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.Infof("loadCompactSequences: %d todo", len(cseqs))
 178         for _, cseq := range cseqs {
 179                 log.Infof("loadCompactSequences: checking %s", cseq.Name)
 180                 for _, tseq := range cseq.TileSequences {
 181                         for i, libref := range tseq {
 182                                 if libref.Variant == 0 {
 183                                         // No variant (e.g., import
 184                                         // dropped tiles with
 185                                         // no-calls) = no translation.
 186                                         continue
 187                                 }
 188                                 v, ok := variantmap[libref]
 189                                 if !ok {
 190                                         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)
 191                                 }
 192                                 tseq[i].Variant = v
 193                         }
 194                 }
 195                 if tilelib.encoder != nil {
 196                         if err := tilelib.encoder.Encode(LibraryEntry{
 197                                 CompactSequences: []CompactSequence{cseq},
 198                         }); err != nil {
 199                                 return err
 200                         }
 201                 }
 202                 log.Infof("loadCompactSequences: checking %s done", cseq.Name)
 203         }
 204         tilelib.mtx.Lock()
 205         defer tilelib.mtx.Unlock()
 206         if tilelib.refseqs == nil {
 207                 tilelib.refseqs = map[string]map[string][]tileLibRef{}
 208         }
 209         for _, cseq := range cseqs {
 210                 tilelib.refseqs[cseq.Name] = cseq.TileSequences
 211         }
 212         log.Info("loadCompactSequences: done")
 213         return nil
 214 }
 215
 216 func (tilelib *tileLibrary) LoadDir(ctx context.Context, path string, onLoadGenome func(CompactGenome)) error {
 217         var files []string
 218         var walk func(string) error
 219         walk = func(path string) error {
 220                 f, err := open(path)
 221                 if err != nil {
 222                         return err
 223                 }
 224                 defer f.Close()
 225                 fis, err := f.Readdir(-1)
 226                 if err != nil {
 227                         files = append(files, path)
 228                         return nil
 229                 }
 230                 for _, fi := range fis {
 231                         if fi.Name() == "." || fi.Name() == ".." {
 232                                 continue
 233                         } else if child := path + "/" + fi.Name(); fi.IsDir() {
 234                                 err = walk(child)
 235                                 if err != nil {
 236                                         return err
 237                                 }
 238                         } else if strings.HasSuffix(child, ".gob") || strings.HasSuffix(child, ".gob.gz") {
 239                                 files = append(files, child)
 240                         }
 241                 }
 242                 return nil
 243         }
 244         log.Infof("LoadDir: walk dir %s", path)
 245         err := walk(path)
 246         if err != nil {
 247                 return err
 248         }
 249         ctx, cancel := context.WithCancel(ctx)
 250         defer cancel()
 251         var mtx sync.Mutex
 252         allcgs := make([][]CompactGenome, len(files))
 253         allcseqs := make([][]CompactSequence, len(files))
 254         allvariantmap := map[tileLibRef]tileVariantID{}
 255         errs := make(chan error, len(files))
 256         log.Infof("LoadDir: read %d files", len(files))
 257         for fileno, path := range files {
 258                 fileno, path := fileno, path
 259                 go func() {
 260                         f, err := open(path)
 261                         if err != nil {
 262                                 errs <- err
 263                                 return
 264                         }
 265                         defer f.Close()
 266                         defer log.Infof("LoadDir: finished reading %s", path)
 267
 268                         var variantmap = map[tileLibRef]tileVariantID{}
 269                         var cgs []CompactGenome
 270                         var cseqs []CompactSequence
 271                         err = DecodeLibrary(f, strings.HasSuffix(path, ".gz"), func(ent *LibraryEntry) error {
 272                                 if ctx.Err() != nil {
 273                                         return ctx.Err()
 274                                 }
 275                                 if len(ent.TagSet) > 0 {
 276                                         mtx.Lock()
 277                                         if tilelib.taglib == nil || tilelib.taglib.Len() != len(ent.TagSet) {
 278                                                 // load first set of tags, or
 279                                                 // report mismatch if 2 sets
 280                                                 // have different #tags.
