tilelib.go

   1 package lightning
   2
   3 import (
   4         "bufio"
   5         "bytes"
   6         "context"
   7         "encoding/gob"
   8         "fmt"
   9         "io"
  10         "os"
  11         "regexp"
  12         "runtime"
  13         "sort"
  14         "strings"
  15         "sync"
  16         "sync/atomic"
  17
  18         "github.com/klauspost/pgzip"
  19         log "github.com/sirupsen/logrus"
  20         "golang.org/x/crypto/blake2b"
  21 )
  22
  23 type tileVariantID uint16 // 1-based
  24
  25 type tileLibRef struct {
  26         Tag     tagID
  27         Variant tileVariantID
  28 }
  29
  30 type tileSeq map[string][]tileLibRef
  31
  32 func (tseq tileSeq) Variants() ([]tileVariantID, int, int) {
  33         maxtag := 0
  34         for _, refs := range tseq {
  35                 for _, ref := range refs {
  36                         if maxtag < int(ref.Tag) {
  37                                 maxtag = int(ref.Tag)
  38                         }
  39                 }
  40         }
  41         vars := make([]tileVariantID, maxtag+1)
  42         var kept, dropped int
  43         for _, refs := range tseq {
  44                 for _, ref := range refs {
  45                         if vars[int(ref.Tag)] != 0 {
  46                                 dropped++
  47                         } else {
  48                                 kept++
  49                         }
  50                         vars[int(ref.Tag)] = ref.Variant
  51                 }
  52         }
  53         return vars, kept, dropped
  54 }
  55
  56 type tileLibrary struct {
  57         retainNoCalls       bool
  58         skipOOO             bool
  59         retainTileSequences bool
  60
  61         taglib         *tagLibrary
  62         variant        [][][blake2b.Size256]byte
  63         refseqs        map[string]map[string][]tileLibRef
  64         compactGenomes map[string][]tileVariantID
  65         // count [][]int
  66         seq      map[[blake2b.Size256]byte][]byte
  67         variants int64
  68         // if non-nil, write out any tile variants added while tiling
  69         encoder *gob.Encoder
  70
  71         mtx   sync.RWMutex
  72         vlock []sync.Locker
  73 }
  74
  75 func (tilelib *tileLibrary) loadTagSet(newtagset [][]byte) error {
  76         // Loading a tagset means either passing it through to the
  77         // output (if it's the first one we've seen), or just ensuring
  78         // it doesn't disagree with what we already have.
  79         if len(newtagset) == 0 {
  80                 return nil
  81         }
  82         tilelib.mtx.Lock()
  83         defer tilelib.mtx.Unlock()
  84         if tilelib.taglib == nil || tilelib.taglib.Len() == 0 {
  85                 tilelib.taglib = &tagLibrary{}
  86                 err := tilelib.taglib.setTags(newtagset)
  87                 if err != nil {
  88                         return err
  89                 }
  90                 if tilelib.encoder != nil {
  91                         err = tilelib.encoder.Encode(LibraryEntry{
  92                                 TagSet: newtagset,
  93                         })
  94                         if err != nil {
  95                                 return err
  96                         }
  97                 }
  98         } else if tilelib.taglib.Len() != len(newtagset) {
  99                 return fmt.Errorf("cannot merge libraries with differing tagsets")
 100         } else {
 101                 current := tilelib.taglib.Tags()
 102                 for i := range newtagset {
 103                         if !bytes.Equal(newtagset[i], current[i]) {
 104                                 return fmt.Errorf("cannot merge libraries with differing tagsets")
 105                         }
 106                 }
 107         }
 108         return nil
 109 }
 110
 111 func (tilelib *tileLibrary) loadTileVariants(tvs []TileVariant, variantmap map[tileLibRef]tileVariantID) error {
 112         for _, tv := range tvs {
 113                 // Assign a new variant ID (unique across all inputs)
 114                 // for each input variant.
