tilelib.go

   1 // Copyright (C) The Lightning Authors. All rights reserved.
   2 //
   3 // SPDX-License-Identifier: AGPL-3.0
   4
   5 package lightning
   6
   7 import (
   8         "bufio"
   9         "bytes"
  10         "context"
  11         "encoding/gob"
  12         "fmt"
  13         "io"
  14         "os"
  15         "regexp"
  16         "runtime"
  17         "sort"
  18         "strings"
  19         "sync"
  20         "sync/atomic"
  21
  22         "github.com/klauspost/pgzip"
  23         log "github.com/sirupsen/logrus"
  24         "golang.org/x/crypto/blake2b"
  25 )
  26
  27 type tileVariantID uint16 // 1-based
  28
  29 type tileLibRef struct {
  30         Tag     tagID
  31         Variant tileVariantID
  32 }
  33
  34 type tileSeq map[string][]tileLibRef
  35
  36 func (tseq tileSeq) Variants() ([]tileVariantID, int, int) {
  37         maxtag := 0
  38         for _, refs := range tseq {
  39                 for _, ref := range refs {
  40                         if maxtag < int(ref.Tag) {
  41                                 maxtag = int(ref.Tag)
  42                         }
  43                 }
  44         }
  45         vars := make([]tileVariantID, maxtag+1)
  46         var kept, dropped int
  47         for _, refs := range tseq {
  48                 for _, ref := range refs {
  49                         if vars[int(ref.Tag)] != 0 {
  50                                 dropped++
  51                         } else {
  52                                 kept++
  53                         }
  54                         vars[int(ref.Tag)] = ref.Variant
  55                 }
  56         }
  57         return vars, kept, dropped
  58 }
  59
  60 type tileLibrary struct {
  61         retainNoCalls       bool
  62         skipOOO             bool
  63         retainTileSequences bool
  64         useDups             bool
  65
  66         taglib         *tagLibrary
  67         variant        [][][blake2b.Size256]byte
  68         refseqs        map[string]map[string][]tileLibRef
  69         compactGenomes map[string][]tileVariantID
  70         seq2           map[[2]byte]map[[blake2b.Size256]byte][]byte
  71         seq2lock       map[[2]byte]sync.Locker
  72         variants       int64
  73         // if non-nil, write out any tile variants added while tiling
  74         encoder *gob.Encoder
  75         // set Ref flag when writing new variants to encoder
  76         encodeRef bool
  77
  78         onAddTileVariant func(libref tileLibRef, hash [blake2b.Size256]byte, seq []byte) error
  79         onAddGenome      func(CompactGenome) error
  80         onAddRefseq      func(CompactSequence) error
  81
  82         mtx   sync.RWMutex
  83         vlock []sync.Locker
  84 }
  85
  86 func (tilelib *tileLibrary) loadTagSet(newtagset [][]byte) error {
  87         // Loading a tagset means either passing it through to the
  88         // output (if it's the first one we've seen), or just ensuring
  89         // it doesn't disagree with what we already have.
  90         if len(newtagset) == 0 {
  91                 return nil
  92         }
  93         tilelib.mtx.Lock()
  94         defer tilelib.mtx.Unlock()
  95         if tilelib.taglib == nil || tilelib.taglib.Len() == 0 {
  96                 tilelib.taglib = &tagLibrary{}
  97                 err := tilelib.taglib.setTags(newtagset)
  98                 if err != nil {
  99                         return err
 100                 }
 101                 if tilelib.encoder != nil {
 102                         err = tilelib.encoder.Encode(LibraryEntry{
 103                                 TagSet: newtagset,
 104                         })
 105                         if err != nil {
 106                                 return err
 107                         }
 108                 }
 109         } else if tilelib.taglib.Len() != len(newtagset) {
 110                 return fmt.Errorf("cannot merge libraries with differing tagsets")
 111         } else {
 112                 current := tilelib.taglib.Tags()
 113                 for i := range newtagset {
 114                         if !bytes.Equal(newtagset[i], current[i]) {
 115                                 return fmt.Errorf("cannot merge libraries with differing tagsets")
 116                         }
 117                 }
 118         }
 119         return nil
 120 }
 121
 122 func (tilelib *tileLibrary) loadTileVariants(tvs []TileVariant, variantmap map[tileLibRef]tileVariantID) error {
 123         for _, tv := range tvs {
 124                 // Assign a new variant ID (unique across all inputs)
 125                 // for each input variant.
