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         seq2           map[[2]byte]map[[blake2b.Size256]byte][]byte
  66         seq2lock       map[[2]byte]sync.Locker
  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         allcgs := make([][]CompactGenome, len(files))
 250         allcseqs := make([][]CompactSequence, len(files))
 251         allvariantmap := make([]map[tileLibRef]tileVariantID, len(files))
 252         errs := make(chan error, len(files))
 253         log.Infof("LoadDir: read %d files", len(files))
 254         for fileno, path := range files {
 255                 fileno, path := fileno, 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
 265                         var variantmap = map[tileLibRef]tileVariantID{}
 266                         var cgs []CompactGenome
 267                         var cseqs []CompactSequence
 268                         errs <- DecodeLibrary(f, strings.HasSuffix(path, ".gz"), func(ent *LibraryEntry) error {
 269                                 if ctx.Err() != nil {
 270                                         return ctx.Err()
 271                                 }
 272                                 if len(ent.TagSet) > 0 {
 273                                         mtx.Lock()
 274                                         if tilelib.taglib == nil || tilelib.taglib.Len() != len(ent.TagSet) {
 275                                                 // load first set of tags, or
 276                                                 // report mismatch if 2 sets
 277                                                 // have different #tags.
 278                                                 if err := tilelib.loadTagSet(ent.TagSet); err != nil {
 279                                                         mtx.Unlock()
 280                                                         return err
 281                                                 }
 282                                         }
 283                                         mtx.Unlock()
 284                                 }
 285                                 for _, tv := range ent.TileVariants {
 286                                         variantmap[tileLibRef{Tag: tv.Tag, Variant: tv.Variant}] = tilelib.getRef(tv.Tag, tv.Sequence).Variant
 287                                 }
 288                                 cgs = append(cgs, ent.CompactGenomes...)
 289                                 cseqs = append(cseqs, ent.CompactSequences...)
 290                                 return nil
 291                         })
 292                         allcgs[fileno] = cgs
 293                         allcseqs[fileno] = cseqs
 294                         allvariantmap[fileno] = variantmap
 295                 }()
 296         }
 297         for range files {
 298                 err := <-errs
 299                 if err != nil {
 300                         return err
 301                 }
 302         }
 303         log.Info("LoadDir: merge variantmap")
 304         variantmap := map[tileLibRef]tileVariantID{}
 305         for _, m := range allvariantmap {
 306                 for k, v := range m {
 307                         variantmap[k] = v
 308                 }
 309         }
 310         log.Info("LoadDir: loadCompactGenomes")
 311         for _, cgs := range allcgs {
 312                 err = tilelib.loadCompactGenomes(cgs, variantmap, onLoadGenome)
 313                 if err != nil {
 314                         return err
 315                 }
 316         }
 317         log.Info("LoadDir: loadCompactSequences")
 318         for _, cseqs := range allcseqs {
 319                 err = tilelib.loadCompactSequences(cseqs, variantmap)
 320                 if err != nil {
 321                         return err
 322                 }
 323         }
 324         log.Info("LoadDir done")
 325         return nil
 326 }
 327
 328 func (tilelib *tileLibrary) WriteDir(dir string) error {
 329         nfiles := 128
 330         files := make([]*os.File, nfiles)
 331         for i := range files {
 332                 f, err := os.OpenFile(fmt.Sprintf("%s/library.%04d.gob.gz", dir, i), os.O_CREATE|os.O_WRONLY, 0666)
 333                 if err != nil {
 334                         return err
 335                 }
 336                 defer f.Close()
 337                 files[i] = f
 338         }
 339         bufws := make([]*bufio.Writer, nfiles)
 340         for i := range bufws {
 341                 bufws[i] = bufio.NewWriterSize(files[i], 1<<26)
 342         }
 343         zws := make([]*pgzip.Writer, nfiles)
 344         for i := range zws {
 345                 zws[i] = pgzip.NewWriter(bufws[i])
 346                 defer zws[i].Close()
 347         }
 348         encoders := make([]*gob.Encoder, nfiles)
 349         for i := range encoders {
 350                 encoders[i] = gob.NewEncoder(zws[i])
 351         }
 352
 353         cgnames := make([]string, 0, len(tilelib.compactGenomes))
 354         for name := range tilelib.compactGenomes {
 355                 cgnames = append(cgnames, name)
 356         }
 357         sort.Strings(cgnames)
 358
