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 fi.IsDir() {
 231                                 err = walk(path + "/" + fi.Name())
 232                                 if err != nil {
 233                                         return err
 234                                 }
 235                         } else if strings.HasSuffix(path, ".gob") || strings.HasSuffix(path, ".gob.gz") {
 236                                 files = append(files, path)
 237                         }
 238                 }
 239                 return nil
 240         }
 241         err := walk(path)
 242         if err != nil {
 243                 return err
 244         }
 245         ctx, cancel := context.WithCancel(ctx)
 246         defer cancel()
 247         var mtx sync.Mutex
 248         cgs := []CompactGenome{}
 249         cseqs := []CompactSequence{}
 250         variantmap := map[tileLibRef]tileVariantID{}
 251         errs := make(chan error, len(files))
 252         for _, path := range files {
 253                 path := path
 254                 go func() {
 255                         f, err := open(path)
 256                         if err != nil {
 257                                 errs <- err
 258                                 return
 259                         }
 260                         defer f.Close()
 261                         errs <- DecodeLibrary(f, strings.HasSuffix(path, ".gz"), func(ent *LibraryEntry) error {
 262                                 if ctx.Err() != nil {
 263                                         return ctx.Err()
 264                                 }
 265                                 mtx.Lock()
 266                                 defer mtx.Unlock()
 267                                 if err := tilelib.loadTagSet(ent.TagSet); err != nil {
 268                                         return err
 269                                 }
 270                                 if err := tilelib.loadTileVariants(ent.TileVariants, variantmap); err != nil {
 271                                         return err
 272                                 }
 273                                 cgs = append(cgs, ent.CompactGenomes...)
 274                                 cseqs = append(cseqs, ent.CompactSequences...)
 275                                 return nil
 276                         })
 277                 }()
 278         }
 279         for range files {
 280                 err := <-errs
 281                 if err != nil {
 282                         return err
 283                 }
 284         }
 285         err = tilelib.loadCompactGenomes(cgs, variantmap, onLoadGenome)
 286         if err != nil {
 287                 return err
 288         }
 289         err = tilelib.loadCompactSequences(cseqs, variantmap)
 290         if err != nil {
 291                 return err
 292         }
 293         return nil
 294 }
 295
 296 func (tilelib *tileLibrary) WriteDir(dir string) error {
 297         nfiles := 128
 298         files := make([]*os.File, nfiles)
 299         for i := range files {
 300                 f, err := os.OpenFile(fmt.Sprintf("%s/library.%04d.gob.gz", dir, i), os.O_CREATE|os.O_WRONLY, 0666)
 301                 if err != nil {
 302                         return err
 303                 }
 304                 defer f.Close()
 305                 files[i] = f
 306         }
 307         bufws := make([]*bufio.Writer, nfiles)
 308         for i := range bufws {
 309                 bufws[i] = bufio.NewWriterSize(files[i], 1<<26)
 310         }
 311         zws := make([]*pgzip.Writer, nfiles)
 312         for i := range zws {
 313                 zws[i] = pgzip.NewWriter(bufws[i])
 314                 defer zws[i].Close()
 315         }
 316         encoders := make([]*gob.Encoder, nfiles)
 317         for i := range encoders {
 318                 encoders[i] = gob.NewEncoder(zws[i])
 319         }
 320         ctx, cancel := context.WithCancel(context.Background())
 321         defer cancel()
 322         errs := make(chan error, nfiles)
 323         for start := range files {
 324                 start := start
 325                 go func() {
 326                         err := encoders[start].Encode(LibraryEntry{TagSet: tilelib.taglib.Tags()})
 327                         if err != nil {
 328                                 errs <- err
 329                                 return
 330                         }
 331                         if start == 0 {