 281                                                 if err := tilelib.loadTagSet(ent.TagSet); err != nil {
 282                                                         mtx.Unlock()
 283                                                         return err
 284                                                 }
 285                                         }
 286                                         mtx.Unlock()
 287                                 }
 288                                 for _, tv := range ent.TileVariants {
 289                                         variantmap[tileLibRef{Tag: tv.Tag, Variant: tv.Variant}] = tilelib.getRef(tv.Tag, tv.Sequence).Variant
 290                                 }
 291                                 cgs = append(cgs, ent.CompactGenomes...)
 292                                 cseqs = append(cseqs, ent.CompactSequences...)
 293                                 return nil
 294                         })
 295                         allcgs[fileno] = cgs
 296                         allcseqs[fileno] = cseqs
 297                         mtx.Lock()
 298                         defer mtx.Unlock()
 299                         for k, v := range variantmap {
 300                                 allvariantmap[k] = v
 301                         }
 302                         errs <- err
 303                 }()
 304         }
 305         for range files {
 306                 err := <-errs
 307                 if err != nil {
 308                         return err
 309                 }
 310         }
 311
 312         log.Info("LoadDir: loadCompactGenomes")
 313         var flatcgs []CompactGenome
 314         for _, cgs := range allcgs {
 315                 flatcgs = append(flatcgs, cgs...)
 316         }
 317         err = tilelib.loadCompactGenomes(flatcgs, allvariantmap, onLoadGenome)
 318         if err != nil {
 319                 return err
 320         }
 321
 322         log.Info("LoadDir: loadCompactSequences")
 323         var flatcseqs []CompactSequence
 324         for _, cseqs := range allcseqs {
 325                 flatcseqs = append(flatcseqs, cseqs...)
 326         }
 327         err = tilelib.loadCompactSequences(flatcseqs, allvariantmap)
 328         if err != nil {
 329                 return err
 330         }
 331
 332         log.Info("LoadDir done")
 333         return nil
 334 }
 335
 336 func (tilelib *tileLibrary) WriteDir(dir string) error {
 337         ntilefiles := 128
 338         nfiles := ntilefiles + len(tilelib.refseqs)
 339         files := make([]*os.File, nfiles)
 340         for i := range files {
 341                 f, err := os.OpenFile(fmt.Sprintf("%s/library.%04d.gob.gz", dir, i), os.O_CREATE|os.O_WRONLY, 0666)
 342                 if err != nil {
 343                         return err
 344                 }
 345                 defer f.Close()
 346                 files[i] = f
 347         }
 348         bufws := make([]*bufio.Writer, nfiles)
 349         for i := range bufws {
 350                 bufws[i] = bufio.NewWriterSize(files[i], 1<<26)
 351         }
 352         zws := make([]*pgzip.Writer, nfiles)
 353         for i := range zws {
 354                 zws[i] = pgzip.NewWriter(bufws[i])
 355                 defer zws[i].Close()
 356         }
 357         encoders := make([]*gob.Encoder, nfiles)
 358         for i := range encoders {
 359                 encoders[i] = gob.NewEncoder(zws[i])
 360         }
 361
 362         cgnames := make([]string, 0, len(tilelib.compactGenomes))
 363         for name := range tilelib.compactGenomes {
 364                 cgnames = append(cgnames, name)
 365         }
 366         sort.Strings(cgnames)
 367
 368         refnames := make([]string, 0, len(tilelib.refseqs))
 369         for name := range tilelib.refseqs {
 370                 refnames = append(refnames, name)
 371         }
 372         sort.Strings(refnames)
 373
 374         log.Infof("WriteDir: writing %d files", nfiles)
 375         ctx, cancel := context.WithCancel(context.Background())
 376         defer cancel()
 377         errs := make(chan error, nfiles)
 378         for start := range files {
 379                 start := start
 380                 go func() {
 381                         err := encoders[start].Encode(LibraryEntry{TagSet: tilelib.taglib.Tags()})
 382                         if err != nil {
 383                                 errs <- err