 115                 variantmap[tileLibRef{Tag: tv.Tag, Variant: tv.Variant}] = tilelib.getRef(tv.Tag, tv.Sequence).Variant
 116         }
 117         return nil
 118 }
 119
 120 func (tilelib *tileLibrary) loadCompactGenomes(cgs []CompactGenome, variantmap map[tileLibRef]tileVariantID, onLoadGenome func(CompactGenome)) error {
 121         log.Debugf("loadCompactGenomes: %d", len(cgs))
 122         var wg sync.WaitGroup
 123         errs := make(chan error, 1)
 124         for _, cg := range cgs {
 125                 wg.Add(1)
 126                 cg := cg
 127                 go func() {
 128                         defer wg.Done()
 129                         for i, variant := range cg.Variants {
 130                                 if len(errs) > 0 {
 131                                         return
 132                                 }
 133                                 if variant == 0 {
 134                                         continue
 135                                 }
 136                                 tag := tagID(i / 2)
 137                                 newvariant, ok := variantmap[tileLibRef{Tag: tag, Variant: variant}]
 138                                 if !ok {
 139                                         err := fmt.Errorf("oops: genome %q has variant %d for tag %d, but that variant was not in its library", cg.Name, variant, tag)
 140                                         select {
 141                                         case errs <- err:
 142                                         default:
 143                                         }
 144                                         return
 145                                 }
 146                                 log.Tracef("loadCompactGenomes: cg %s tag %d variant %d => %d", cg.Name, tag, variant, newvariant)
 147                                 cg.Variants[i] = newvariant
 148                         }
 149                         if onLoadGenome != nil {
 150                                 onLoadGenome(cg)
 151                         }
 152                         if tilelib.encoder != nil {
 153                                 err := tilelib.encoder.Encode(LibraryEntry{
 154                                         CompactGenomes: []CompactGenome{cg},
 155                                 })
 156                                 if err != nil {
 157                                         select {
 158                                         case errs <- err:
 159                                         default:
 160                                         }
 161                                         return
 162                                 }
 163                         }
 164                         if tilelib.compactGenomes != nil {
 165                                 tilelib.mtx.Lock()
 166                                 defer tilelib.mtx.Unlock()
 167                                 tilelib.compactGenomes[cg.Name] = cg.Variants
 168                         }
 169                 }()
 170         }
 171         wg.Wait()
 172         go close(errs)
 173         return <-errs
 174 }
 175
 176 func (tilelib *tileLibrary) loadCompactSequences(cseqs []CompactSequence, variantmap map[tileLibRef]tileVariantID) error {
 177         log.Debugf("loadCompactSequences: %d", len(cseqs))
 178         for _, cseq := range cseqs {
 179                 for _, tseq := range cseq.TileSequences {
 180                         for i, libref := range tseq {
 181                                 if libref.Variant == 0 {
 182                                         // No variant (e.g., import
 183                                         // dropped tiles with
 184                                         // no-calls) = no translation.
 185                                         continue
 186                                 }
 187                                 v, ok := variantmap[libref]
 188                                 if !ok {
 189                                         return fmt.Errorf("oops: CompactSequence %q has variant %d for tag %d, but that variant was not in its library", cseq.Name, libref.Variant, libref.Tag)
 190                                 }
 191                                 tseq[i].Variant = v
 192                         }
 193                 }
 194                 if tilelib.encoder != nil {
 195                         if err := tilelib.encoder.Encode(LibraryEntry{
 196                                 CompactSequences: []CompactSequence{cseq},
 197                         }); err != nil {
 198                                 return err
 199                         }
 200                 }
 201         }
 202         tilelib.mtx.Lock()
 203         defer tilelib.mtx.Unlock()
 204         if tilelib.refseqs == nil {
 205                 tilelib.refseqs = map[string]map[string][]tileLibRef{}
 206         }
 207         for _, cseq := range cseqs {
 208                 tilelib.refseqs[cseq.Name] = cseq.TileSequences
 209         }
 210         return nil
 211 }
 212
 213 func (tilelib *tileLibrary) LoadDir(ctx context.Context, path string, onLoadGenome func(CompactGenome)) error {
 214         var files []string
 215         var walk func(string) error
 216         walk = func(path string) error {
 217                 f, err := open(path)
 218                 if err != nil {
 219                         return err
 220                 }
 221                 defer f.Close()
 222                 fis, err := f.Readdir(-1)
 223                 if err != nil {