 332                                 // For now, just write all the genomes and refs
 333                                 // to the first file
 334                                 for name, cg := range tilelib.compactGenomes {
 335                                         err := encoders[start].Encode(LibraryEntry{CompactGenomes: []CompactGenome{{
 336                                                 Name:     name,
 337                                                 Variants: cg,
 338                                         }}})
 339                                         if err != nil {
 340                                                 errs <- err
 341                                                 return
 342                                         }
 343                                 }
 344                                 for name, tseqs := range tilelib.refseqs {
 345                                         err := encoders[start].Encode(LibraryEntry{CompactSequences: []CompactSequence{{
 346                                                 Name:          name,
 347                                                 TileSequences: tseqs,
 348                                         }}})
 349                                         if err != nil {
 350                                                 errs <- err
 351                                                 return
 352                                         }
 353                                 }
 354                         }
 355                         tvs := []TileVariant{}
 356                         for tag := start; tag < len(tilelib.variant) && ctx.Err() == nil; tag += nfiles {
 357                                 tvs = tvs[:0]
 358                                 for idx, hash := range tilelib.variant[tag] {
 359                                         tvs = append(tvs, TileVariant{
 360                                                 Tag:      tagID(tag),
 361                                                 Variant:  tileVariantID(idx + 1),
 362                                                 Blake2b:  hash,
 363                                                 Sequence: tilelib.seq[hash],
 364                                         })
 365                                 }
 366                                 err := encoders[start].Encode(LibraryEntry{TileVariants: tvs})
 367                                 if err != nil {
 368                                         errs <- err
 369                                         return
 370                                 }
 371                         }
 372                         errs <- nil
 373                 }()
 374         }
 375         for range files {
 376                 err := <-errs
 377                 if err != nil {
 378                         return err
 379                 }
 380         }
 381         for i := range zws {
 382                 err := zws[i].Close()
 383                 if err != nil {
 384                         return err
 385                 }
 386                 err = bufws[i].Flush()
 387                 if err != nil {
 388                         return err
 389                 }
 390                 err = files[i].Close()
 391                 if err != nil {
 392                         return err
 393                 }
 394         }
 395         return nil
 396 }
 397
 398 // Load library data from rdr. Tile variants might be renumbered in
 399 // the process; in that case, genomes variants will be renumbered to
 400 // match.
 401 //
 402 // If onLoadGenome is non-nil, call it on each CompactGenome entry.
 403 func (tilelib *tileLibrary) LoadGob(ctx context.Context, rdr io.Reader, gz bool, onLoadGenome func(CompactGenome)) error {
 404         cgs := []CompactGenome{}
 405         cseqs := []CompactSequence{}
 406         variantmap := map[tileLibRef]tileVariantID{}
 407         err := DecodeLibrary(rdr, gz, func(ent *LibraryEntry) error {
 408                 if ctx.Err() != nil {
 409                         return ctx.Err()
 410                 }
 411                 if err := tilelib.loadTagSet(ent.TagSet); err != nil {
 412                         return err
 413                 }
 414                 if err := tilelib.loadTileVariants(ent.TileVariants, variantmap); err != nil {
 415                         return err
 416                 }
 417                 cgs = append(cgs, ent.CompactGenomes...)
 418                 cseqs = append(cseqs, ent.CompactSequences...)
