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         for _, path := range files {
 255                 path := path
 256                 go func() {
 257                         f, err := open(path)
 258                         if err != nil {
 259                                 errs <- err
 260                                 return
 261                         }
 262                         defer f.Close()
 263                         defer log.Infof("LoadDir: finished reading %s", path)
 264                         errs <- DecodeLibrary(f, strings.HasSuffix(path, ".gz"), func(ent *LibraryEntry) error {
 265                                 if ctx.Err() != nil {
 266                                         return ctx.Err()
 267                                 }
 268                                 if len(ent.TagSet) > 0 {
 269                                         mtx.Lock()
 270                                         if tilelib.taglib == nil || tilelib.taglib.Len() != len(ent.TagSet) {
 271                                                 // load first set of tags, or
 272                                                 // report mismatch if 2 sets
 273                                                 // have different #tags.
 274                                                 if err := tilelib.loadTagSet(ent.TagSet); err != nil {
 275                                                         mtx.Unlock()
 276                                                         return err
 277                                                 }
 278                                         }
 279                                         mtx.Unlock()
 280                                 }
 281                                 variantmapadd := map[tileLibRef]tileVariantID{}
 282                                 for _, tv := range ent.TileVariants {
 283                                         variantmapadd[tileLibRef{Tag: tv.Tag, Variant: tv.Variant}] = tilelib.getRef(tv.Tag, tv.Sequence).Variant
 284                                 }
 285                                 mtx.Lock()
 286                                 cgs = append(cgs, ent.CompactGenomes...)
 287                                 cseqs = append(cseqs, ent.CompactSequences...)
 288                                 for k, v := range variantmapadd {
 289                                         variantmap[k] = v
 290                                 }
 291                                 mtx.Unlock()
 292                                 return nil
 293                         })
 294                 }()
 295         }
 296         for range files {
 297                 err := <-errs
 298                 if err != nil {
 299                         return err
 300                 }
 301         }
 302         log.Info("LoadDir: loadCompactGenomes")
 303         err = tilelib.loadCompactGenomes(cgs, variantmap, onLoadGenome)
 304         if err != nil {
 305                 return err
 306         }
 307         log.Info("LoadDir: loadCompactSequences")
 308         err = tilelib.loadCompactSequences(cseqs, variantmap)
 309         if err != nil {
 310                 return err
 311         }
 312         log.Info("LoadDir done")
 313         return nil
 314 }
 315
 316 func (tilelib *tileLibrary) WriteDir(dir string) error {
 317         nfiles := 128
 318         files := make([]*os.File, nfiles)
 319         for i := range files {
 320                 f, err := os.OpenFile(fmt.Sprintf("%s/library.%04d.gob.gz", dir, i), os.O_CREATE|os.O_WRONLY, 0666)
 321                 if err != nil {
 322                         return err
 323                 }
 324                 defer f.Close()
 325                 files[i] = f
 326         }
 327         bufws := make([]*bufio.Writer, nfiles)
 328         for i := range bufws {
 329                 bufws[i] = bufio.NewWriterSize(files[i], 1<<26)
 330         }
 331         zws := make([]*pgzip.Writer, nfiles)
 332         for i := range zws {
 333                 zws[i] = pgzip.NewWriter(bufws[i])
 334                 defer zws[i].Close()
 335         }
 336         encoders := make([]*gob.Encoder, nfiles)
 337         for i := range encoders {
 338                 encoders[i] = gob.NewEncoder(zws[i])
 339         }
 340
 341         log.Infof("WriteDir: writing %d files", nfiles)
 342         ctx, cancel := context.WithCancel(context.Background())
 343         defer cancel()
 344         errs := make(chan error, nfiles)
 345         for start := range files {
 346                 start := start
 347                 go func() {
 348                         err := encoders[start].Encode(LibraryEntry{TagSet: tilelib.taglib.Tags()})
 349                         if err != nil {