 126                 variantmap[tileLibRef{Tag: tv.Tag, Variant: tv.Variant}] = tilelib.getRef(tv.Tag, tv.Sequence, tv.Ref).Variant
 127         }
 128         return nil
 129 }
 130
 131 func (tilelib *tileLibrary) loadCompactGenomes(cgs []CompactGenome, variantmap map[tileLibRef]tileVariantID) error {
 132         log.Debugf("loadCompactGenomes: %d", len(cgs))
 133         var wg sync.WaitGroup
 134         errs := make(chan error, 1)
 135         for _, cg := range cgs {
 136                 wg.Add(1)
 137                 cg := cg
 138                 go func() {
 139                         defer wg.Done()
 140                         for i, variant := range cg.Variants {
 141                                 if len(errs) > 0 {
 142                                         return
 143                                 }
 144                                 if variant == 0 {
 145                                         continue
 146                                 }
 147                                 tag := tagID(i / 2)
 148                                 newvariant, ok := variantmap[tileLibRef{Tag: tag, Variant: variant}]
 149                                 if !ok {
 150                                         err := fmt.Errorf("oops: genome %q has variant %d for tag %d, but that variant was not in its library", cg.Name, variant, tag)
 151                                         select {
 152                                         case errs <- err:
 153                                         default:
 154                                         }
 155                                         return
 156                                 }
 157                                 // log.Tracef("loadCompactGenomes: cg %s tag %d variant %d => %d", cg.Name, tag, variant, newvariant)
 158                                 cg.Variants[i] = newvariant
 159                         }
 160                         if tilelib.onAddGenome != nil {
 161                                 err := tilelib.onAddGenome(cg)
 162                                 if err != nil {
 163                                         select {
 164                                         case errs <- err:
 165                                         default:
 166                                         }
 167                                         return
 168                                 }
 169                         }
 170                         if tilelib.encoder != nil {
 171                                 err := tilelib.encoder.Encode(LibraryEntry{
 172                                         CompactGenomes: []CompactGenome{cg},
 173                                 })
 174                                 if err != nil {
 175                                         select {
 176                                         case errs <- err:
 177                                         default:
 178                                         }
 179                                         return
 180                                 }
 181                         }
 182                         if tilelib.compactGenomes != nil {
 183                                 tilelib.mtx.Lock()
 184                                 defer tilelib.mtx.Unlock()
 185                                 tilelib.compactGenomes[cg.Name] = cg.Variants
 186                         }
 187                 }()
 188         }
 189         wg.Wait()
 190         go close(errs)
 191         return <-errs
 192 }
 193
 194 func (tilelib *tileLibrary) loadCompactSequences(cseqs []CompactSequence, variantmap map[tileLibRef]tileVariantID) error {
 195         log.Infof("loadCompactSequences: %d todo", len(cseqs))
 196         for _, cseq := range cseqs {
 197                 log.Infof("loadCompactSequences: checking %s", cseq.Name)
 198                 for _, tseq := range cseq.TileSequences {
 199                         for i, libref := range tseq {
 200                                 if libref.Variant == 0 {
 201                                         // No variant (e.g., import
 202                                         // dropped tiles with
 203                                         // no-calls) = no translation.