 359         log.Infof("WriteDir: writing %d files", nfiles)
 360         ctx, cancel := context.WithCancel(context.Background())
 361         defer cancel()
 362         errs := make(chan error, nfiles)
 363         for start := range files {
 364                 start := start
 365                 go func() {
 366                         err := encoders[start].Encode(LibraryEntry{TagSet: tilelib.taglib.Tags()})
 367                         if err != nil {
 368                                 errs <- err
 369                                 return
 370                         }
 371                         if start == 0 {
 372                                 // For now, just write all the refs to
 373                                 // the first file
 374                                 for name, tseqs := range tilelib.refseqs {
 375                                         err := encoders[start].Encode(LibraryEntry{CompactSequences: []CompactSequence{{
 376                                                 Name:          name,
 377                                                 TileSequences: tseqs,
 378                                         }}})
 379                                         if err != nil {
 380                                                 errs <- err
 381                                                 return
 382                                         }
 383                                 }
 384                         }
 385                         for i := start; i < len(cgnames); i += nfiles {
 386                                 err := encoders[start].Encode(LibraryEntry{CompactGenomes: []CompactGenome{{
 387                                         Name:     cgnames[i],
 388                                         Variants: tilelib.compactGenomes[cgnames[i]],
 389                                 }}})
 390                                 if err != nil {
 391                                         errs <- err
 392                                         return
 393                                 }
 394                         }
 395                         tvs := []TileVariant{}
 396                         for tag := start; tag < len(tilelib.variant) && ctx.Err() == nil; tag += nfiles {
 397                                 tvs = tvs[:0]
 398                                 for idx, hash := range tilelib.variant[tag] {
 399                                         tvs = append(tvs, TileVariant{
 400                                                 Tag:      tagID(tag),
 401                                                 Variant:  tileVariantID(idx + 1),
 402                                                 Blake2b:  hash,
 403                                                 Sequence: tilelib.hashSequence(hash),
 404                                         })
 405                                 }
 406                                 err := encoders[start].Encode(LibraryEntry{TileVariants: tvs})
 407                                 if err != nil {
 408                                         errs <- err
 409                                         return
 410                                 }
 411                         }
 412                         errs <- nil
 413                 }()
 414         }
 415         for range files {
 416                 err := <-errs
 417                 if err != nil {
 418                         return err
 419                 }
 420         }
 421         log.Info("WriteDir: flushing")
 422         for i := range zws {
 423                 err := zws[i].Close()
 424                 if err != nil {
 425                         return err
 426                 }
 427                 err = bufws[i].Flush()
 428                 if err != nil {
 429                         return err
 430                 }
 431                 err = files[i].Close()
 432                 if err != nil {
 433                         return err
 434                 }
 435         }
 436         log.Info("WriteDir: done")
 437         return nil
 438 }
 439
 440 // Load library data from rdr. Tile variants might be renumbered in
 441 // the process; in that case, genomes variants will be renumbered to
 442 // match.
 443 //
 444 // If onLoadGenome is non-nil, call it on each CompactGenome entry.
 445 func (tilelib *tileLibrary) LoadGob(ctx context.Context, rdr io.Reader, gz bool, onLoadGenome func(CompactGenome)) error {
 446         cgs := []CompactGenome{}
 447         cseqs := []CompactSequence{}
 448         variantmap := map[tileLibRef]tileVariantID{}
 449         err := DecodeLibrary(rdr, gz, func(ent *LibraryEntry) error {
 450                 if ctx.Err() != nil {
 451                         return ctx.Err()
 452                 }
 453                 if err := tilelib.loadTagSet(ent.TagSet); err != nil {
 454                         return err
 455                 }
 456                 if err := tilelib.loadTileVariants(ent.TileVariants, variantmap); err != nil {
 457                         return err
 458                 }
 459                 cgs = append(cgs, ent.CompactGenomes...)
 460                 cseqs = append(cseqs, ent.CompactSequences...)