 224                         files = append(files, path)
 225                         return nil
 226                 }
 227                 for _, fi := range fis {
 228                         if fi.Name() == "." || fi.Name() == ".." {
 229                                 continue
 230                         } else if child := path + "/" + fi.Name(); fi.IsDir() {
 231                                 err = walk(child)
 232                                 if err != nil {
 233                                         return err
 234                                 }
 235                         } else if strings.HasSuffix(child, ".gob") || strings.HasSuffix(child, ".gob.gz") {
 236                                 files = append(files, child)
 237                         }
 238                 }
 239                 return nil
 240         }
 241         log.Infof("LoadDir: walk dir %s", path)
 242         err := walk(path)
 243         if err != nil {
 244                 return err
 245         }
 246         ctx, cancel := context.WithCancel(ctx)
 247         defer cancel()
 248         var mtx sync.Mutex
 249         cgs := []CompactGenome{}
 250         cseqs := []CompactSequence{}
 251         variantmap := map[tileLibRef]tileVariantID{}
 252         errs := make(chan error, len(files))
 253         log.Infof("LoadDir: read %d files", len(files))
 254         throttle := throttle{Max: runtime.GOMAXPROCS(0)}
 255         for _, path := range files {
 256                 path := path
 257                 go func() {
 258                         throttle.Acquire()
 259                         defer throttle.Release()
 260                         f, err := open(path)
 261                         if err != nil {
 262                                 errs <- err
 263                                 return
 264                         }
 265                         defer f.Close()
 266                         defer log.Infof("LoadDir: finished reading %s", path)
 267                         errs <- DecodeLibrary(f, strings.HasSuffix(path, ".gz"), func(ent *LibraryEntry) error {
 268                                 if ctx.Err() != nil {
 269                                         return ctx.Err()
 270                                 }
 271                                 if len(ent.TagSet) > 0 {
 272                                         mtx.Lock()
 273                                         if tilelib.taglib == nil || tilelib.taglib.Len() != len(ent.TagSet) {
 274                                                 // load first set of tags, or
 275                                                 // report mismatch if 2 sets
 276                                                 // have different #tags.
 277                                                 if err := tilelib.loadTagSet(ent.TagSet); err != nil {
 278                                                         mtx.Unlock()
 279                                                         return err
 280                                                 }
 281                                         }
 282                                         mtx.Unlock()
 283                                 }
 284                                 variantmapadd := map[tileLibRef]tileVariantID{}
 285                                 for _, tv := range ent.TileVariants {
 286                                         variantmapadd[tileLibRef{Tag: tv.Tag, Variant: tv.Variant}] = tilelib.getRef(tv.Tag, tv.Sequence).Variant
 287                                 }
 288                                 mtx.Lock()
 289                                 cgs = append(cgs, ent.CompactGenomes...)
 290                                 cseqs = append(cseqs, ent.CompactSequences...)
 291                                 for k, v := range variantmapadd {
 292                                         variantmap[k] = v
 293                                 }
 294                                 mtx.Unlock()
 295                                 return nil
 296                         })
 297                 }()
 298         }
 299         for range files {
 300                 err := <-errs
 301                 if err != nil {
 302                         return err
 303                 }
 304         }
 305         log.Info("LoadDir: loadCompactGenomes")
 306         err = tilelib.loadCompactGenomes(cgs, variantmap, onLoadGenome)
 307         if err != nil {
 308                 return err
 309         }
 310         log.Info("LoadDir: loadCompactSequences")
 311         err = tilelib.loadCompactSequences(cseqs, variantmap)
 312         if err != nil {
 313                 return err
 314         }
 315         log.Info("LoadDir done")
 316         return nil
 317 }
 318
 319 func (tilelib *tileLibrary) WriteDir(dir string) error {
 320         nfiles := 128
 321         files := make([]*os.File, nfiles)
 322         for i := range files {
 323                 f, err := os.OpenFile(fmt.Sprintf("%s/library.%04d.gob.gz", dir, i), os.O_CREATE|os.O_WRONLY, 0666)
 324                 if err != nil {
 325                         return err
 326                 }
 327                 defer f.Close()
 328                 files[i] = f
 329         }
 330         bufws := make([]*bufio.Writer, nfiles)
 331         for i := range bufws {
 332                 bufws[i] = bufio.NewWriterSize(files[i], 1<<26)
 333         }
 334         zws := make([]*pgzip.Writer, nfiles)
 335         for i := range zws {
 336                 zws[i] = pgzip.NewWriter(bufws[i])
