+ defer f.Close()
+ fis, err := f.Readdir(-1)
+ if err != nil {
+ return []string{path}, nil
+ }
+ for _, fi := range fis {
+ if fi.Name() == "." || fi.Name() == ".." {
+ continue
+ } else if child := path + "/" + fi.Name(); fi.IsDir() {
+ add, err := allGobFiles(child)
+ if err != nil {
+ return nil, err
+ }
+ files = append(files, add...)
+ } else if strings.HasSuffix(child, ".gob") || strings.HasSuffix(child, ".gob.gz") {
+ files = append(files, child)
+ }
+ }
+ return files, nil
+}
+
+func (tilelib *tileLibrary) LoadDir(ctx context.Context, path string) error {
+ log.Infof("LoadDir: walk dir %s", path)
+ files, err := allGobFiles(path)
+ if err != nil {
+ return err
+ }
+ ctx, cancel := context.WithCancel(ctx)
+ defer cancel()
+ var mtx sync.Mutex
+ allcgs := make([][]CompactGenome, len(files))
+ allcseqs := make([][]CompactSequence, len(files))
+ allvariantmap := map[tileLibRef]tileVariantID{}
+ errs := make(chan error, len(files))
+ log.Infof("LoadDir: read %d files", len(files))
+ for fileno, path := range files {
+ fileno, path := fileno, path
+ go func() {
+ f, err := open(path)
+ if err != nil {
+ errs <- err
+ return
+ }
+ defer f.Close()
+ defer log.Infof("LoadDir: finished reading %s", path)
+
+ var variantmap = map[tileLibRef]tileVariantID{}
+ var cgs []CompactGenome
+ var cseqs []CompactSequence
+ err = DecodeLibrary(f, strings.HasSuffix(path, ".gz"), func(ent *LibraryEntry) error {
+ if ctx.Err() != nil {
+ return ctx.Err()
+ }
+ if len(ent.TagSet) > 0 {
+ mtx.Lock()
+ if tilelib.taglib == nil || tilelib.taglib.Len() != len(ent.TagSet) {
+ // load first set of tags, or
+ // report mismatch if 2 sets
+ // have different #tags.
+ if err := tilelib.loadTagSet(ent.TagSet); err != nil {
+ mtx.Unlock()
+ return err
+ }
+ }
+ mtx.Unlock()
+ }
+ for _, tv := range ent.TileVariants {
+ variantmap[tileLibRef{Tag: tv.Tag, Variant: tv.Variant}] = tilelib.getRef(tv.Tag, tv.Sequence, tv.Ref).Variant
+ }
+ cgs = append(cgs, ent.CompactGenomes...)
+ cseqs = append(cseqs, ent.CompactSequences...)
+ return nil
+ })
+ allcgs[fileno] = cgs
+ allcseqs[fileno] = cseqs
+ mtx.Lock()
+ defer mtx.Unlock()
+ for k, v := range variantmap {
+ allvariantmap[k] = v
+ }
+ errs <- err
+ }()
+ }
+ for range files {
+ err := <-errs
+ if err != nil {
+ return err
+ }
+ }
+
+ log.Info("LoadDir: loadCompactGenomes")
+ var flatcgs []CompactGenome
+ for _, cgs := range allcgs {
+ flatcgs = append(flatcgs, cgs...)
+ }
+ err = tilelib.loadCompactGenomes(flatcgs, allvariantmap)
+ if err != nil {
+ return err
+ }
+
+ log.Info("LoadDir: loadCompactSequences")
+ var flatcseqs []CompactSequence
+ for _, cseqs := range allcseqs {
+ flatcseqs = append(flatcseqs, cseqs...)