 461                 return nil
 462         })
 463         if err != nil {
 464                 return err
 465         }
 466         if ctx.Err() != nil {
 467                 return ctx.Err()
 468         }
 469         err = tilelib.loadCompactGenomes(cgs, variantmap, onLoadGenome)
 470         if err != nil {
 471                 return err
 472         }
 473         err = tilelib.loadCompactSequences(cseqs, variantmap)
 474         if err != nil {
 475                 return err
 476         }
 477         return nil
 478 }
 479
 480 func (tilelib *tileLibrary) dump(out io.Writer) {
 481         printTV := func(tag int, variant tileVariantID) {
 482                 if variant < 1 {
 483                         fmt.Fprintf(out, " -")
 484                 } else if tag >= len(tilelib.variant) {
 485                         fmt.Fprintf(out, " (!tag=%d)", tag)
 486                 } else if int(variant) > len(tilelib.variant[tag]) {
 487                         fmt.Fprintf(out, " (tag=%d,!variant=%d)", tag, variant)
 488                 } else {
 489                         fmt.Fprintf(out, " %x", tilelib.variant[tag][variant-1][:8])
 490                 }
 491         }
 492         for refname, refseqs := range tilelib.refseqs {
 493                 for seqname, seq := range refseqs {
 494                         fmt.Fprintf(out, "ref %s %s", refname, seqname)
 495                         for _, libref := range seq {
 496                                 printTV(int(libref.Tag), libref.Variant)
 497                         }
 498                         fmt.Fprintf(out, "\n")
 499                 }
 500         }
 501         for name, cg := range tilelib.compactGenomes {
 502                 fmt.Fprintf(out, "cg %s", name)
 503                 for tag, variant := range cg {
 504                         printTV(tag/2, variant)
 505                 }
 506                 fmt.Fprintf(out, "\n")
 507         }
 508 }
 509
 510 type importStats struct {
 511         InputFile              string
 512         InputLabel             string
 513         InputLength            int
 514         InputCoverage          int
 515         PathLength             int
 516         DroppedOutOfOrderTiles int
 517 }
 518
 519 func (tilelib *tileLibrary) TileFasta(filelabel string, rdr io.Reader, matchChromosome *regexp.Regexp) (tileSeq, []importStats, error) {
 520         ret := tileSeq{}
 521         type jobT struct {
 522                 label string
 523                 fasta []byte
 524         }
 525         todo := make(chan jobT, 1)
 526         scanner := bufio.NewScanner(rdr)
 527         go func() {
 528                 defer close(todo)
 529                 var fasta []byte
 530                 var seqlabel string
 531                 for scanner.Scan() {
 532                         buf := scanner.Bytes()
 533                         if len(buf) > 0 && buf[0] == '>' {
 534                                 todo <- jobT{seqlabel, append([]byte(nil), fasta...)}
 535                                 seqlabel, fasta = strings.SplitN(string(buf[1:]), " ", 2)[0], fasta[:0]
 536                                 log.Debugf("%s %s reading fasta", filelabel, seqlabel)
 537                         } else {
 538                                 fasta = append(fasta, bytes.ToLower(buf)...)
 539                         }
 540                 }
 541                 todo <- jobT{seqlabel, fasta}
 542         }()
 543         type foundtag struct {
 544                 pos   int
 545                 tagid tagID
 546         }
 547         found := make([]foundtag, 2000000)
 548         path := make([]tileLibRef, 2000000)
 549         totalFoundTags := 0
 550         totalPathLen := 0
 551         skippedSequences := 0
 552         taglen := tilelib.taglib.TagLen()
 553         var stats []importStats
 554         for job := range todo {
 555                 if len(job.fasta) == 0 {
 556                         continue
 557                 } else if !matchChromosome.MatchString(job.label) {
 558                         skippedSequences++
 559                         continue
 560                 }
 561                 log.Debugf("%s %s tiling", filelabel, job.label)
 562
 563                 found = found[:0]
 564                 tilelib.taglib.FindAll(job.fasta, func(tagid tagID, pos, taglen int) {
 565                         found = append(found, foundtag{pos: pos, tagid: tagid})
 566                 })
 567                 totalFoundTags += len(found)
 568                 if len(found) == 0 {
 569                         log.Warnf("%s %s no tags found", filelabel, job.label)
 570                 }
 571
 572                 skipped := 0
 573                 if tilelib.skipOOO {
 574                         log.Infof("%s %s keeping longest increasing subsequence", filelabel, job.label)
 575                         keep := longestIncreasingSubsequence(len(found), func(i int) int { return int(found[i].tagid) })
 576                         for i, x := range keep {
 577                                 found[i] = found[x]
 578                         }