 337                 defer zws[i].Close()
 338         }
 339         encoders := make([]*gob.Encoder, nfiles)
 340         for i := range encoders {
 341                 encoders[i] = gob.NewEncoder(zws[i])
 342         }
 343
 344         cgnames := make([]string, 0, len(tilelib.compactGenomes))
 345         for name := range tilelib.compactGenomes {
 346                 cgnames = append(cgnames, name)
 347         }
 348         sort.Strings(cgnames)
 349
 350         log.Infof("WriteDir: writing %d files", nfiles)
 351         ctx, cancel := context.WithCancel(context.Background())
 352         defer cancel()
 353         errs := make(chan error, nfiles)
 354         for start := range files {
 355                 start := start
 356                 go func() {
 357                         err := encoders[start].Encode(LibraryEntry{TagSet: tilelib.taglib.Tags()})
 358                         if err != nil {
 359                                 errs <- err
 360                                 return
 361                         }
 362                         if start == 0 {
 363                                 // For now, just write all the refs to
 364                                 // the first file
 365                                 for name, tseqs := range tilelib.refseqs {
 366                                         err := encoders[start].Encode(LibraryEntry{CompactSequences: []CompactSequence{{
 367                                                 Name:          name,
 368                                                 TileSequences: tseqs,
 369                                         }}})
 370                                         if err != nil {
 371                                                 errs <- err
 372                                                 return
 373                                         }
 374                                 }
 375                         }
 376                         for i := start; i < len(cgnames); i += nfiles {
 377                                 err := encoders[start].Encode(LibraryEntry{CompactGenomes: []CompactGenome{{
 378                                         Name:     cgnames[i],
 379                                         Variants: tilelib.compactGenomes[cgnames[i]],
 380                                 }}})
 381                                 if err != nil {
 382                                         errs <- err
 383                                         return
 384                                 }
 385                         }
 386                         tvs := []TileVariant{}
 387                         for tag := start; tag < len(tilelib.variant) && ctx.Err() == nil; tag += nfiles {
 388                                 tvs = tvs[:0]
 389                                 for idx, hash := range tilelib.variant[tag] {
 390                                         tvs = append(tvs, TileVariant{
 391                                                 Tag:      tagID(tag),
 392                                                 Variant:  tileVariantID(idx + 1),
 393                                                 Blake2b:  hash,
 394                                                 Sequence: tilelib.seq[hash],
 395                                         })
 396                                 }
 397                                 err := encoders[start].Encode(LibraryEntry{TileVariants: tvs})
 398                                 if err != nil {
 399                                         errs <- err
 400                                         return
 401                                 }
 402                         }
 403                         errs <- nil
 404                 }()
 405         }
 406         for range files {
 407                 err := <-errs
 408                 if err != nil {
 409                         return err
 410                 }
 411         }
 412         log.Info("WriteDir: flushing")
 413         for i := range zws {
 414                 err := zws[i].Close()
 415                 if err != nil {
 416                         return err
 417                 }
 418                 err = bufws[i].Flush()
 419                 if err != nil {
 420                         return err
 421                 }
 422                 err = files[i].Close()
 423                 if err != nil {
 424                         return err
 425                 }
 426         }
 427         log.Info("WriteDir: done")
 428         return nil
 429 }
 430
 431 // Load library data from rdr. Tile variants might be renumbered in
 432 // the process; in that case, genomes variants will be renumbered to
 433 // match.
 434 //
 435 // If onLoadGenome is non-nil, call it on each CompactGenome entry.
 436 func (tilelib *tileLibrary) LoadGob(ctx context.Context, rdr io.Reader, gz bool, onLoadGenome func(CompactGenome)) error {
 437         cgs := []CompactGenome{}
 438         cseqs := []CompactSequence{}
 439         variantmap := map[tileLibRef]tileVariantID{}
 440         err := DecodeLibrary(rdr, gz, func(ent *LibraryEntry) error {
 441                 if ctx.Err() != nil {
 442                         return ctx.Err()
 443                 }
 444                 if err := tilelib.loadTagSet(ent.TagSet); err != nil {
 445                         return err
 446                 }
 447                 if err := tilelib.loadTileVariants(ent.TileVariants, variantmap); err != nil {
 448                         return err
 449                 }
 450                 cgs = append(cgs, ent.CompactGenomes...)