 204                                         continue
 205                                 }
 206                                 v, ok := variantmap[libref]
 207                                 if !ok {
 208                                         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)
 209                                 }
 210                                 tseq[i].Variant = v
 211                         }
 212                 }
 213                 if tilelib.encoder != nil {
 214                         if err := tilelib.encoder.Encode(LibraryEntry{
 215                                 CompactSequences: []CompactSequence{cseq},
 216                         }); err != nil {
 217                                 return err
 218                         }
 219                 }
 220                 if tilelib.onAddRefseq != nil {
 221                         err := tilelib.onAddRefseq(cseq)
 222                         if err != nil {
 223                                 return err
 224                         }
 225                 }
 226                 log.Infof("loadCompactSequences: checking %s done", cseq.Name)
 227         }
 228         tilelib.mtx.Lock()
 229         defer tilelib.mtx.Unlock()
 230         if tilelib.refseqs == nil {
 231                 tilelib.refseqs = map[string]map[string][]tileLibRef{}
 232         }
 233         for _, cseq := range cseqs {
 234                 tilelib.refseqs[cseq.Name] = cseq.TileSequences
 235         }
 236         log.Info("loadCompactSequences: done")
 237         return nil
 238 }
 239
 240 func allFiles(path string, re *regexp.Regexp) ([]string, error) {
 241         var files []string
 242         f, err := open(path)
 243         if err != nil {
 244                 return nil, err
 245         }
 246         defer f.Close()
 247         fis, err := f.Readdir(-1)
 248         if err != nil {
 249                 return []string{path}, nil
 250         }
 251         for _, fi := range fis {
 252                 if fi.Name() == "." || fi.Name() == ".." {
 253                         continue
 254                 } else if child := path + "/" + fi.Name(); fi.IsDir() {
 255                         add, err := allFiles(child, re)
 256                         if err != nil {
 257                                 return nil, err
 258                         }
 259                         files = append(files, add...)
 260                 } else if re == nil || re.MatchString(child) {
 261                         files = append(files, child)
 262                 }
 263         }
 264         sort.Strings(files)
 265         return files, nil
 266 }
 267
 268 var matchGobFile = regexp.MustCompile(`\.gob(\.gz)?$`)
 269
 270 func (tilelib *tileLibrary) LoadDir(ctx context.Context, path string) error {
 271         log.Infof("LoadDir: walk dir %s", path)
 272         files, err := allFiles(path, matchGobFile)
 273         if err != nil {
 274                 return err
 275         }
 276         ctx, cancel := context.WithCancel(ctx)
 277         defer cancel()
 278         var mtx sync.Mutex
 279         allcgs := make([][]CompactGenome, len(files))
 280         allcseqs := make([][]CompactSequence, len(files))
 281         allvariantmap := map[tileLibRef]tileVariantID{}
 282         errs := make(chan error, len(files))
 283         log.Infof("LoadDir: read %d files", len(files))
 284         for fileno, path := range files {
 285                 fileno, path := fileno, path
 286                 go func() {
 287                         f, err := open(path)
 288                         if err != nil {
 289                                 errs <- err
 290                                 return
 291                         }
 292                         defer f.Close()
 293                         defer log.Infof("LoadDir: finished reading %s", path)
 294
 295                         var variantmap = map[tileLibRef]tileVariantID{}
 296                         var cgs []CompactGenome
 297                         var cseqs []CompactSequence
 298                         err = DecodeLibrary(f, strings.HasSuffix(path, ".gz"), func(ent *LibraryEntry) error {
 299                                 if ctx.Err() != nil {
 300                                         return ctx.Err()
 301                                 }
 302                                 if len(ent.TagSet) > 0 {
 303                                         mtx.Lock()
 304                                         if tilelib.taglib == nil || tilelib.taglib.Len() != len(ent.TagSet) {
 305                                                 // load first set of tags, or
 306                                                 // report mismatch if 2 sets
 307                                                 // have different #tags.