 384                                 return
 385                         }
 386                         if refidx := start - ntilefiles; refidx >= 0 {
 387                                 // write each ref to its own file
 388                                 // (they seem to load very slowly)
 389                                 name := refnames[refidx]
 390                                 errs <- encoders[start].Encode(LibraryEntry{CompactSequences: []CompactSequence{{
 391                                         Name:          name,
 392                                         TileSequences: tilelib.refseqs[name],
 393                                 }}})
 394                                 return
 395                         }
 396                         for i := start; i < len(cgnames); i += ntilefiles {
 397                                 err := encoders[start].Encode(LibraryEntry{CompactGenomes: []CompactGenome{{
 398                                         Name:     cgnames[i],
 399                                         Variants: tilelib.compactGenomes[cgnames[i]],
 400                                 }}})
 401                                 if err != nil {
 402                                         errs <- err
 403                                         return
 404                                 }
 405                         }
 406                         tvs := []TileVariant{}
 407                         for tag := start; tag < len(tilelib.variant) && ctx.Err() == nil; tag += ntilefiles {
 408                                 tvs = tvs[:0]
 409                                 for idx, hash := range tilelib.variant[tag] {
 410                                         tvs = append(tvs, TileVariant{
 411                                                 Tag:      tagID(tag),
 412                                                 Variant:  tileVariantID(idx + 1),
 413                                                 Blake2b:  hash,
 414                                                 Sequence: tilelib.hashSequence(hash),
 415                                         })
 416                                 }
 417                                 err := encoders[start].Encode(LibraryEntry{TileVariants: tvs})
 418                                 if err != nil {
 419                                         errs <- err
 420                                         return
 421                                 }
 422                         }
 423                         errs <- nil
 424                 }()
 425         }
 426         for range files {
 427                 err := <-errs
 428                 if err != nil {
 429                         return err
 430                 }
 431         }
 432         log.Info("WriteDir: flushing")
 433         for i := range zws {
 434                 err := zws[i].Close()
 435                 if err != nil {
 436                         return err
 437                 }
 438                 err = bufws[i].Flush()
 439                 if err != nil {
 440                         return err
 441                 }
 442                 err = files[i].Close()
 443                 if err != nil {
 444                         return err
 445                 }
 446         }
 447         log.Info("WriteDir: done")
 448         return nil
 449 }
 450
 451 // Load library data from rdr. Tile variants might be renumbered in
 452 // the process; in that case, genomes variants will be renumbered to
 453 // match.
 454 //
 455 // If onLoadGenome is non-nil, call it on each CompactGenome entry.
 456 func (tilelib *tileLibrary) LoadGob(ctx context.Context, rdr io.Reader, gz bool, onLoadGenome func(CompactGenome)) error {
 457         cgs := []CompactGenome{}
 458         cseqs := []CompactSequence{}
 459         variantmap := map[tileLibRef]tileVariantID{}
 460         err := DecodeLibrary(rdr, gz, func(ent *LibraryEntry) error {
 461                 if ctx.Err() != nil {
 462                         return ctx.Err()
 463                 }
 464                 if err := tilelib.loadTagSet(ent.TagSet); err != nil {
 465                         return err
 466                 }
 467                 if err := tilelib.loadTileVariants(ent.TileVariants, variantmap); err != nil {
 468                         return err
 469                 }
 470                 cgs = append(cgs, ent.CompactGenomes...)
 471                 cseqs = append(cseqs, ent.CompactSequences...)