 451                 cseqs = append(cseqs, ent.CompactSequences...)
 452                 return nil
 453         })
 454         if err != nil {
 455                 return err
 456         }
 457         if ctx.Err() != nil {
 458                 return ctx.Err()
 459         }
 460         err = tilelib.loadCompactGenomes(cgs, variantmap, onLoadGenome)
 461         if err != nil {
 462                 return err
 463         }
 464         err = tilelib.loadCompactSequences(cseqs, variantmap)
 465         if err != nil {
 466                 return err
 467         }
 468         return nil
 469 }
 470
 471 func (tilelib *tileLibrary) dump(out io.Writer) {
 472         printTV := func(tag int, variant tileVariantID) {
 473                 if variant < 1 {
 474                         fmt.Fprintf(out, " -")
 475                 } else if tag >= len(tilelib.variant) {
 476                         fmt.Fprintf(out, " (!tag=%d)", tag)
 477                 } else if int(variant) > len(tilelib.variant[tag]) {
 478                         fmt.Fprintf(out, " (tag=%d,!variant=%d)", tag, variant)
 479                 } else {
 480                         fmt.Fprintf(out, " %x", tilelib.variant[tag][variant-1][:8])
 481                 }
 482         }
 483         for refname, refseqs := range tilelib.refseqs {
 484                 for seqname, seq := range refseqs {
 485                         fmt.Fprintf(out, "ref %s %s", refname, seqname)
 486                         for _, libref := range seq {
 487                                 printTV(int(libref.Tag), libref.Variant)
 488                         }
 489                         fmt.Fprintf(out, "\n")
 490                 }
 491         }
 492         for name, cg := range tilelib.compactGenomes {
 493                 fmt.Fprintf(out, "cg %s", name)
 494                 for tag, variant := range cg {
 495                         printTV(tag/2, variant)
 496                 }
 497                 fmt.Fprintf(out, "\n")
 498         }
 499 }
 500
 501 type importStats struct {
 502         InputFile              string
 503         InputLabel             string
 504         InputLength            int
 505         InputCoverage          int
 506         PathLength             int
 507         DroppedOutOfOrderTiles int
 508 }
 509
 510 func (tilelib *tileLibrary) TileFasta(filelabel string, rdr io.Reader, matchChromosome *regexp.Regexp) (tileSeq, []importStats, error) {
 511         ret := tileSeq{}
 512         type jobT struct {
 513                 label string
 514                 fasta []byte
 515         }
 516         todo := make(chan jobT, 1)
 517         scanner := bufio.NewScanner(rdr)
 518         go func() {
 519                 defer close(todo)
 520                 var fasta []byte
 521                 var seqlabel string
 522                 for scanner.Scan() {
 523                         buf := scanner.Bytes()
 524                         if len(buf) > 0 && buf[0] == '>' {
 525                                 todo <- jobT{seqlabel, append([]byte(nil), fasta...)}
 526                                 seqlabel, fasta = strings.SplitN(string(buf[1:]), " ", 2)[0], fasta[:0]
 527                                 log.Debugf("%s %s reading fasta", filelabel, seqlabel)
 528                         } else {
 529                                 fasta = append(fasta, bytes.ToLower(buf)...)