 350                                 errs <- err
 351                                 return
 352                         }
 353                         if start == 0 {
 354                                 // For now, just write all the genomes and refs
 355                                 // to the first file
 356                                 for name, cg := range tilelib.compactGenomes {
 357                                         err := encoders[start].Encode(LibraryEntry{CompactGenomes: []CompactGenome{{
 358                                                 Name:     name,
 359                                                 Variants: cg,
 360                                         }}})
 361                                         if err != nil {
 362                                                 errs <- err
 363                                                 return
 364                                         }
 365                                 }
 366                                 for name, tseqs := range tilelib.refseqs {
 367                                         err := encoders[start].Encode(LibraryEntry{CompactSequences: []CompactSequence{{
 368                                                 Name:          name,
 369                                                 TileSequences: tseqs,
 370                                         }}})
 371                                         if err != nil {
 372                                                 errs <- err
 373                                                 return
 374                                         }
 375                                 }
 376                         }
 377                         tvs := []TileVariant{}
 378                         for tag := start; tag < len(tilelib.variant) && ctx.Err() == nil; tag += nfiles {
 379                                 tvs = tvs[:0]
 380                                 for idx, hash := range tilelib.variant[tag] {
 381                                         tvs = append(tvs, TileVariant{
 382                                                 Tag:      tagID(tag),
 383                                                 Variant:  tileVariantID(idx + 1),
 384                                                 Blake2b:  hash,
 385                                                 Sequence: tilelib.seq[hash],
 386                                         })
 387                                 }
 388                                 err := encoders[start].Encode(LibraryEntry{TileVariants: tvs})
 389                                 if err != nil {
 390                                         errs <- err
 391                                         return
 392                                 }
 393                         }
 394                         errs <- nil
 395                 }()
 396         }
 397         for range files {
 398                 err := <-errs
 399                 if err != nil {
 400                         return err
 401                 }
 402         }
 403         log.Info("WriteDir: flushing")
 404         for i := range zws {
 405                 err := zws[i].Close()
 406                 if err != nil {
 407                         return err
 408                 }
 409                 err = bufws[i].Flush()
 410                 if err != nil {
 411                         return err
 412                 }
 413                 err = files[i].Close()
 414                 if err != nil {
 415                         return err
 416                 }
 417         }
 418         log.Info("WriteDir: done")
 419         return nil
 420 }
 421
 422 // Load library data from rdr. Tile variants might be renumbered in
 423 // the process; in that case, genomes variants will be renumbered to
 424 // match.
 425 //
 426 // If onLoadGenome is non-nil, call it on each CompactGenome entry.
 427 func (tilelib *tileLibrary) LoadGob(ctx context.Context, rdr io.Reader, gz bool, onLoadGenome func(CompactGenome)) error {
 428         cgs := []CompactGenome{}
 429         cseqs := []CompactSequence{}
 430         variantmap := map[tileLibRef]tileVariantID{}
 431         err := DecodeLibrary(rdr, gz, func(ent *LibraryEntry) error {
 432                 if ctx.Err() != nil {
 433                         return ctx.Err()
 434                 }
 435                 if err := tilelib.loadTagSet(ent.TagSet); err != nil {
 436                         return err
 437                 }
 438                 if err := tilelib.loadTileVariants(ent.TileVariants, variantmap); err != nil {
 439                         return err
 440                 }
 441                 cgs = append(cgs, ent.CompactGenomes...)
 442                 cseqs = append(cseqs, ent.CompactSequences...)