 472                 return nil
 473         })
 474         if err != nil {
 475                 return err
 476         }
 477         if ctx.Err() != nil {
 478                 return ctx.Err()
 479         }
 480         err = tilelib.loadCompactGenomes(cgs, variantmap, onLoadGenome)
 481         if err != nil {
 482                 return err
 483         }
 484         err = tilelib.loadCompactSequences(cseqs, variantmap)
 485         if err != nil {
 486                 return err
 487         }
 488         return nil
 489 }
 490
 491 func (tilelib *tileLibrary) dump(out io.Writer) {
 492         printTV := func(tag int, variant tileVariantID) {
 493                 if variant < 1 {
 494                         fmt.Fprintf(out, " -")
 495                 } else if tag >= len(tilelib.variant) {
 496                         fmt.Fprintf(out, " (!tag=%d)", tag)
 497                 } else if int(variant) > len(tilelib.variant[tag]) {
 498                         fmt.Fprintf(out, " (tag=%d,!variant=%d)", tag, variant)
 499                 } else {
 500                         fmt.Fprintf(out, " %x", tilelib.variant[tag][variant-1][:8])
 501                 }
 502         }
 503         for refname, refseqs := range tilelib.refseqs {
 504                 for seqname, seq := range refseqs {
 505                         fmt.Fprintf(out, "ref %s %s", refname, seqname)
 506                         for _, libref := range seq {
 507                                 printTV(int(libref.Tag), libref.Variant)
 508                         }
 509                         fmt.Fprintf(out, "\n")
 510                 }
 511         }
 512         for name, cg := range tilelib.compactGenomes {
 513                 fmt.Fprintf(out, "cg %s", name)
 514                 for tag, variant := range cg {
 515                         printTV(tag/2, variant)
 516                 }
 517                 fmt.Fprintf(out, "\n")
 518         }
 519 }
 520
 521 type importStats struct {
 522         InputFile              string
 523         InputLabel             string
 524         InputLength            int
 525         InputCoverage          int
 526         PathLength             int
 527         DroppedOutOfOrderTiles int
 528 }
 529
 530 func (tilelib *tileLibrary) TileFasta(filelabel string, rdr io.Reader, matchChromosome *regexp.Regexp) (tileSeq, []importStats, error) {
 531         ret := tileSeq{}
 532         type jobT struct {
 533                 label string
 534                 fasta []byte
 535         }
 536         todo := make(chan jobT, 1)
 537         scanner := bufio.NewScanner(rdr)
 538         go func() {
 539                 defer close(todo)
 540                 var fasta []byte
 541                 var seqlabel string
 542                 for scanner.Scan() {
 543                         buf := scanner.Bytes()
 544                         if len(buf) > 0 && buf[0] == '>' {
 545                                 todo <- jobT{seqlabel, append([]byte(nil), fasta...)}
 546                                 seqlabel, fasta = strings.SplitN(string(buf[1:]), " ", 2)[0], fasta[:0]
 547                                 log.Debugf("%s %s reading fasta", filelabel, seqlabel)
 548                         } else {
 549                                 fasta = append(fasta, bytes.ToLower(buf)...)
 550                         }
 551                 }
 552                 todo <- jobT{seqlabel, fasta}
 553         }()
 554         type foundtag struct {
 555                 pos   int
 556                 tagid tagID
 557         }
 558         found := make([]foundtag, 2000000)
 559         path := make([]tileLibRef, 2000000)
 560         totalFoundTags := 0
 561         totalPathLen := 0
 562         skippedSequences := 0
 563         taglen := tilelib.taglib.TagLen()
 564         var stats []importStats
 565         for job := range todo {
 566                 if len(job.fasta) == 0 {
 567                         continue
 568                 } else if !matchChromosome.MatchString(job.label) {
 569                         skippedSequences++
 570                         continue
 571                 }
 572                 log.Debugf("%s %s tiling", filelabel, job.label)
 573
 574                 found = found[:0]
 575                 tilelib.taglib.FindAll(job.fasta, func(tagid tagID, pos, taglen int) {
 576                         found = append(found, foundtag{pos: pos, tagid: tagid})
 577                 })
 578                 totalFoundTags += len(found)
 579                 if len(found) == 0 {
 580                         log.Warnf("%s %s no tags found", filelabel, job.label)
 581                 }
 582
 583                 skipped := 0
 584                 if tilelib.skipOOO {
 585                         log.Infof("%s %s keeping longest increasing subsequence", filelabel, job.label)
 586                         keep := longestIncreasingSubsequence(len(found), func(i int) int { return int(found[i].tagid) })
 587                         for i, x := range keep {
 588                                 found[i] = found[x]
 589                         }
 590                         skipped = len(found) - len(keep)
 591                         found = found[:len(keep)]
