTyped {
-	if v := idx.keyIDMap.Get(key); v != nil {
-		if id, _ := v.(tsdb.SeriesIDTyped); !id.IsZero() && !idx.IsDeleted(id.SeriesID()) {
-			return id
-		}
-	}
-	if len(idx.data) == 0 {
-		return tsdb.SeriesIDTyped{}
-	}
-
-	hash := rhh.HashKey(key)
-	for d, pos := int64(0), hash&idx.mask; ; d, pos = d+1, (pos+1)&idx.mask {
-		elem := idx.keyIDData[(pos * SeriesIndexElemSize):]
-		elemOffset := int64(binary.BigEndian.Uint64(elem[:SeriesOffsetSize]))
-		_ = wait(idx.limiter, elem[:SeriesOffsetSize]) // elem size is two uint64s
-
-		if elemOffset == 0 {
-			return tsdb.SeriesIDTyped{}
-		}
-
-		elemKey := ReadSeriesKeyFromSegments(segments, elemOffset+SeriesEntryHeaderSize)
-		elemHash := rhh.HashKey(elemKey)
-		if d > rhh.Dist(elemHash, pos, idx.capacity) {
-			return tsdb.SeriesIDTyped{}
-		} else if elemHash == hash && bytes.Equal(elemKey, key) {
-			id := tsdb.NewSeriesIDTyped(binary.BigEndian.Uint64(elem[SeriesOffsetSize:]))
-			if idx.IsDeleted(id.SeriesID()) {
-				return tsdb.SeriesIDTyped{}
-			}
-			return id
-		}
-	}
-}
-
-func (idx *SeriesIndex) FindIDByNameTags(segments []*SeriesSegment, name []byte, tags models.Tags, buf []byte) tsdb.SeriesIDTyped {
-	id := idx.FindIDBySeriesKey(segments, AppendSeriesKey(buf[:0], name, tags))
-	if _, ok := idx.tombstones[id.SeriesID()]; ok {
-		return tsdb.SeriesIDTyped{}
-	}
-	return id
-}
-
-func (idx *SeriesIndex) FindIDListByNameTags(segments []*SeriesSegment, names [][]byte, tagsSlice []models.Tags, buf []byte) (ids []tsdb.SeriesIDTyped, ok bool) {
-	ids, ok = make([]tsdb.SeriesIDTyped, len(names)), true
-	for i := range names {
-		id := idx.FindIDByNameTags(segments, names[i], tagsSlice[i], buf)
-		if id.IsZero() {
-			ok = false
-			continue
-		}
-		ids[i] = id
-	}
-	return ids, ok
-}
-
-func (idx *SeriesIndex) FindOffsetByID(id tsdb.SeriesID) int64 {
-	if offset := idx.idOffsetMap[id]; offset != 0 {
-		return offset
-	} else if len(idx.data) == 0 {
-		return 0
-	}
-
-	hash := rhh.HashUint64(id.RawID())
-	for d, pos := int64(0), hash&idx.mask; ; d, pos = d+1, (pos+1)&idx.mask {
-		elem := idx.idOffsetData[(pos * SeriesIndexElemSize):]
-		elemID := tsdb.NewSeriesID(binary.BigEndian.Uint64(elem[:SeriesIDSize]))
-		_ = wait(idx.limiter, elem[:SeriesIDSize])
-
-		if elemID == id {
-			return int64(binary.BigEndian.Uint64(elem[SeriesIDSize:]))
-		} else if elemID.IsZero() || d > rhh.Dist(rhh.HashUint64(elemID.RawID()), pos, idx.capacity) {
-			return 0
-		}
-	}
-}
-
-// Clone returns a copy of idx for use during compaction. In-memory maps are not cloned.