 443                 return nil
 444         })
 445         if err != nil {
 446                 return err
 447         }
 448         if ctx.Err() != nil {
 449                 return ctx.Err()
 450         }
 451         err = tilelib.loadCompactGenomes(cgs, variantmap, onLoadGenome)
 452         if err != nil {
 453                 return err
 454         }
 455         err = tilelib.loadCompactSequences(cseqs, variantmap)
 456         if err != nil {
 457                 return err
 458         }
 459         return nil
 460 }
 461
 462 func (tilelib *tileLibrary) dump(out io.Writer) {
 463         printTV := func(tag int, variant tileVariantID) {
 464                 if variant < 1 {
 465                         fmt.Fprintf(out, " -")
 466                 } else if tag >= len(tilelib.variant) {
 467                         fmt.Fprintf(out, " (!tag=%d)", tag)
 468                 } else if int(variant) > len(tilelib.variant[tag]) {
 469                         fmt.Fprintf(out, " (tag=%d,!variant=%d)", tag, variant)
 470                 } else {
 471                         fmt.Fprintf(out, " %x", tilelib.variant[tag][variant-1][:8])
 472                 }
 473         }
 474         for refname, refseqs := range tilelib.refseqs {
 475                 for seqname, seq := range refseqs {
 476                         fmt.Fprintf(out, "ref %s %s", refname, seqname)
 477                         for _, libref := range seq {
 478                                 printTV(int(libref.Tag), libref.Variant)
 479                         }
 480                         fmt.Fprintf(out, "\n")
 481                 }
 482         }
 483         for name, cg := range tilelib.compactGenomes {
 484                 fmt.Fprintf(out, "cg %s", name)
 485                 for tag, variant := range cg {
 486                         printTV(tag/2, variant)
 487                 }
 488                 fmt.Fprintf(out, "\n")
 489         }
 490 }
 491
 492 type importStats struct {
 493         InputFile              string
 494         InputLabel             string
 495         InputLength            int
 496         InputCoverage          int
 497         PathLength             int
 498         DroppedOutOfOrderTiles int
 499 }
 500
 501 func (tilelib *tileLibrary) TileFasta(filelabel string, rdr io.Reader, matchChromosome *regexp.Regexp) (tileSeq, []importStats, error) {
 502         ret := tileSeq{}
 503         type jobT struct {
 504                 label string
 505                 fasta []byte
 506         }
 507         todo := make(chan jobT, 1)
 508         scanner := bufio.NewScanner(rdr)
 509         go func() {
 510                 defer close(todo)
 511                 var fasta []byte
 512                 var seqlabel string
 513                 for scanner.Scan() {
 514                         buf := scanner.Bytes()
 515                         if len(buf) > 0 && buf[0] == '>' {
 516                                 todo <- jobT{seqlabel, append([]byte(nil), fasta...)}
 517                                 seqlabel, fasta = strings.SplitN(string(buf[1:]), " ", 2)[0], fasta[:0]
 518                                 log.Debugf("%s %s reading fasta", filelabel, seqlabel)
 519                         } else {
 520                                 fasta = append(fasta, bytes.ToLower(buf)...)
 521                         }
 522                 }
 523                 todo <- jobT{seqlabel, fasta}
 524         }()
 525         type foundtag struct {
 526                 pos   int
 527                 tagid tagID
 528         }
 529         found := make([]foundtag, 2000000)
 530         path := make([]tileLibRef, 2000000)
 531         totalFoundTags := 0
 532         totalPathLen := 0
 533         skippedSequences := 0
 534         taglen := tilelib.taglib.TagLen()
 535         var stats []importStats
 536         for job := range todo {
 537                 if len(job.fasta) == 0 {
 538                         continue
 539                 } else if !matchChromosome.MatchString(job.label) {
 540                         skippedSequences++
 541                         continue
 542                 }
 543                 log.Debugf("%s %s tiling", filelabel, job.label)
 544
 545                 found = found[:0]
 546                 tilelib.taglib.FindAll(job.fasta, func(tagid tagID, pos, taglen int) {
 547                         found = append(found, foundtag{pos: pos, tagid: tagid})
 548                 })
 549                 totalFoundTags += len(found)
 550                 if len(found) == 0 {
 551                         log.Warnf("%s %s no tags found", filelabel, job.label)
 552                 }
 553
 554                 skipped := 0
 555                 if tilelib.skipOOO {
 556                         log.Infof("%s %s keeping longest increasing subsequence", filelabel, job.label)
 557                         keep := longestIncreasingSubsequence(len(found), func(i int) int { return int(found[i].tagid) })
 558                         for i, x := range keep {
 559                                 found[i] = found[x]
 560                         }