 308                                                 if err := tilelib.loadTagSet(ent.TagSet); err != nil {
 309                                                         mtx.Unlock()
 310                                                         return err
 311                                                 }
 312                                         }
 313                                         mtx.Unlock()
 314                                 }
 315                                 for _, tv := range ent.TileVariants {
 316                                         variantmap[tileLibRef{Tag: tv.Tag, Variant: tv.Variant}] = tilelib.getRef(tv.Tag, tv.Sequence, tv.Ref).Variant
 317                                 }
 318                                 cgs = append(cgs, ent.CompactGenomes...)
 319                                 cseqs = append(cseqs, ent.CompactSequences...)
 320                                 return nil
 321                         })
 322                         allcgs[fileno] = cgs
 323                         allcseqs[fileno] = cseqs
 324                         mtx.Lock()
 325                         defer mtx.Unlock()
 326                         for k, v := range variantmap {
 327                                 allvariantmap[k] = v
 328                         }
 329                         errs <- err
 330                 }()
 331         }
 332         for range files {
 333                 err := <-errs
 334                 if err != nil {
 335                         return err
 336                 }
 337         }
 338
 339         log.Info("LoadDir: loadCompactGenomes")
 340         var flatcgs []CompactGenome
 341         for _, cgs := range allcgs {
 342                 flatcgs = append(flatcgs, cgs...)
 343         }
 344         err = tilelib.loadCompactGenomes(flatcgs, allvariantmap)
 345         if err != nil {
 346                 return err
 347         }
 348
 349         log.Info("LoadDir: loadCompactSequences")
 350         var flatcseqs []CompactSequence
 351         for _, cseqs := range allcseqs {
 352                 flatcseqs = append(flatcseqs, cseqs...)
 353         }
 354         err = tilelib.loadCompactSequences(flatcseqs, allvariantmap)
 355         if err != nil {
 356                 return err
 357         }
 358
 359         log.Info("LoadDir done")
 360         return nil
 361 }
 362
 363 func (tilelib *tileLibrary) WriteDir(dir string) error {
 364         ntilefiles := 128
 365         nfiles := ntilefiles + len(tilelib.refseqs)
 366         files := make([]*os.File, nfiles)
 367         for i := range files {
 368                 f, err := os.OpenFile(fmt.Sprintf("%s/library.%04d.gob.gz", dir, i), os.O_CREATE|os.O_WRONLY, 0666)
 369                 if err != nil {
 370                         return err
 371                 }
 372                 defer f.Close()
 373                 files[i] = f
 374         }
 375         bufws := make([]*bufio.Writer, nfiles)
 376         for i := range bufws {
 377                 bufws[i] = bufio.NewWriterSize(files[i], 1<<26)
 378         }
 379         zws := make([]*pgzip.Writer, nfiles)
 380         for i := range zws {
 381                 zws[i] = pgzip.NewWriter(bufws[i])
 382                 defer zws[i].Close()
 383         }
 384         encoders := make([]*gob.Encoder, nfiles)
 385         for i := range encoders {
 386                 encoders[i] = gob.NewEncoder(zws[i])
 387         }
 388
 389         cgnames := make([]string, 0, len(tilelib.compactGenomes))
 390         for name := range tilelib.compactGenomes {
 391                 cgnames = append(cgnames, name)
 392         }
 393         sort.Strings(cgnames)
 394
 395         refnames := make([]string, 0, len(tilelib.refseqs))
 396         for name := range tilelib.refseqs {
 397                 refnames = append(refnames, name)
 398         }
 399         sort.Strings(refnames)
 400
 401         log.Infof("WriteDir: writing %d files", nfiles)
 402         ctx, cancel := context.WithCancel(context.Background())
 403         defer cancel()
 404         errs := make(chan error, nfiles)
 405         for start := range files {
 406                 start := start
 407                 go func() {
 408                         err := encoders[start].Encode(LibraryEntry{TagSet: tilelib.taglib.Tags()})
 409                         if err != nil {
