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