 561                         skipped = len(found) - len(keep)
 562                         found = found[:len(keep)]
 563                 }
 564
 565                 log.Infof("%s %s getting %d librefs", filelabel, job.label, len(found))
 566                 throttle := &throttle{Max: runtime.NumCPU()}
 567                 path = path[:len(found)]
 568                 var lowquality int64
 569                 for i, f := range found {
 570                         i, f := i, f
 571                         throttle.Acquire()
 572                         go func() {
 573                                 defer throttle.Release()
 574                                 var startpos, endpos int
 575                                 if i == 0 {
 576                                         startpos = 0
 577                                 } else {
 578                                         startpos = f.pos
 579                                 }
 580                                 if i == len(found)-1 {
 581                                         endpos = len(job.fasta)
 582                                 } else {
 583                                         endpos = found[i+1].pos + taglen
 584                                 }
 585                                 path[i] = tilelib.getRef(f.tagid, job.fasta[startpos:endpos])
 586                                 if countBases(job.fasta[startpos:endpos]) != endpos-startpos {
 587                                         atomic.AddInt64(&lowquality, 1)
 588                                 }
 589                         }()
 590                 }
 591                 throttle.Wait()
 592
 593                 log.Infof("%s %s copying path", filelabel, job.label)
 594
 595                 pathcopy := make([]tileLibRef, len(path))
 596                 copy(pathcopy, path)
 597                 ret[job.label] = pathcopy
 598
 599                 basesIn := countBases(job.fasta)
 600                 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)
 601                 stats = append(stats, importStats{
 602                         InputFile:              filelabel,
 603                         InputLabel:             job.label,
 604                         InputLength:            len(job.fasta),
 605                         InputCoverage:          basesIn,
 606                         PathLength:             len(path),
 607                         DroppedOutOfOrderTiles: skipped,
 608                 })
 609
 610                 totalPathLen += len(path)
 611         }
 612         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)
 613         return ret, stats, scanner.Err()
 614 }
 615
 616 func (tilelib *tileLibrary) Len() int64 {
 617         return atomic.LoadInt64(&tilelib.variants)
 618 }
 619
 620 // Return a tileLibRef for a tile with the given tag and sequence,
 621 // adding the sequence to the library if needed.
 622 func (tilelib *tileLibrary) getRef(tag tagID, seq []byte) tileLibRef {
 623         dropSeq := false
 624         if !tilelib.retainNoCalls {
 625                 for _, b := range seq {
 626                         if b != 'a' && b != 'c' && b != 'g' && b != 't' {
 627                                 dropSeq = true
 628                                 break
 629                         }
 630                 }
 631         }
 632         seqhash := blake2b.Sum256(seq)
 633         var vlock sync.Locker
 634
 635         tilelib.mtx.RLock()
 636         if len(tilelib.vlock) > int(tag) {
 637                 vlock = tilelib.vlock[tag]
 638         }
 639         tilelib.mtx.RUnlock()
 640
 641         if vlock != nil {
 642                 vlock.Lock()
 643                 for i, varhash := range tilelib.variant[tag] {
 644                         if varhash == seqhash {
 645                                 vlock.Unlock()
 646                                 return tileLibRef{Tag: tag, Variant: tileVariantID(i + 1)}
 647                         }
 648                 }
 649                 vlock.Unlock()
 650         } else {
 651                 tilelib.mtx.Lock()
 652                 if tilelib.variant == nil && tilelib.taglib != nil {
 653                         tilelib.variant = make([][][blake2b.Size256]byte, tilelib.taglib.Len())
 654                         tilelib.vlock = make([]sync.Locker, tilelib.taglib.Len())
 655                         for i := range tilelib.vlock {
 656                                 tilelib.vlock[i] = new(sync.Mutex)
 657                         }
 658                 }
 659                 if int(tag) >= len(tilelib.variant) {
 660                         oldlen := len(tilelib.vlock)
 661                         for i := 0; i < oldlen; i++ {
 662                                 tilelib.vlock[i].Lock()
 663                         }
 664                         // If we haven't seen the tag library yet (as
 665                         // in a merge), tilelib.taglib.Len() is
 666                         // zero. We can still behave correctly, we
 667                         // just need to expand the tilelib.variant and
 668                         // tilelib.vlock slices as needed.