 410                                 errs <- err
 411                                 return
 412                         }
 413                         if refidx := start - ntilefiles; refidx >= 0 {
 414                                 // write each ref to its own file
 415                                 // (they seem to load very slowly)
 416                                 name := refnames[refidx]
 417                                 errs <- encoders[start].Encode(LibraryEntry{CompactSequences: []CompactSequence{{
 418                                         Name:          name,
 419                                         TileSequences: tilelib.refseqs[name],
 420                                 }}})
 421                                 return
 422                         }
 423                         for i := start; i < len(cgnames); i += ntilefiles {
 424                                 err := encoders[start].Encode(LibraryEntry{CompactGenomes: []CompactGenome{{
 425                                         Name:     cgnames[i],
 426                                         Variants: tilelib.compactGenomes[cgnames[i]],
 427                                 }}})
 428                                 if err != nil {
 429                                         errs <- err
 430                                         return
 431                                 }
 432                         }
 433                         tvs := []TileVariant{}
 434                         for tag := start; tag < len(tilelib.variant) && ctx.Err() == nil; tag += ntilefiles {
 435                                 tvs = tvs[:0]
 436                                 for idx, hash := range tilelib.variant[tag] {
 437                                         tvs = append(tvs, TileVariant{
 438                                                 Tag:      tagID(tag),
 439                                                 Variant:  tileVariantID(idx + 1),
 440                                                 Blake2b:  hash,
 441                                                 Sequence: tilelib.hashSequence(hash),
 442                                         })
 443                                 }
 444                                 err := encoders[start].Encode(LibraryEntry{TileVariants: tvs})
 445                                 if err != nil {
 446                                         errs <- err
 447                                         return
 448                                 }
 449                         }
 450                         errs <- nil
 451                 }()
 452         }
 453         for range files {
 454                 err := <-errs
 455                 if err != nil {
 456                         return err
 457                 }
 458         }
 459         log.Info("WriteDir: flushing")
 460         for i := range zws {
 461                 err := zws[i].Close()
 462                 if err != nil {
 463                         return err
 464                 }
 465                 err = bufws[i].Flush()
 466                 if err != nil {
 467                         return err
 468                 }
 469                 err = files[i].Close()
 470                 if err != nil {
 471                         return err
 472                 }
 473         }
 474         log.Info("WriteDir: done")
 475         return nil
 476 }
 477
 478 // Load library data from rdr. Tile variants might be renumbered in
 479 // the process; in that case, genomes variants will be renumbered to
 480 // match.
 481 func (tilelib *tileLibrary) LoadGob(ctx context.Context, rdr io.Reader, gz bool) error {
 482         cgs := []CompactGenome{}
 483         cseqs := []CompactSequence{}
 484         variantmap := map[tileLibRef]tileVariantID{}
 485         err := DecodeLibrary(rdr, gz, func(ent *LibraryEntry) error {
 486                 if ctx.Err() != nil {
 487                         return ctx.Err()
 488                 }
 489                 if err := tilelib.loadTagSet(ent.TagSet); err != nil {
 490                         return err
 491                 }
 492                 if err := tilelib.loadTileVariants(ent.TileVariants, variantmap); err != nil {
 493                         return err
 494                 }
 495                 cgs = append(cgs, ent.CompactGenomes...)
 496                 cseqs = append(cseqs, ent.CompactSequences...)