+ }
+ err = tilelib.loadCompactSequences(flatcseqs, allvariantmap)
+ if err != nil {
+ return err
+ }
+
+ log.Info("LoadDir done")
+ return nil
+}
+
+func (tilelib *tileLibrary) WriteDir(dir string) error {
+ ntilefiles := 128
+ nfiles := ntilefiles + len(tilelib.refseqs)
+ files := make([]*os.File, nfiles)
+ for i := range files {
+ f, err := os.OpenFile(fmt.Sprintf("%s/library.%04d.gob.gz", dir, i), os.O_CREATE|os.O_WRONLY, 0666)
+ if err != nil {
+ return err
+ }
+ defer f.Close()
+ files[i] = f
+ }
+ bufws := make([]*bufio.Writer, nfiles)
+ for i := range bufws {
+ bufws[i] = bufio.NewWriterSize(files[i], 1<<26)
+ }
+ zws := make([]*pgzip.Writer, nfiles)
+ for i := range zws {
+ zws[i] = pgzip.NewWriter(bufws[i])
+ defer zws[i].Close()
+ }
+ encoders := make([]*gob.Encoder, nfiles)
+ for i := range encoders {
+ encoders[i] = gob.NewEncoder(zws[i])
+ }
+
+ cgnames := make([]string, 0, len(tilelib.compactGenomes))
+ for name := range tilelib.compactGenomes {
+ cgnames = append(cgnames, name)
+ }
+ sort.Strings(cgnames)
+
+ refnames := make([]string, 0, len(tilelib.refseqs))
+ for name := range tilelib.refseqs {
+ refnames = append(refnames, name)
+ }
+ sort.Strings(refnames)
+
+ log.Infof("WriteDir: writing %d files", nfiles)
+ ctx, cancel := context.WithCancel(context.Background())
+ defer cancel()
+ errs := make(chan error, nfiles)
+ for start := range files {
+ start := start
+ go func() {
+ err := encoders[start].Encode(LibraryEntry{TagSet: tilelib.taglib.Tags()})
+ if err != nil {
+ errs <- err
+ return
+ }
+ if refidx := start - ntilefiles; refidx >= 0 {
+ // write each ref to its own file
+ // (they seem to load very slowly)
+ name := refnames[refidx]
+ errs <- encoders[start].Encode(LibraryEntry{CompactSequences: []CompactSequence{{
+ Name: name,
+ TileSequences: tilelib.refseqs[name],
+ }}})
+ return
+ }
+ for i := start; i < len(cgnames); i += ntilefiles {
+ err := encoders[start].Encode(LibraryEntry{CompactGenomes: []CompactGenome{{
+ Name: cgnames[i],
+ Variants: tilelib.compactGenomes[cgnames[i]],
+ }}})
+ if err != nil {
+ errs <- err
+ return
+ }
+ }
+ tvs := []TileVariant{}
+ for tag := start; tag < len(tilelib.variant) && ctx.Err() == nil; tag += ntilefiles {
+ tvs = tvs[:0]
+ for idx, hash := range tilelib.variant[tag] {
+ tvs = append(tvs, TileVariant{
+ Tag: tagID(tag),
+ Variant: tileVariantID(idx + 1),
+ Blake2b: hash,
+ Sequence: tilelib.hashSequence(hash),
+ })
+ }
+ err := encoders[start].Encode(LibraryEntry{TileVariants: tvs})
+ if err != nil {
+ errs <- err
+ return
+ }
+ }
+ errs <- nil
+ }()
+ }
+ for range files {
+ err := <-errs
+ if err != nil {
+ return err
+ }
+ }
+ log.Info("WriteDir: flushing")
+ for i := range zws {
+ err := zws[i].Close()
+ if err != nil {
+ return err
+ }
+ err = bufws[i].Flush()
+ if err != nil {
+ return err
+ }
+ err = files[i].Close()
+ if err != nil {
+ return err
+ }
+ }
+ log.Info("WriteDir: done")
+ return nil
+}
+
+// Load library data from rdr. Tile variants might be renumbered in
+// the process; in that case, genomes variants will be renumbered to
+// match.
+func (tilelib *tileLibrary) LoadGob(ctx context.Context, rdr io.Reader, gz bool) error {
+ cgs := []CompactGenome{}
+ cseqs := []CompactSequence{}
+ variantmap := map[tileLibRef]tileVariantID{}
+ err := DecodeLibrary(rdr, gz, func(ent *LibraryEntry) error {
+ if ctx.Err() != nil {
+ return ctx.Err()
+ }
+ if err := tilelib.loadTagSet(ent.TagSet); err != nil {
+ return err
+ }
+ if err := tilelib.loadTileVariants(ent.TileVariants, variantmap); err != nil {
+ return err
+ }
+ cgs = append(cgs, ent.CompactGenomes...)
+ cseqs = append(cseqs, ent.CompactSequences...)
+ return nil
+ })
+ if err != nil {
+ return err
+ }
+ if ctx.Err() != nil {
+ return ctx.Err()
+ }
+ err = tilelib.loadCompactGenomes(cgs, variantmap)
+ if err != nil {
+ return err
+ }
+ err = tilelib.loadCompactSequences(cseqs, variantmap)
+ if err != nil {
+ return err
+ }
+ return nil
+}
+
+func (tilelib *tileLibrary) dump(out io.Writer) {
+ printTV := func(tag int, variant tileVariantID) {
+ if variant < 1 {
+ fmt.Fprintf(out, " -")
+ } else if tag >= len(tilelib.variant) {
+ fmt.Fprintf(out, " (!tag=%d)", tag)
+ } else if int(variant) > len(tilelib.variant[tag]) {
+ fmt.Fprintf(out, " (tag=%d,!variant=%d)", tag, variant)
+ } else {
+ fmt.Fprintf(out, " %x", tilelib.variant[tag][variant-1][:8])
+ }
+ }
+ for refname, refseqs := range tilelib.refseqs {
+ for seqname, seq := range refseqs {
+ fmt.Fprintf(out, "ref %s %s", refname, seqname)
+ for _, libref := range seq {
+ printTV(int(libref.Tag), libref.Variant)
+ }
+ fmt.Fprintf(out, "\n")
+ }
+ }
+ for name, cg := range tilelib.compactGenomes {
+ fmt.Fprintf(out, "cg %s", name)
+ for tag, variant := range cg {
+ printTV(tag/2, variant)
+ }
+ fmt.Fprintf(out, "\n")
+ }
+}
+
+type importStats struct {
+ InputFile string
+ InputLabel string
+ InputLength int
+ InputCoverage int
+ PathLength int
+ DroppedOutOfOrderTiles int
+}
+
+func (tilelib *tileLibrary) TileFasta(filelabel string, rdr io.Reader, matchChromosome *regexp.Regexp, isRef bool) (tileSeq, []importStats, error) {
+ ret := tileSeq{}
+ type jobT struct {
+ label string
+ fasta []byte
+ }
+ todo := make(chan jobT, 1)
+ scanner := bufio.NewScanner(rdr)
+ go func() {
+ defer close(todo)
+ var fasta []byte
+ var seqlabel string
+ for scanner.Scan() {
+ buf := scanner.Bytes()
+ if len(buf) > 0 && buf[0] == '>' {
+ todo <- jobT{seqlabel, append([]byte(nil), fasta...)}
+ seqlabel, fasta = strings.SplitN(string(buf[1:]), " ", 2)[0], fasta[:0]
+ log.Debugf("%s %s reading fasta", filelabel, seqlabel)
+ } else {
+ fasta = append(fasta, bytes.ToLower(buf)...)