 669                         if int(tag) >= cap(tilelib.variant) {
 670                                 // Allocate 2x capacity.
 671                                 newslice := make([][][blake2b.Size256]byte, int(tag)+1, (int(tag)+1)*2)
 672                                 copy(newslice, tilelib.variant)
 673                                 tilelib.variant = newslice[:int(tag)+1]
 674                                 newvlock := make([]sync.Locker, int(tag)+1, (int(tag)+1)*2)
 675                                 copy(newvlock, tilelib.vlock)
 676                                 tilelib.vlock = newvlock[:int(tag)+1]
 677                         } else {
 678                                 // Use previously allocated capacity,
 679                                 // avoiding copy.
 680                                 tilelib.variant = tilelib.variant[:int(tag)+1]
 681                                 tilelib.vlock = tilelib.vlock[:int(tag)+1]
 682                         }
 683                         for i := oldlen; i < len(tilelib.vlock); i++ {
 684                                 tilelib.vlock[i] = new(sync.Mutex)
 685                         }
 686                         for i := 0; i < oldlen; i++ {
 687                                 tilelib.vlock[i].Unlock()
 688                         }
 689                 }
 690                 vlock = tilelib.vlock[tag]
 691                 tilelib.mtx.Unlock()
 692         }
 693
 694         vlock.Lock()
 695         for i, varhash := range tilelib.variant[tag] {
 696                 if varhash == seqhash {
 697                         vlock.Unlock()
 698                         return tileLibRef{Tag: tag, Variant: tileVariantID(i + 1)}
 699                 }
 700         }
 701         atomic.AddInt64(&tilelib.variants, 1)
 702         tilelib.variant[tag] = append(tilelib.variant[tag], seqhash)
 703         variant := tileVariantID(len(tilelib.variant[tag]))
 704         vlock.Unlock()
 705
 706         if tilelib.retainTileSequences && !dropSeq {
 707                 tilelib.mtx.Lock()
 708                 if tilelib.seq == nil {
 709                         tilelib.seq = map[[blake2b.Size256]byte][]byte{}
 710                 }
 711                 tilelib.seq[seqhash] = append([]byte(nil), seq...)
 712                 tilelib.mtx.Unlock()
 713         }
 714
 715         if tilelib.encoder != nil {
 716                 saveSeq := seq
 717                 if dropSeq {
 718                         // Save the hash, but not the sequence
 719                         saveSeq = nil
 720                 }
 721                 tilelib.encoder.Encode(LibraryEntry{
 722                         TileVariants: []TileVariant{{
 723                                 Tag:      tag,
 724                                 Variant:  variant,
 725                                 Blake2b:  seqhash,
 726                                 Sequence: saveSeq,
 727                         }},
 728                 })
 729         }
 730         return tileLibRef{Tag: tag, Variant: variant}
 731 }
 732
 733 func (tilelib *tileLibrary) TileVariantSequence(libref tileLibRef) []byte {
 734         if libref.Variant == 0 || len(tilelib.variant) <= int(libref.Tag) || len(tilelib.variant[libref.Tag]) < int(libref.Variant) {
 735                 return nil
 736         }
 737         return tilelib.seq[tilelib.variant[libref.Tag][libref.Variant-1]]
 738 }
 739
 740 // Tidy deletes unreferenced tile variants and renumbers variants so
 741 // more common variants have smaller IDs.