 579                         skipped = len(found) - len(keep)
 580                         found = found[:len(keep)]
 581                 }
 582
 583                 log.Infof("%s %s getting %d librefs", filelabel, job.label, len(found))
 584                 throttle := &throttle{Max: runtime.NumCPU()}
 585                 path = path[:len(found)]
 586                 var lowquality int64
 587                 for i, f := range found {
 588                         i, f := i, f
 589                         throttle.Acquire()
 590                         go func() {
 591                                 defer throttle.Release()
 592                                 var startpos, endpos int
 593                                 if i == 0 {
 594                                         startpos = 0
 595                                 } else {
 596                                         startpos = f.pos
 597                                 }
 598                                 if i == len(found)-1 {
 599                                         endpos = len(job.fasta)
 600                                 } else {
 601                                         endpos = found[i+1].pos + taglen
 602                                 }
 603                                 path[i] = tilelib.getRef(f.tagid, job.fasta[startpos:endpos])
 604                                 if countBases(job.fasta[startpos:endpos]) != endpos-startpos {
 605                                         atomic.AddInt64(&lowquality, 1)
 606                                 }
 607                         }()
 608                 }
 609                 throttle.Wait()
 610
 611                 log.Infof("%s %s copying path", filelabel, job.label)
 612
 613                 pathcopy := make([]tileLibRef, len(path))
 614                 copy(pathcopy, path)
 615                 ret[job.label] = pathcopy
 616
 617                 basesIn := countBases(job.fasta)
 618                 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)
 619                 stats = append(stats, importStats{
 620                         InputFile:              filelabel,
 621                         InputLabel:             job.label,
 622                         InputLength:            len(job.fasta),
 623                         InputCoverage:          basesIn,
 624                         PathLength:             len(path),
 625                         DroppedOutOfOrderTiles: skipped,
 626                 })
 627
 628                 totalPathLen += len(path)
 629         }
 630         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)
 631         return ret, stats, scanner.Err()
 632 }
 633
 634 func (tilelib *tileLibrary) Len() int64 {
 635         return atomic.LoadInt64(&tilelib.variants)
 636 }
 637
 638 // Return a tileLibRef for a tile with the given tag and sequence,
 639 // adding the sequence to the library if needed.
 640 func (tilelib *tileLibrary) getRef(tag tagID, seq []byte) tileLibRef {
 641         dropSeq := false
 642         if !tilelib.retainNoCalls {
 643                 for _, b := range seq {
 644                         if b != 'a' && b != 'c' && b != 'g' && b != 't' {
 645                                 dropSeq = true
 646                                 break
 647                         }
 648                 }
 649         }
 650         seqhash := blake2b.Sum256(seq)
 651         var vlock sync.Locker
 652
 653         tilelib.mtx.RLock()
 654         if len(tilelib.vlock) > int(tag) {
 655                 vlock = tilelib.vlock[tag]
 656         }
 657         tilelib.mtx.RUnlock()
 658
 659         if vlock != nil {
 660                 vlock.Lock()
 661                 for i, varhash := range tilelib.variant[tag] {
 662                         if varhash == seqhash {
 663                                 vlock.Unlock()
 664                                 return tileLibRef{Tag: tag, Variant: tileVariantID(i + 1)}
 665                         }
 666                 }
 667                 vlock.Unlock()
 668         } else {
 669                 tilelib.mtx.Lock()
 670                 if tilelib.variant == nil && tilelib.taglib != nil {
 671                         tilelib.variant = make([][][blake2b.Size256]byte, tilelib.taglib.Len())
 672                         tilelib.vlock = make([]sync.Locker, tilelib.taglib.Len())
 673                         for i := range tilelib.vlock {
 674                                 tilelib.vlock[i] = new(sync.Mutex)
 675                         }
 676                 }
 677                 if int(tag) >= len(tilelib.variant) {
 678                         oldlen := len(tilelib.vlock)
 679                         for i := 0; i < oldlen; i++ {
 680                                 tilelib.vlock[i].Lock()
 681                         }
 682                         // If we haven't seen the tag library yet (as
 683                         // in a merge), tilelib.taglib.Len() is
 684                         // zero. We can still behave correctly, we
 685                         // just need to expand the tilelib.variant and
 686                         // tilelib.vlock slices as needed.