+ }
+ }
+ todo <- jobT{seqlabel, fasta}
+ }()
+ type foundtag struct {
+ pos int
+ tagid tagID
+ }
+ found := make([]foundtag, 2000000)
+ path := make([]tileLibRef, 2000000)
+ totalFoundTags := 0
+ totalPathLen := 0
+ skippedSequences := 0
+ taglen := tilelib.taglib.TagLen()
+ var stats []importStats
+ for job := range todo {
+ if len(job.fasta) == 0 {
+ continue
+ } else if !matchChromosome.MatchString(job.label) {
+ skippedSequences++
+ continue
+ }
+ log.Debugf("%s %s tiling", filelabel, job.label)
+
+ found = found[:0]
+ tilelib.taglib.FindAll(job.fasta, func(tagid tagID, pos, taglen int) {
+ found = append(found, foundtag{pos: pos, tagid: tagid})
+ })
+ totalFoundTags += len(found)
+ if len(found) == 0 {
+ log.Warnf("%s %s no tags found", filelabel, job.label)
+ }
+
+ skipped := 0
+ if tilelib.skipOOO {
+ log.Infof("%s %s keeping longest increasing subsequence", filelabel, job.label)
+ keep := longestIncreasingSubsequence(len(found), func(i int) int { return int(found[i].tagid) })
+ for i, x := range keep {
+ found[i] = found[x]
+ }
+ skipped = len(found) - len(keep)
+ found = found[:len(keep)]
+ }
+
+ log.Infof("%s %s getting %d librefs", filelabel, job.label, len(found))
+ throttle := &throttle{Max: runtime.NumCPU()}
+ path = path[:len(found)]
+ var lowquality int64
+ for i, f := range found {
+ i, f := i, f
+ throttle.Acquire()
+ go func() {
+ defer throttle.Release()
+ var startpos, endpos int
+ if i == 0 {
+ startpos = 0
+ } else {
+ startpos = f.pos
+ }
+ if i == len(found)-1 {
+ endpos = len(job.fasta)
+ } else {
+ endpos = found[i+1].pos + taglen
+ }
+ path[i] = tilelib.getRef(f.tagid, job.fasta[startpos:endpos], isRef)
+ if countBases(job.fasta[startpos:endpos]) != endpos-startpos {
+ atomic.AddInt64(&lowquality, 1)
+ }
+ }()
+ }
+ throttle.Wait()
+
+ log.Infof("%s %s copying path", filelabel, job.label)
+
+ pathcopy := make([]tileLibRef, len(path))
+ copy(pathcopy, path)
+ ret[job.label] = pathcopy
+
+ basesIn := countBases(job.fasta)
+ log.Infof("%s %s fasta in %d coverage in %d path len %d low-quality %d skipped-out-of-order %d", filelabel, job.label, len(job.fasta), basesIn, len(path), lowquality, skipped)
+ stats = append(stats, importStats{
+ InputFile: filelabel,
+ InputLabel: job.label,
+ InputLength: len(job.fasta),
+ InputCoverage: basesIn,
+ PathLength: len(path),
+ DroppedOutOfOrderTiles: skipped,
+ })
+
+ totalPathLen += len(path)
+ }
+ log.Printf("%s tiled with total path len %d in %d sequences (skipped %d sequences that did not match chromosome regexp, skipped %d out-of-order tags)", filelabel, totalPathLen, len(ret), skippedSequences, totalFoundTags-totalPathLen)
+ return ret, stats, scanner.Err()
+}
+
+func (tilelib *tileLibrary) Len() int64 {
+ return atomic.LoadInt64(&tilelib.variants)