 742 func (tilelib *tileLibrary) Tidy() {
 743         log.Print("Tidy: compute inref")
 744         inref := map[tileLibRef]bool{}
 745         for _, refseq := range tilelib.refseqs {
 746                 for _, librefs := range refseq {
 747                         for _, libref := range librefs {
 748                                 inref[libref] = true
 749                         }
 750                 }
 751         }
 752         log.Print("Tidy: compute remap")
 753         remap := make([][]tileVariantID, len(tilelib.variant))
 754         throttle := throttle{Max: runtime.NumCPU() + 1}
 755         for tag, oldvariants := range tilelib.variant {
 756                 tag, oldvariants := tagID(tag), oldvariants
 757                 if tag%1000000 == 0 {
 758                         log.Printf("Tidy: tag %d", tag)
 759                 }
 760                 throttle.Acquire()
 761                 go func() {
 762                         defer throttle.Release()
 763                         uses := make([]int, len(oldvariants))
 764                         for _, cg := range tilelib.compactGenomes {
 765                                 for phase := 0; phase < 2; phase++ {
 766                                         cgi := int(tag)*2 + phase
 767                                         if cgi < len(cg) && cg[cgi] > 0 {
 768                                                 uses[cg[cgi]-1]++
 769                                         }
 770                                 }
 771                         }
 772
 773                         // Compute desired order of variants:
 774                         // neworder[x] == index in oldvariants that
 775                         // should move to position x.
 776                         neworder := make([]int, len(oldvariants))
 777                         for i := range neworder {
 778                                 neworder[i] = i
 779                         }
 780                         sort.Slice(neworder, func(i, j int) bool {
 781                                 if cmp := uses[neworder[i]] - uses[neworder[j]]; cmp != 0 {
 782                                         return cmp > 0
 783                                 } else {
 784                                         return bytes.Compare(oldvariants[neworder[i]][:], oldvariants[neworder[j]][:]) < 0
 785                                 }
 786                         })
 787
 788                         // Replace tilelib.variant[tag] with a new
 789                         // re-ordered slice of hashes, and make a
 790                         // mapping from old to new variant IDs.
 791                         remaptag := make([]tileVariantID, len(oldvariants)+1)
 792                         newvariants := make([][blake2b.Size256]byte, 0, len(neworder))
 793                         for _, oldi := range neworder {
 794                                 if uses[oldi] > 0 || inref[tileLibRef{Tag: tag, Variant: tileVariantID(oldi + 1)}] {
 795                                         newvariants = append(newvariants, oldvariants[oldi])
 796                                         remaptag[oldi+1] = tileVariantID(len(newvariants))
 797                                 }
 798                         }
 799                         tilelib.variant[tag] = newvariants
 800                         remap[tag] = remaptag
 801                 }()
 802         }
 803         throttle.Wait()
 804
 805         // Apply remap to genomes and reference sequences, so they
 806         // refer to the same tile variants using the changed IDs.
 807         log.Print("Tidy: apply remap")
 808         var wg sync.WaitGroup
 809         for _, cg := range tilelib.compactGenomes {
 810                 cg := cg
 811                 wg.Add(1)
 812                 go func() {
 813                         defer wg.Done()
 814                         for idx, variant := range cg {
 815                                 cg[idx] = remap[tagID(idx/2)][variant]
 816                         }
 817                 }()
 818         }
 819         for _, refcs := range tilelib.refseqs {
 820                 for _, refseq := range refcs {
 821                         refseq := refseq
 822                         wg.Add(1)
 823                         go func() {
 824                                 defer wg.Done()
 825                                 for i, tv := range refseq {
 826                                         refseq[i].Variant = remap[tv.Tag][tv.Variant]
 827                                 }
 828                         }()
 829                 }
 830         }
 831         wg.Wait()
 832         log.Print("Tidy: done")
 833 }
 834
 835 func countBases(seq []byte) int {
 836         n := 0
 837         for _, c := range seq {
 838                 if isbase[c] {
 839                         n++
 840                 }
 841         }
 842         return n
 843 }