 592                 }
 593
 594                 log.Infof("%s %s getting %d librefs", filelabel, job.label, len(found))
 595                 throttle := &throttle{Max: runtime.NumCPU()}
 596                 path = path[:len(found)]
 597                 var lowquality int64
 598                 for i, f := range found {
 599                         i, f := i, f
 600                         throttle.Acquire()
 601                         go func() {
 602                                 defer throttle.Release()
 603                                 var startpos, endpos int
 604                                 if i == 0 {
 605                                         startpos = 0
 606                                 } else {
 607                                         startpos = f.pos
 608                                 }
 609                                 if i == len(found)-1 {
 610                                         endpos = len(job.fasta)
 611                                 } else {
 612                                         endpos = found[i+1].pos + taglen
 613                                 }
 614                                 path[i] = tilelib.getRef(f.tagid, job.fasta[startpos:endpos])
 615                                 if countBases(job.fasta[startpos:endpos]) != endpos-startpos {
 616                                         atomic.AddInt64(&lowquality, 1)
 617                                 }
 618                         }()
 619                 }
 620                 throttle.Wait()
 621
 622                 log.Infof("%s %s copying path", filelabel, job.label)
 623
 624                 pathcopy := make([]tileLibRef, len(path))
 625                 copy(pathcopy, path)
 626                 ret[job.label] = pathcopy
 627
 628                 basesIn := countBases(job.fasta)
 629                 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)
 630                 stats = append(stats, importStats{
 631                         InputFile:              filelabel,
 632                         InputLabel:             job.label,
 633                         InputLength:            len(job.fasta),
 634                         InputCoverage:          basesIn,
 635                         PathLength:             len(path),
 636                         DroppedOutOfOrderTiles: skipped,
 637                 })
 638
 639                 totalPathLen += len(path)
 640         }
 641         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)
 642         return ret, stats, scanner.Err()
 643 }
 644
 645 func (tilelib *tileLibrary) Len() int64 {
 646         return atomic.LoadInt64(&tilelib.variants)
 647 }
 648
 649 // Return a tileLibRef for a tile with the given tag and sequence,
 650 // adding the sequence to the library if needed.
 651 func (tilelib *tileLibrary) getRef(tag tagID, seq []byte) tileLibRef {
 652         dropSeq := false
 653         if !tilelib.retainNoCalls {
 654                 for _, b := range seq {
 655                         if b != 'a' && b != 'c' && b != 'g' && b != 't' {
 656                                 dropSeq = true
 657                                 break
 658                         }
 659                 }
 660         }
 661         seqhash := blake2b.Sum256(seq)
 662         var vlock sync.Locker
 663
 664         tilelib.mtx.RLock()
 665         if len(tilelib.vlock) > int(tag) {
 666                 vlock = tilelib.vlock[tag]
 667         }
 668         tilelib.mtx.RUnlock()
 669
 670         if vlock != nil {
 671                 vlock.Lock()
 672                 for i, varhash := range tilelib.variant[tag] {
 673                         if varhash == seqhash {
 674                                 vlock.Unlock()
 675                                 return tileLibRef{Tag: tag, Variant: tileVariantID(i + 1)}
 676                         }
 677                 }
 678                 vlock.Unlock()
 679         } else {
 680                 tilelib.mtx.Lock()
 681                 if tilelib.variant == nil && tilelib.taglib != nil {
 682                         tilelib.variant = make([][][blake2b.Size256]byte, tilelib.taglib.Len())
 683                         tilelib.vlock = make([]sync.Locker, tilelib.taglib.Len())
 684                         for i := range tilelib.vlock {
 685                                 tilelib.vlock[i] = new(sync.Mutex)
 686                         }
 687                 }
 688                 if int(tag) >= len(tilelib.variant) {
 689                         oldlen := len(tilelib.vlock)
 690                         for i := 0; i < oldlen; i++ {
 691                                 tilelib.vlock[i].Lock()
 692                         }
 693                         // If we haven't seen the tag library yet (as
 694                         // in a merge), tilelib.taglib.Len() is
 695                         // zero. We can still behave correctly, we
 696                         // just need to expand the tilelib.variant and
 697                         // tilelib.vlock slices as needed.
 698                         if int(tag) >= cap(tilelib.variant) {
 699                                 // Allocate 2x capacity.