 530                         }
 531                 }
 532                 todo <- jobT{seqlabel, fasta}
 533         }()
 534         type foundtag struct {
 535                 pos   int
 536                 tagid tagID
 537         }
 538         found := make([]foundtag, 2000000)
 539         path := make([]tileLibRef, 2000000)
 540         totalFoundTags := 0
 541         totalPathLen := 0
 542         skippedSequences := 0
 543         taglen := tilelib.taglib.TagLen()
 544         var stats []importStats
 545         for job := range todo {
 546                 if len(job.fasta) == 0 {
 547                         continue
 548                 } else if !matchChromosome.MatchString(job.label) {
 549                         skippedSequences++
 550                         continue
 551                 }
 552                 log.Debugf("%s %s tiling", filelabel, job.label)
 553
 554                 found = found[:0]
 555                 tilelib.taglib.FindAll(job.fasta, func(tagid tagID, pos, taglen int) {
 556                         found = append(found, foundtag{pos: pos, tagid: tagid})
 557                 })
 558                 totalFoundTags += len(found)
 559                 if len(found) == 0 {
 560                         log.Warnf("%s %s no tags found", filelabel, job.label)
 561                 }
 562
 563                 skipped := 0
 564                 if tilelib.skipOOO {
 565                         log.Infof("%s %s keeping longest increasing subsequence", filelabel, job.label)
 566                         keep := longestIncreasingSubsequence(len(found), func(i int) int { return int(found[i].tagid) })
 567                         for i, x := range keep {
 568                                 found[i] = found[x]
 569                         }
 570                         skipped = len(found) - len(keep)
 571                         found = found[:len(keep)]
 572                 }
 573
 574                 log.Infof("%s %s getting %d librefs", filelabel, job.label, len(found))
 575                 throttle := &throttle{Max: runtime.NumCPU()}
 576                 path = path[:len(found)]
 577                 var lowquality int64
 578                 for i, f := range found {
 579                         i, f := i, f
 580                         throttle.Acquire()
 581                         go func() {
 582                                 defer throttle.Release()
 583                                 var startpos, endpos int
 584                                 if i == 0 {
 585                                         startpos = 0
 586                                 } else {
 587                                         startpos = f.pos
 588                                 }
 589                                 if i == len(found)-1 {
 590                                         endpos = len(job.fasta)
 591                                 } else {
 592                                         endpos = found[i+1].pos + taglen
 593                                 }
 594                                 path[i] = tilelib.getRef(f.tagid, job.fasta[startpos:endpos])
 595                                 if countBases(job.fasta[startpos:endpos]) != endpos-startpos {
 596                                         atomic.AddInt64(&lowquality, 1)
 597                                 }
 598                         }()
 599                 }
 600                 throttle.Wait()
 601
 602                 log.Infof("%s %s copying path", filelabel, job.label)
 603
 604                 pathcopy := make([]tileLibRef, len(path))
 605                 copy(pathcopy, path)
 606                 ret[job.label] = pathcopy
 607
 608                 basesIn := countBases(job.fasta)
 609                 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)
 610                 stats = append(stats, importStats{
 611                         InputFile:              filelabel,
 612                         InputLabel:             job.label,
 613                         InputLength:            len(job.fasta),
 614                         InputCoverage:          basesIn,
 615                         PathLength:             len(path),
 616                         DroppedOutOfOrderTiles: skipped,
 617                 })
 618
 619                 totalPathLen += len(path)
 620         }
 621         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)
 622         return ret, stats, scanner.Err()
 623 }
 624
 625 func (tilelib *tileLibrary) Len() int64 {
 626         return atomic.LoadInt64(&tilelib.variants)
 627 }
 628
 629 // Return a tileLibRef for a tile with the given tag and sequence,
 630 // adding the sequence to the library if needed.
 631 func (tilelib *tileLibrary) getRef(tag tagID, seq []byte) tileLibRef {
 632         dropSeq := false
 633         if !tilelib.retainNoCalls {
 634                 for _, b := range seq {
 635                         if b != 'a' && b != 'c' && b != 'g' && b != 't' {
 636                                 dropSeq = true
 637                                 break
 638                         }
 639                 }
 640         }
 641         seqhash := blake2b.Sum256(seq)
 642         var vlock sync.Locker
 643
 644         tilelib.mtx.RLock()
 645         if len(tilelib.vlock) > int(tag) {
 646                 vlock = tilelib.vlock[tag]
 647         }
 648         tilelib.mtx.RUnlock()
 649
 650         if vlock != nil {
 651                 vlock.Lock()
 652                 for i, varhash := range tilelib.variant[tag] {
 653                         if varhash == seqhash {
 654                                 vlock.Unlock()
 655                                 return tileLibRef{Tag: tag, Variant: tileVariantID(i + 1)}
 656                         }
 657                 }
 658                 vlock.Unlock()
 659         } else {
 660                 tilelib.mtx.Lock()
 661                 if tilelib.variant == nil && tilelib.taglib != nil {
 662                         tilelib.variant = make([][][blake2b.Size256]byte, tilelib.taglib.Len())
 663                         tilelib.vlock = make([]sync.Locker, tilelib.taglib.Len())
 664                         for i := range tilelib.vlock {
 665                                 tilelib.vlock[i] = new(sync.Mutex)
 666                         }
 667                 }
 668                 if int(tag) >= len(tilelib.variant) {
 669                         oldlen := len(tilelib.vlock)
 670                         for i := 0; i < oldlen; i++ {
 671                                 tilelib.vlock[i].Lock()
 672                         }
 673                         // If we haven't seen the tag library yet (as
 674                         // in a merge), tilelib.taglib.Len() is
 675                         // zero. We can still behave correctly, we
 676                         // just need to expand the tilelib.variant and
 677                         // tilelib.vlock slices as needed.