 687                         if int(tag) >= cap(tilelib.variant) {
 688                                 // Allocate 2x capacity.
 689                                 newslice := make([][][blake2b.Size256]byte, int(tag)+1, (int(tag)+1)*2)
 690                                 copy(newslice, tilelib.variant)
 691                                 tilelib.variant = newslice[:int(tag)+1]
 692                                 newvlock := make([]sync.Locker, int(tag)+1, (int(tag)+1)*2)
 693                                 copy(newvlock, tilelib.vlock)
 694                                 tilelib.vlock = newvlock[:int(tag)+1]
 695                         } else {
 696                                 // Use previously allocated capacity,
 697                                 // avoiding copy.
 698                                 tilelib.variant = tilelib.variant[:int(tag)+1]
 699                                 tilelib.vlock = tilelib.vlock[:int(tag)+1]
 700                         }
 701                         for i := oldlen; i < len(tilelib.vlock); i++ {
 702                                 tilelib.vlock[i] = new(sync.Mutex)
 703                         }
 704                         for i := 0; i < oldlen; i++ {
 705                                 tilelib.vlock[i].Unlock()
 706                         }
 707                 }
 708                 vlock = tilelib.vlock[tag]
 709                 tilelib.mtx.Unlock()
 710         }
 711
 712         vlock.Lock()
 713         for i, varhash := range tilelib.variant[tag] {
 714                 if varhash == seqhash {
 715                         vlock.Unlock()
 716                         return tileLibRef{Tag: tag, Variant: tileVariantID(i + 1)}
 717                 }
 718         }
 719         atomic.AddInt64(&tilelib.variants, 1)
 720         tilelib.variant[tag] = append(tilelib.variant[tag], seqhash)
 721         variant := tileVariantID(len(tilelib.variant[tag]))
 722         vlock.Unlock()
 723
 724         if tilelib.retainTileSequences && !dropSeq {
 725                 seqCopy := append([]byte(nil), seq...)
 726                 if tilelib.seq2 == nil {
 727                         tilelib.mtx.Lock()
 728                         if tilelib.seq2 == nil {
 729                                 tilelib.seq2lock = map[[2]byte]sync.Locker{}
 730                                 m := map[[2]byte]map[[blake2b.Size256]byte][]byte{}
 731                                 var k [2]byte
 732                                 for i := 0; i < 256; i++ {
 733                                         k[0] = byte(i)
 734                                         for j := 0; j < 256; j++ {
 735                                                 k[1] = byte(j)
 736                                                 m[k] = map[[blake2b.Size256]byte][]byte{}
 737                                                 tilelib.seq2lock[k] = &sync.Mutex{}
 738                                         }
 739                                 }
 740                                 tilelib.seq2 = m
 741                         }
 742                         tilelib.mtx.Unlock()
 743                 }
 744                 var k [2]byte
 745                 copy(k[:], seqhash[:])
 746                 locker := tilelib.seq2lock[k]
 747                 locker.Lock()
 748                 tilelib.seq2[k][seqhash] = seqCopy
 749                 locker.Unlock()
 750         }
 751
 752         if tilelib.encoder != nil {
 753                 saveSeq := seq
 754                 if dropSeq {
 755                         // Save the hash, but not the sequence
 756                         saveSeq = nil
 757                 }
 758                 tilelib.encoder.Encode(LibraryEntry{
 759                         TileVariants: []TileVariant{{
 760                                 Tag:      tag,
 761                                 Variant:  variant,
 762                                 Blake2b:  seqhash,
 763                                 Sequence: saveSeq,
 764                         }},
 765                 })
 766         }
 767         return tileLibRef{Tag: tag, Variant: variant}
 768 }
 769
 770 func (tilelib *tileLibrary) hashSequence(hash [blake2b.Size256]byte) []byte {
 771         var partition [2]byte
 772         copy(partition[:], hash[:])
 773         return tilelib.seq2[partition][hash]
 774 }
 775
 776 func (tilelib *tileLibrary) TileVariantSequence(libref tileLibRef) []byte {
 777         if libref.Variant == 0 || len(tilelib.variant) <= int(libref.Tag) || len(tilelib.variant[libref.Tag]) < int(libref.Variant) {
 778                 return nil
 779         }
 780         return tilelib.hashSequence(tilelib.variant[libref.Tag][libref.Variant-1])
 781 }
 782
 783 // Tidy deletes unreferenced tile variants and renumbers variants so
 784 // more common variants have smaller IDs.