 419                 return nil
 420         })
 421         if err != nil {
 422                 return err
 423         }
 424         if ctx.Err() != nil {
 425                 return ctx.Err()
 426         }
 427         err = tilelib.loadCompactGenomes(cgs, variantmap, onLoadGenome)
 428         if err != nil {
 429                 return err
 430         }
 431         err = tilelib.loadCompactSequences(cseqs, variantmap)
 432         if err != nil {
 433                 return err
 434         }
 435         return nil
 436 }
 437
 438 func (tilelib *tileLibrary) dump(out io.Writer) {
 439         printTV := func(tag int, variant tileVariantID) {
 440                 if variant < 1 {
 441                         fmt.Fprintf(out, " -")
 442                 } else if tag >= len(tilelib.variant) {
 443                         fmt.Fprintf(out, " (!tag=%d)", tag)
 444                 } else if int(variant) > len(tilelib.variant[tag]) {
 445                         fmt.Fprintf(out, " (tag=%d,!variant=%d)", tag, variant)
 446                 } else {
 447                         fmt.Fprintf(out, " %x", tilelib.variant[tag][variant-1][:8])
 448                 }
 449         }
 450         for refname, refseqs := range tilelib.refseqs {
 451                 for seqname, seq := range refseqs {
 452                         fmt.Fprintf(out, "ref %s %s", refname, seqname)
 453                         for _, libref := range seq {
 454                                 printTV(int(libref.Tag), libref.Variant)
 455                         }
 456                         fmt.Fprintf(out, "\n")
 457                 }
 458         }
 459         for name, cg := range tilelib.compactGenomes {
 460                 fmt.Fprintf(out, "cg %s", name)
 461                 for tag, variant := range cg {
 462                         printTV(tag/2, variant)
 463                 }
 464                 fmt.Fprintf(out, "\n")
 465         }
 466 }
 467
 468 type importStats struct {
 469         InputFile              string
 470         InputLabel             string
 471         InputLength            int
 472         InputCoverage          int
 473         PathLength             int
 474         DroppedOutOfOrderTiles int
 475 }
 476
 477 func (tilelib *tileLibrary) TileFasta(filelabel string, rdr io.Reader, matchChromosome *regexp.Regexp) (tileSeq, []importStats, error) {
 478         ret := tileSeq{}
 479         type jobT struct {
 480                 label string
 481                 fasta []byte
 482         }
 483         todo := make(chan jobT, 1)
 484         scanner := bufio.NewScanner(rdr)
 485         go func() {
 486                 defer close(todo)
 487                 var fasta []byte
 488                 var seqlabel string
 489                 for scanner.Scan() {
 490                         buf := scanner.Bytes()
 491                         if len(buf) > 0 && buf[0] == '>' {
 492                                 todo <- jobT{seqlabel, append([]byte(nil), fasta...)}
 493                                 seqlabel, fasta = strings.SplitN(string(buf[1:]), " ", 2)[0], fasta[:0]
 494                                 log.Debugf("%s %s reading fasta", filelabel, seqlabel)
 495                         } else {
 496                                 fasta = append(fasta, bytes.ToLower(buf)...)
 497                         }
 498                 }
 499                 todo <- jobT{seqlabel, fasta}
 500         }()
 501         type foundtag struct {
 502                 pos   int
 503                 tagid tagID
 504         }
 505         found := make([]foundtag, 2000000)
 506         path := make([]tileLibRef, 2000000)
 507         totalFoundTags := 0
 508         totalPathLen := 0
 509         skippedSequences := 0
 510         taglen := tilelib.taglib.TagLen()
 511         var stats []importStats
 512         for job := range todo {
 513                 if len(job.fasta) == 0 {
 514                         continue
 515                 } else if !matchChromosome.MatchString(job.label) {
 516                         skippedSequences++
 517                         continue
 518                 }
 519                 log.Debugf("%s %s tiling", filelabel, job.label)
 520
 521                 found = found[:0]
 522                 tilelib.taglib.FindAll(job.fasta, func(tagid tagID, pos, taglen int) {
 523                         found = append(found, foundtag{pos: pos, tagid: tagid})
 524                 })
 525                 totalFoundTags += len(found)
 526                 if len(found) == 0 {
 527                         log.Warnf("%s %s no tags found", filelabel, job.label)
 528                 }
 529
 530                 skipped := 0
 531                 if tilelib.skipOOO {
 532                         log.Infof("%s %s keeping longest increasing subsequence", filelabel, job.label)
 533                         keep := longestIncreasingSubsequence(len(found), func(i int) int { return int(found[i].tagid) })
 534                         for i, x := range keep {
 535                                 found[i] = found[x]
 536                         }