 678                         if int(tag) >= cap(tilelib.variant) {
 679                                 // Allocate 2x capacity.
 680                                 newslice := make([][][blake2b.Size256]byte, int(tag)+1, (int(tag)+1)*2)
 681                                 copy(newslice, tilelib.variant)
 682                                 tilelib.variant = newslice[:int(tag)+1]
 683                                 newvlock := make([]sync.Locker, int(tag)+1, (int(tag)+1)*2)
 684                                 copy(newvlock, tilelib.vlock)
 685                                 tilelib.vlock = newvlock[:int(tag)+1]
 686                         } else {
 687                                 // Use previously allocated capacity,
 688                                 // avoiding copy.
 689                                 tilelib.variant = tilelib.variant[:int(tag)+1]
 690                                 tilelib.vlock = tilelib.vlock[:int(tag)+1]
 691                         }
 692                         for i := oldlen; i < len(tilelib.vlock); i++ {
 693                                 tilelib.vlock[i] = new(sync.Mutex)
 694                         }
 695                         for i := 0; i < oldlen; i++ {
 696                                 tilelib.vlock[i].Unlock()
 697                         }
 698                 }
 699                 vlock = tilelib.vlock[tag]
 700                 tilelib.mtx.Unlock()
 701         }
 702
 703         vlock.Lock()
 704         for i, varhash := range tilelib.variant[tag] {
 705                 if varhash == seqhash {
 706                         vlock.Unlock()
 707                         return tileLibRef{Tag: tag, Variant: tileVariantID(i + 1)}
 708                 }
 709         }
 710         atomic.AddInt64(&tilelib.variants, 1)
 711         tilelib.variant[tag] = append(tilelib.variant[tag], seqhash)
 712         variant := tileVariantID(len(tilelib.variant[tag]))
 713         vlock.Unlock()
 714
 715         if tilelib.retainTileSequences && !dropSeq {
 716                 seqCopy := append([]byte(nil), seq...)
 717                 tilelib.mtx.Lock()
 718                 if tilelib.seq == nil {
 719                         tilelib.seq = map[[blake2b.Size256]byte][]byte{}
 720                 }
 721                 tilelib.seq[seqhash] = seqCopy
 722                 tilelib.mtx.Unlock()
 723         }
 724
 725         if tilelib.encoder != nil {
 726                 saveSeq := seq
 727                 if dropSeq {
 728                         // Save the hash, but not the sequence
 729                         saveSeq = nil
 730                 }
 731                 tilelib.encoder.Encode(LibraryEntry{
 732                         TileVariants: []TileVariant{{
 733                                 Tag:      tag,
 734                                 Variant:  variant,
 735                                 Blake2b:  seqhash,
 736                                 Sequence: saveSeq,
 737                         }},
 738                 })
 739         }
 740         return tileLibRef{Tag: tag, Variant: variant}
 741 }
 742
 743 func (tilelib *tileLibrary) TileVariantSequence(libref tileLibRef) []byte {
 744         if libref.Variant == 0 || len(tilelib.variant) <= int(libref.Tag) || len(tilelib.variant[libref.Tag]) < int(libref.Variant) {
 745                 return nil
 746         }
 747         return tilelib.seq[tilelib.variant[libref.Tag][libref.Variant-1]]
 748 }
 749
 750 // Tidy deletes unreferenced tile variants and renumbers variants so
 751 // more common variants have smaller IDs.