 785 func (tilelib *tileLibrary) Tidy() {
 786         log.Print("Tidy: compute inref")
 787         inref := map[tileLibRef]bool{}
 788         for _, refseq := range tilelib.refseqs {
 789                 for _, librefs := range refseq {
 790                         for _, libref := range librefs {
 791                                 inref[libref] = true
 792                         }
 793                 }
 794         }
 795         log.Print("Tidy: compute remap")
 796         remap := make([][]tileVariantID, len(tilelib.variant))
 797         throttle := throttle{Max: runtime.NumCPU() + 1}
 798         for tag, oldvariants := range tilelib.variant {
 799                 tag, oldvariants := tagID(tag), oldvariants
 800                 if tag%1000000 == 0 {
 801                         log.Printf("Tidy: tag %d", tag)
 802                 }
 803                 throttle.Acquire()
 804                 go func() {
 805                         defer throttle.Release()
 806                         uses := make([]int, len(oldvariants))
 807                         for _, cg := range tilelib.compactGenomes {
 808                                 for phase := 0; phase < 2; phase++ {
 809                                         cgi := int(tag)*2 + phase
 810                                         if cgi < len(cg) && cg[cgi] > 0 {
 811                                                 uses[cg[cgi]-1]++
 812                                         }
 813                                 }
 814                         }
 815
 816                         // Compute desired order of variants:
 817                         // neworder[x] == index in oldvariants that
 818                         // should move to position x.
 819                         neworder := make([]int, len(oldvariants))
 820                         for i := range neworder {
 821                                 neworder[i] = i
 822                         }
 823                         sort.Slice(neworder, func(i, j int) bool {
 824                                 if cmp := uses[neworder[i]] - uses[neworder[j]]; cmp != 0 {
 825                                         return cmp > 0
 826                                 } else {
 827                                         return bytes.Compare(oldvariants[neworder[i]][:], oldvariants[neworder[j]][:]) < 0
 828                                 }
 829                         })
 830
 831                         // Replace tilelib.variant[tag] with a new
 832                         // re-ordered slice of hashes, and make a
 833                         // mapping from old to new variant IDs.
 834                         remaptag := make([]tileVariantID, len(oldvariants)+1)
 835                         newvariants := make([][blake2b.Size256]byte, 0, len(neworder))
 836                         for _, oldi := range neworder {
 837                                 if uses[oldi] > 0 || inref[tileLibRef{Tag: tag, Variant: tileVariantID(oldi + 1)}] {
 838                                         newvariants = append(newvariants, oldvariants[oldi])
 839                                         remaptag[oldi+1] = tileVariantID(len(newvariants))
 840                                 }
 841                         }
 842                         tilelib.variant[tag] = newvariants
 843                         remap[tag] = remaptag
 844                 }()
 845         }
 846         throttle.Wait()
 847
 848         // Apply remap to genomes and reference sequences, so they
 849         // refer to the same tile variants using the changed IDs.
 850         log.Print("Tidy: apply remap")
 851         var wg sync.WaitGroup
 852         for _, cg := range tilelib.compactGenomes {
 853                 cg := cg
 854                 wg.Add(1)
 855                 go func() {
 856                         defer wg.Done()
 857                         for idx, variant := range cg {
 858                                 cg[idx] = remap[tagID(idx/2)][variant]
 859                         }
 860                 }()
 861         }
 862         for _, refcs := range tilelib.refseqs {
 863                 for _, refseq := range refcs {
 864                         refseq := refseq
 865                         wg.Add(1)
 866                         go func() {
 867                                 defer wg.Done()
 868                                 for i, tv := range refseq {
 869                                         refseq[i].Variant = remap[tv.Tag][tv.Variant]
 870                                 }
 871                         }()
 872                 }
 873         }
 874         wg.Wait()
 875         log.Print("Tidy: done")
 876 }
 877
 878 func countBases(seq []byte) int {
 879         n := 0
 880         for _, c := range seq {
 881                 if isbase[c] {
 882                         n++
 883                 }
 884         }
 885         return n
 886 }