 700                                 newslice := make([][][blake2b.Size256]byte, int(tag)+1, (int(tag)+1)*2)
 701                                 copy(newslice, tilelib.variant)
 702                                 tilelib.variant = newslice[:int(tag)+1]
 703                                 newvlock := make([]sync.Locker, int(tag)+1, (int(tag)+1)*2)
 704                                 copy(newvlock, tilelib.vlock)
 705                                 tilelib.vlock = newvlock[:int(tag)+1]
 706                         } else {
 707                                 // Use previously allocated capacity,
 708                                 // avoiding copy.
 709                                 tilelib.variant = tilelib.variant[:int(tag)+1]
 710                                 tilelib.vlock = tilelib.vlock[:int(tag)+1]
 711                         }
 712                         for i := oldlen; i < len(tilelib.vlock); i++ {
 713                                 tilelib.vlock[i] = new(sync.Mutex)
 714                         }
 715                         for i := 0; i < oldlen; i++ {
 716                                 tilelib.vlock[i].Unlock()
 717                         }
 718                 }
 719                 vlock = tilelib.vlock[tag]
 720                 tilelib.mtx.Unlock()
 721         }
 722
 723         vlock.Lock()
 724         for i, varhash := range tilelib.variant[tag] {
 725                 if varhash == seqhash {
 726                         vlock.Unlock()
 727                         return tileLibRef{Tag: tag, Variant: tileVariantID(i + 1)}
 728                 }
 729         }
 730         atomic.AddInt64(&tilelib.variants, 1)
 731         tilelib.variant[tag] = append(tilelib.variant[tag], seqhash)
 732         variant := tileVariantID(len(tilelib.variant[tag]))
 733         vlock.Unlock()
 734
 735         if tilelib.retainTileSequences && !dropSeq {
 736                 seqCopy := append([]byte(nil), seq...)
 737                 if tilelib.seq2 == nil {
 738                         tilelib.mtx.Lock()
 739                         if tilelib.seq2 == nil {
 740                                 tilelib.seq2lock = map[[2]byte]sync.Locker{}
 741                                 m := map[[2]byte]map[[blake2b.Size256]byte][]byte{}
 742                                 var k [2]byte
 743                                 for i := 0; i < 256; i++ {
 744                                         k[0] = byte(i)
 745                                         for j := 0; j < 256; j++ {
 746                                                 k[1] = byte(j)
 747                                                 m[k] = map[[blake2b.Size256]byte][]byte{}
 748                                                 tilelib.seq2lock[k] = &sync.Mutex{}
 749                                         }
 750                                 }
 751                                 tilelib.seq2 = m
 752                         }
 753                         tilelib.mtx.Unlock()
 754                 }
 755                 var k [2]byte
 756                 copy(k[:], seqhash[:])
 757                 locker := tilelib.seq2lock[k]
 758                 locker.Lock()
 759                 tilelib.seq2[k][seqhash] = seqCopy
 760                 locker.Unlock()
 761         }
 762
 763         if tilelib.encoder != nil {
 764                 saveSeq := seq
 765                 if dropSeq {
 766                         // Save the hash, but not the sequence
 767                         saveSeq = nil
 768                 }
 769                 tilelib.encoder.Encode(LibraryEntry{
 770                         TileVariants: []TileVariant{{
 771                                 Tag:      tag,
 772                                 Variant:  variant,
 773                                 Blake2b:  seqhash,
 774                                 Sequence: saveSeq,
 775                         }},
 776                 })
 777         }
 778         return tileLibRef{Tag: tag, Variant: variant}
 779 }
 780
 781 func (tilelib *tileLibrary) hashSequence(hash [blake2b.Size256]byte) []byte {
 782         var partition [2]byte
 783         copy(partition[:], hash[:])
 784         return tilelib.seq2[partition][hash]
 785 }
 786
 787 func (tilelib *tileLibrary) TileVariantSequence(libref tileLibRef) []byte {
 788         if libref.Variant == 0 || len(tilelib.variant) <= int(libref.Tag) || len(tilelib.variant[libref.Tag]) < int(libref.Variant) {
 789                 return nil
 790         }
 791         return tilelib.hashSequence(tilelib.variant[libref.Tag][libref.Variant-1])
 792 }
 793
 794 // Tidy deletes unreferenced tile variants and renumbers variants so
 795 // more common variants have smaller IDs.