 752 func (tilelib *tileLibrary) Tidy() {
 753         log.Print("Tidy: compute inref")
 754         inref := map[tileLibRef]bool{}
 755         for _, refseq := range tilelib.refseqs {
 756                 for _, librefs := range refseq {
 757                         for _, libref := range librefs {
 758                                 inref[libref] = true
 759                         }
 760                 }
 761         }
 762         log.Print("Tidy: compute remap")
 763         remap := make([][]tileVariantID, len(tilelib.variant))
 764         throttle := throttle{Max: runtime.NumCPU() + 1}
 765         for tag, oldvariants := range tilelib.variant {
 766                 tag, oldvariants := tagID(tag), oldvariants
 767                 if tag%1000000 == 0 {
 768                         log.Printf("Tidy: tag %d", tag)
 769                 }
 770                 throttle.Acquire()
 771                 go func() {
 772                         defer throttle.Release()
 773                         uses := make([]int, len(oldvariants))
 774                         for _, cg := range tilelib.compactGenomes {
 775                                 for phase := 0; phase < 2; phase++ {
 776                                         cgi := int(tag)*2 + phase
 777                                         if cgi < len(cg) && cg[cgi] > 0 {
 778                                                 uses[cg[cgi]-1]++
 779                                         }
 780                                 }
 781                         }
 782
 783                         // Compute desired order of variants:
 784                         // neworder[x] == index in oldvariants that
 785                         // should move to position x.
 786                         neworder := make([]int, len(oldvariants))
 787                         for i := range neworder {
 788                                 neworder[i] = i
 789                         }
 790                         sort.Slice(neworder, func(i, j int) bool {
 791                                 if cmp := uses[neworder[i]] - uses[neworder[j]]; cmp != 0 {
 792                                         return cmp > 0
 793                                 } else {
 794                                         return bytes.Compare(oldvariants[neworder[i]][:], oldvariants[neworder[j]][:]) < 0
 795                                 }
 796                         })
 797
 798                         // Replace tilelib.variant[tag] with a new
 799                         // re-ordered slice of hashes, and make a
 800                         // mapping from old to new variant IDs.
 801                         remaptag := make([]tileVariantID, len(oldvariants)+1)
 802                         newvariants := make([][blake2b.Size256]byte, 0, len(neworder))
 803                         for _, oldi := range neworder {
 804                                 if uses[oldi] > 0 || inref[tileLibRef{Tag: tag, Variant: tileVariantID(oldi + 1)}] {
 805                                         newvariants = append(newvariants, oldvariants[oldi])
 806                                         remaptag[oldi+1] = tileVariantID(len(newvariants))
 807                                 }
 808                         }
 809                         tilelib.variant[tag] = newvariants
 810                         remap[tag] = remaptag
 811                 }()
 812         }
 813         throttle.Wait()
 814
 815         // Apply remap to genomes and reference sequences, so they
 816         // refer to the same tile variants using the changed IDs.
 817         log.Print("Tidy: apply remap")
 818         var wg sync.WaitGroup
 819         for _, cg := range tilelib.compactGenomes {
 820                 cg := cg
 821                 wg.Add(1)
 822                 go func() {
 823                         defer wg.Done()
 824                         for idx, variant := range cg {
 825                                 cg[idx] = remap[tagID(idx/2)][variant]
 826                         }
 827                 }()
 828         }
 829         for _, refcs := range tilelib.refseqs {
 830                 for _, refseq := range refcs {
 831                         refseq := refseq
 832                         wg.Add(1)
 833                         go func() {
 834                                 defer wg.Done()
 835                                 for i, tv := range refseq {
 836                                         refseq[i].Variant = remap[tv.Tag][tv.Variant]
 837                                 }
 838                         }()
 839                 }
 840         }
 841         wg.Wait()
 842         log.Print("Tidy: done")
 843 }
 844
 845 func countBases(seq []byte) int {
 846         n := 0
 847         for _, c := range seq {
 848                 if isbase[c] {
 849                         n++
 850                 }
 851         }
 852         return n
 853 }