-func (idx *SeriesIndex) Clone() *SeriesIndex {
-	tombstones := make(map[tsdb.SeriesID]struct{}, len(idx.tombstones))
-	for id := range idx.tombstones {
-		tombstones[id] = struct{}{}
-	}
-
-	idOffsetMap := make(map[tsdb.SeriesID]int64)
-	for k, v := range idx.idOffsetMap {
-		idOffsetMap[k] = v
-	}
-
-	return &SeriesIndex{
-		path:         idx.path,
-		count:        idx.count,
-		capacity:     idx.capacity,
-		mask:         idx.mask,
-		maxSeriesID:  idx.maxSeriesID,
-		maxOffset:    idx.maxOffset,
-		data:         idx.data,
-		keyIDData:    idx.keyIDData,
-		idOffsetData: idx.idOffsetData,
-		tombstones:   tombstones,
-		idOffsetMap:  idOffsetMap,
-	}
-}
-
-// SeriesIndexHeader represents the header of a series index.
-type SeriesIndexHeader struct {
-	Version uint8
-
-	MaxSeriesID tsdb.SeriesID
-	MaxOffset   int64
-
-	Count    uint64
-	Capacity int64
-
-	KeyIDMap struct {
-		Offset int64
-		Size   int64
-	}
-
-	IDOffsetMap struct {
-		Offset int64
-		Size   int64
-	}
-}
-
-// NewSeriesIndexHeader returns a new instance of SeriesIndexHeader.
-func NewSeriesIndexHeader() SeriesIndexHeader {
-	return SeriesIndexHeader{Version: SeriesIndexVersion}
-}
-
-// ReadSeriesIndexHeader returns the header from data.
-func ReadSeriesIndexHeader(data []byte) (hdr SeriesIndexHeader, err error) {
-	r := bytes.NewReader(data)
-
-	// Read magic number.
-	magic := make([]byte, len(SeriesIndexMagic))
-	if _, err := io.ReadFull(r, magic); err != nil {
-		return hdr, err
-	} else if !bytes.Equal([]byte(SeriesIndexMagic), magic) {
-		return hdr, ErrInvalidSeriesIndex
-	}
-
-	// Read version.
-	if err := binary.Read(r, binary.BigEndian, &hdr.Version); err != nil {
-		return hdr, err
-	}
-
-	// Read max offset.
-	if err := binary.Read(r, binary.BigEndian, &hdr.MaxSeriesID.ID); err != nil {
-		return hdr, err
-	} else if err := binary.Read(r, binary.BigEndian, &hdr.MaxOffset); err != nil {
-		return hdr, err
-	}
-
-	// Read count & capacity.
-	if err := binary.Read(r, binary.BigEndian, &hdr.Count); err != nil {
-		return hdr, err
-	} else if err := binary.Read(r, binary.BigEndian, &hdr.Capacity); err != nil {
-		return hdr, err
-	}
-
-	// Read key/id map position.
-	if err := binary.Read(r, binary.BigEndian, &hdr.KeyIDMap.Offset); err != nil {
-		return hdr, err
-	} else if err := binary.Read(r, binary.BigEndian, &hdr.KeyIDMap.Size); err != nil {
-		return hdr, err
-	}
-
-	// Read offset/id map position.
-	if err := binary.Read(r, binary.BigEndian, &hdr.IDOffsetMap.Offset); err != nil {
-		return hdr, err
-	} else if err := binary.Read(r, binary.BigEndian, &hdr.IDOffsetMap.Size); err != nil {
-		return hdr, err
-	}
-	return hdr, nil
-}
-
-// WriteTo writes the header to w.
-func (hdr *SeriesIndexHeader) WriteTo(w io.Writer) (n int64, err error) {
-	var buf bytes.Buffer
-	buf.WriteString(SeriesIndexMagic)
-	binary.Write(&buf, binary.BigEndian, hdr.Version)
-	binary.Write(&buf, binary.BigEndian, hdr.MaxSeriesID)
-	binary.Write(&buf, binary.BigEndian, hdr.MaxOffset)
-	binary.Write(&buf, binary.BigEndian, hdr.Count)
-	binary.Write(&buf, binary.BigEndian, hdr.Capacity)
-	binary.Write(&buf, binary.BigEndian, hdr.KeyIDMap.Offset)
-	binary.Write(&buf, binary.BigEndian, hdr.KeyIDMap.Size)
-	binary.Write(&buf, binary.BigEndian, hdr.IDOffsetMap.Offset)
-	binary.Write(&buf, binary.BigEndian, hdr.IDOffsetMap.Size)
-	return buf.WriteTo(w)
-}