 537                         skipped = len(found) - len(keep)
 538                         found = found[:len(keep)]
 539                 }
 540
 541                 log.Infof("%s %s getting %d librefs", filelabel, job.label, len(found))
 542                 throttle := &throttle{Max: runtime.NumCPU()}
 543                 path = path[:len(found)]
 544                 var lowquality int64
 545                 for i, f := range found {
 546                         i, f := i, f
 547                         throttle.Acquire()
 548                         go func() {
 549                                 defer throttle.Release()
 550                                 var startpos, endpos int
 551                                 if i == 0 {
 552                                         startpos = 0
 553                                 } else {
 554                                         startpos = f.pos
 555                                 }
 556                                 if i == len(found)-1 {
 557                                         endpos = len(job.fasta)
 558                                 } else {
 559                                         endpos = found[i+1].pos + taglen
 560                                 }
 561                                 path[i] = tilelib.getRef(f.tagid, job.fasta[startpos:endpos])
 562                                 if countBases(job.fasta[startpos:endpos]) != endpos-startpos {
 563                                         atomic.AddInt64(&lowquality, 1)
 564                                 }
 565                         }()
 566                 }
 567                 throttle.Wait()
 568
 569                 log.Infof("%s %s copying path", filelabel, job.label)
 570
 571                 pathcopy := make([]tileLibRef, len(path))
 572                 copy(pathcopy, path)
 573                 ret[job.label] = pathcopy
 574
 575                 basesIn := countBases(job.fasta)
 576                 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)
 577                 stats = append(stats, importStats{
 578                         InputFile:              filelabel,
 579                         InputLabel:             job.label,
 580                         InputLength:            len(job.fasta),
 581                         InputCoverage:          basesIn,
 582                         PathLength:             len(path),
 583                         DroppedOutOfOrderTiles: skipped,
 584                 })
 585
 586                 totalPathLen += len(path)
 587         }
 588         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)
 589         return ret, stats, scanner.Err()
 590 }
 591
 592 func (tilelib *tileLibrary) Len() int64 {
 593         return atomic.LoadInt64(&tilelib.variants)
 594 }
 595
 596 // Return a tileLibRef for a tile with the given tag and sequence,
 597 // adding the sequence to the library if needed.
 598 func (tilelib *tileLibrary) getRef(tag tagID, seq []byte) tileLibRef {
 599         dropSeq := false
 600         if !tilelib.retainNoCalls {
 601                 for _, b := range seq {
 602                         if b != 'a' && b != 'c' && b != 'g' && b != 't' {
 603                                 dropSeq = true
 604                                 break
 605                         }
 606                 }
 607         }
 608         seqhash := blake2b.Sum256(seq)
 609         var vlock sync.Locker
 610
 611         tilelib.mtx.RLock()
 612         if len(tilelib.vlock) > int(tag) {
 613                 vlock = tilelib.vlock[tag]
 614         }
 615         tilelib.mtx.RUnlock()
 616
 617         if vlock != nil {
 618                 vlock.Lock()
 619                 for i, varhash := range tilelib.variant[tag] {
 620                         if varhash == seqhash {
 621                                 vlock.Unlock()
 622                                 return tileLibRef{Tag: tag, Variant: tileVariantID(i + 1)}
 623                         }
 624                 }
 625                 vlock.Unlock()
 626         } else {
 627                 tilelib.mtx.Lock()
 628                 if tilelib.variant == nil && tilelib.taglib != nil {
 629                         tilelib.variant = make([][][blake2b.Size256]byte, tilelib.taglib.Len())
 630                         tilelib.vlock = make([]sync.Locker, tilelib.taglib.Len())
 631                         for i := range tilelib.vlock {
 632                                 tilelib.vlock[i] = new(sync.Mutex)
 633                         }
 634                 }
 635                 if int(tag) >= len(tilelib.variant) {
 636                         oldlen := len(tilelib.vlock)
 637                         for i := 0; i < oldlen; i++ {
 638                                 tilelib.vlock[i].Lock()
 639                         }
 640                         // If we haven't seen the tag library yet (as
 641                         // in a merge), tilelib.taglib.Len() is
 642                         // zero. We can still behave correctly, we
 643                         // just need to expand the tilelib.variant and
 644                         // tilelib.vlock slices as needed.
