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