 796 func (tilelib *tileLibrary) Tidy() {
 797         log.Print("Tidy: compute inref")
 798         inref := map[tileLibRef]bool{}
 799         for _, refseq := range tilelib.refseqs {
 800                 for _, librefs := range refseq {
 801                         for _, libref := range librefs {
 802                                 inref[libref] = true
 803                         }
 804                 }
 805         }
 806         log.Print("Tidy: compute remap")
 807         remap := make([][]tileVariantID, len(tilelib.variant))
 808         throttle := throttle{Max: runtime.NumCPU() + 1}
 809         for tag, oldvariants := range tilelib.variant {
 810                 tag, oldvariants := tagID(tag), oldvariants
 811                 if tag%1000000 == 0 {
 812                         log.Printf("Tidy: tag %d", tag)
 813                 }
 814                 throttle.Acquire()
 815                 go func() {
 816                         defer throttle.Release()
 817                         uses := make([]int, len(oldvariants))
 818                         for _, cg := range tilelib.compactGenomes {
 819                                 for phase := 0; phase < 2; phase++ {
 820                                         cgi := int(tag)*2 + phase
 821                                         if cgi < len(cg) && cg[cgi] > 0 {
 822                                                 uses[cg[cgi]-1]++
 823                                         }
 824                                 }
 825                         }
 826
 827                         // Compute desired order of variants:
 828                         // neworder[x] == index in oldvariants that
 829                         // should move to position x.
 830                         neworder := make([]int, len(oldvariants))
 831                         for i := range neworder {
 832                                 neworder[i] = i
 833                         }
 834                         sort.Slice(neworder, func(i, j int) bool {
 835                                 if cmp := uses[neworder[i]] - uses[neworder[j]]; cmp != 0 {
 836                                         return cmp > 0
 837                                 } else {
 838                                         return bytes.Compare(oldvariants[neworder[i]][:], oldvariants[neworder[j]][:]) < 0
 839                                 }
 840                         })
 841
 842                         // Replace tilelib.variant[tag] with a new
 843                         // re-ordered slice of hashes, and make a
 844                         // mapping from old to new variant IDs.
 845                         remaptag := make([]tileVariantID, len(oldvariants)+1)
 846                         newvariants := make([][blake2b.Size256]byte, 0, len(neworder))
 847                         for _, oldi := range neworder {
 848                                 if uses[oldi] > 0 || inref[tileLibRef{Tag: tag, Variant: tileVariantID(oldi + 1)}] {
 849                                         newvariants = append(newvariants, oldvariants[oldi])
 850                                         remaptag[oldi+1] = tileVariantID(len(newvariants))
 851                                 }
 852                         }
 853                         tilelib.variant[tag] = newvariants
 854                         remap[tag] = remaptag
 855                 }()
 856         }
 857         throttle.Wait()
 858
 859         // Apply remap to genomes and reference sequences, so they
 860         // refer to the same tile variants using the changed IDs.
 861         log.Print("Tidy: apply remap")
 862         var wg sync.WaitGroup
 863         for _, cg := range tilelib.compactGenomes {
 864                 cg := cg
 865                 wg.Add(1)
 866                 go func() {
 867                         defer wg.Done()
 868                         for idx, variant := range cg {
 869                                 cg[idx] = remap[tagID(idx/2)][variant]
 870                         }
 871                 }()
 872         }
 873         for _, refcs := range tilelib.refseqs {
 874                 for _, refseq := range refcs {
 875                         refseq := refseq
 876                         wg.Add(1)
 877                         go func() {
 878                                 defer wg.Done()
 879                                 for i, tv := range refseq {
 880                                         refseq[i].Variant = remap[tv.Tag][tv.Variant]
 881                                 }
 882                         }()
 883                 }
 884         }
 885         wg.Wait()
 886         log.Print("Tidy: done")
 887 }
 888
 889 func countBases(seq []byte) int {
 890         n := 0
 891         for _, c := range seq {
 892                 if isbase[c] {
 893                         n++
 894                 }
 895         }
 896         return n
 897 }