 645                         if int(tag) >= cap(tilelib.variant) {
 646                                 // Allocate 2x capacity.
 647                                 newslice := make([][][blake2b.Size256]byte, int(tag)+1, (int(tag)+1)*2)
 648                                 copy(newslice, tilelib.variant)
 649                                 tilelib.variant = newslice[:int(tag)+1]
 650                                 newvlock := make([]sync.Locker, int(tag)+1, (int(tag)+1)*2)
 651                                 copy(newvlock, tilelib.vlock)
 652                                 tilelib.vlock = newvlock[:int(tag)+1]
 653                         } else {
 654                                 // Use previously allocated capacity,
 655                                 // avoiding copy.
 656                                 tilelib.variant = tilelib.variant[:int(tag)+1]
 657                                 tilelib.vlock = tilelib.vlock[:int(tag)+1]
 658                         }
 659                         for i := oldlen; i < len(tilelib.vlock); i++ {
 660                                 tilelib.vlock[i] = new(sync.Mutex)
 661                         }
 662                         for i := 0; i < oldlen; i++ {
 663                                 tilelib.vlock[i].Unlock()
 664                         }
 665                 }
 666                 vlock = tilelib.vlock[tag]
 667                 tilelib.mtx.Unlock()
 668         }
 669
 670         vlock.Lock()
 671         for i, varhash := range tilelib.variant[tag] {
 672                 if varhash == seqhash {
 673                         vlock.Unlock()
 674                         return tileLibRef{Tag: tag, Variant: tileVariantID(i + 1)}
 675                 }
 676         }
 677         atomic.AddInt64(&tilelib.variants, 1)
 678         tilelib.variant[tag] = append(tilelib.variant[tag], seqhash)
 679         variant := tileVariantID(len(tilelib.variant[tag]))
 680         vlock.Unlock()
 681
 682         if tilelib.retainTileSequences && !dropSeq {
 683                 tilelib.mtx.Lock()
 684                 if tilelib.seq == nil {
 685                         tilelib.seq = map[[blake2b.Size256]byte][]byte{}
 686                 }
 687                 tilelib.seq[seqhash] = append([]byte(nil), seq...)
 688                 tilelib.mtx.Unlock()
 689         }
 690
 691         if tilelib.encoder != nil {
 692                 saveSeq := seq
 693                 if dropSeq {
 694                         // Save the hash, but not the sequence
 695                         saveSeq = nil
 696                 }
 697                 tilelib.encoder.Encode(LibraryEntry{
 698                         TileVariants: []TileVariant{{
 699                                 Tag:      tag,
 700                                 Variant:  variant,
 701                                 Blake2b:  seqhash,
 702                                 Sequence: saveSeq,
 703                         }},
 704                 })
 705         }
 706         return tileLibRef{Tag: tag, Variant: variant}
 707 }
 708
 709 func (tilelib *tileLibrary) TileVariantSequence(libref tileLibRef) []byte {
 710         if libref.Variant == 0 || len(tilelib.variant) <= int(libref.Tag) || len(tilelib.variant[libref.Tag]) < int(libref.Variant) {
 711                 return nil
 712         }
 713         return tilelib.seq[tilelib.variant[libref.Tag][libref.Variant-1]]
 714 }
 715
 716 // Tidy deletes unreferenced tile variants and renumbers variants so
 717 // more common variants have smaller IDs.
 718 func (tilelib *tileLibrary) Tidy() {
 719         log.Print("Tidy: compute inref")
 720         inref := map[tileLibRef]bool{}
 721         for _, refseq := range tilelib.refseqs {
 722                 for _, librefs := range refseq {
 723                         for _, libref := range librefs {
 724                                 inref[libref] = true
 725                         }
 726                 }
 727         }
 728         log.Print("Tidy: compute remap")
 729         remap := make([][]tileVariantID, len(tilelib.variant))
 730         throttle := throttle{Max: runtime.NumCPU() + 1}
 731         for tag, oldvariants := range tilelib.variant {
 732                 tag, oldvariants := tagID(tag), oldvariants
 733                 if tag%1000000 == 0 {
 734                         log.Printf("Tidy: tag %d", tag)
 735                 }
 736                 throttle.Acquire()
 737                 go func() {
 738                         defer throttle.Release()
 739                         uses := make([]int, len(oldvariants))
 740                         for _, cg := range tilelib.compactGenomes {
 741                                 for phase := 0; phase < 2; phase++ {
 742                                         cgi := int(tag)*2 + phase
 743                                         if cgi < len(cg) && cg[cgi] > 0 {
 744                                                 uses[cg[cgi]-1]++
 745                                         }
 746                                 }
 747                         }
 748
 749                         // Compute desired order of variants:
 750                         // neworder[x] == index in oldvariants that
 751                         // should move to position x.
 752                         neworder := make([]int, len(oldvariants))
 753                         for i := range neworder {
 754                                 neworder[i] = i
 755                         }
 756                         sort.Slice(neworder, func(i, j int) bool {
 757                                 if cmp := uses[neworder[i]] - uses[neworder[j]]; cmp != 0 {
 758                                         return cmp > 0
 759                                 } else {
 760                                         return bytes.Compare(oldvariants[neworder[i]][:], oldvariants[neworder[j]][:]) < 0
 761                                 }
 762                         })
 763
 764                         // Replace tilelib.variant[tag] with a new
 765                         // re-ordered slice of hashes, and make a
 766                         // mapping from old to new variant IDs.
 767                         remaptag := make([]tileVariantID, len(oldvariants)+1)
 768                         newvariants := make([][blake2b.Size256]byte, 0, len(neworder))
 769                         for _, oldi := range neworder {
 770                                 if uses[oldi] > 0 || inref[tileLibRef{Tag: tag, Variant: tileVariantID(oldi + 1)}] {
 771                                         newvariants = append(newvariants, oldvariants[oldi])
 772                                         remaptag[oldi+1] = tileVariantID(len(newvariants))
 773                                 }
 774                         }
 775                         tilelib.variant[tag] = newvariants
 776                         remap[tag] = remaptag
 777                 }()
 778         }
 779         throttle.Wait()
 780
 781         // Apply remap to genomes and reference sequences, so they
 782         // refer to the same tile variants using the changed IDs.
 783         log.Print("Tidy: apply remap")
 784         var wg sync.WaitGroup
 785         for _, cg := range tilelib.compactGenomes {
 786                 cg := cg
 787                 wg.Add(1)
 788                 go func() {
 789                         defer wg.Done()
 790                         for idx, variant := range cg {
 791                                 cg[idx] = remap[tagID(idx/2)][variant]
 792                         }
 793                 }()
 794         }
 795         for _, refcs := range tilelib.refseqs {
 796                 for _, refseq := range refcs {
 797                         refseq := refseq
 798                         wg.Add(1)
 799                         go func() {
 800                                 defer wg.Done()
 801                                 for i, tv := range refseq {
 802                                         refseq[i].Variant = remap[tv.Tag][tv.Variant]
 803                                 }
 804                         }()
 805                 }
 806         }
 807         wg.Wait()
 808         log.Print("Tidy: done")
 809 }
 810
 811 func countBases(seq []byte) int {
 812         n := 0
 813         for _, c := range seq {
 814                 if isbase[c] {
 815                         n++
 816                 }
 817         }
 818         return n
 819 }