 497                 return nil
 498         })
 499         if err != nil {
 500                 return err
 501         }
 502         if ctx.Err() != nil {
 503                 return ctx.Err()
 504         }
 505         err = tilelib.loadCompactGenomes(cgs, variantmap)
 506         if err != nil {
 507                 return err
 508         }
 509         err = tilelib.loadCompactSequences(cseqs, variantmap)
 510         if err != nil {
 511                 return err
 512         }
 513         return nil
 514 }
 515
 516 func (tilelib *tileLibrary) dump(out io.Writer) {
 517         printTV := func(tag int, variant tileVariantID) {
 518                 if variant < 1 {
 519                         fmt.Fprintf(out, " -")
 520                 } else if tag >= len(tilelib.variant) {
 521                         fmt.Fprintf(out, " (!tag=%d)", tag)
 522                 } else if int(variant) > len(tilelib.variant[tag]) {
 523                         fmt.Fprintf(out, " (tag=%d,!variant=%d)", tag, variant)
 524                 } else {
 525                         fmt.Fprintf(out, " %x", tilelib.variant[tag][variant-1][:8])
 526                 }
 527         }
 528         for refname, refseqs := range tilelib.refseqs {
 529                 for seqname, seq := range refseqs {
 530                         fmt.Fprintf(out, "ref %s %s", refname, seqname)
 531                         for _, libref := range seq {
 532                                 printTV(int(libref.Tag), libref.Variant)
 533                         }
 534                         fmt.Fprintf(out, "\n")
 535                 }
 536         }
 537         for name, cg := range tilelib.compactGenomes {
 538                 fmt.Fprintf(out, "cg %s", name)
 539                 for tag, variant := range cg {
 540                         printTV(tag/2, variant)
 541                 }
 542                 fmt.Fprintf(out, "\n")
 543         }
 544 }
 545
 546 type importStats struct {
 547         InputFile             string
 548         InputLabel            string
 549         InputLength           int
 550         InputCoverage         int
 551         PathLength            int
 552         DroppedRepeatedTags   int
 553         DroppedOutOfOrderTags int
 554 }
 555
 556 func (tilelib *tileLibrary) TileFasta(filelabel string, rdr io.Reader, matchChromosome *regexp.Regexp, isRef bool) (tileSeq, []importStats, error) {
 557         ret := tileSeq{}
 558         type jobT struct {
 559                 label string
 560                 fasta []byte
 561         }
 562         todo := make(chan jobT, 1)
 563         scanner := bufio.NewScanner(rdr)
 564         scanner.Buffer(make([]byte, 256), 1<<29) // 512 MiB, in case fasta does not have line breaks
 565         go func() {
 566                 defer close(todo)
 567                 var fasta []byte
 568                 var seqlabel string
 569                 for scanner.Scan() {
 570                         buf := scanner.Bytes()
 571                         if len(buf) > 0 && buf[0] == '>' {
 572                                 todo <- jobT{seqlabel, append([]byte(nil), fasta...)}
 573                                 seqlabel, fasta = strings.SplitN(string(buf[1:]), " ", 2)[0], fasta[:0]
 574                                 log.Debugf("%s %s reading fasta", filelabel, seqlabel)
 575                         } else if len(buf) > 0 && buf[0] == '#' {
 576                                 // ignore testdata comment
 577                         } else {
 578                                 fasta = append(fasta, bytes.ToLower(buf)...)
 579                         }
 580                 }
 581                 todo <- jobT{seqlabel, fasta}
 582         }()
 583         type foundtag struct {
 584                 pos   int
 585                 tagid tagID
 586         }
 587         found := make([]foundtag, 2000000)
 588         path := make([]tileLibRef, 2000000)
 589         totalFoundTags := 0
 590         totalPathLen := 0
 591         skippedSequences := 0
 592         taglen := tilelib.taglib.TagLen()
 593         var stats []importStats
 594         for job := range todo {
 595                 if len(job.fasta) == 0 {
 596                         continue
 597                 } else if !matchChromosome.MatchString(job.label) {
 598                         skippedSequences++
 599                         continue
 600                 }
 601                 log.Debugf("%s %s tiling", filelabel, job.label)
 602
 603                 found = found[:0]
 604                 tilelib.taglib.FindAll(job.fasta, func(tagid tagID, pos, taglen int) {
 605                         found = append(found, foundtag{pos: pos, tagid: tagid})
 606                 })
 607                 totalFoundTags += len(found)
 608                 if len(found) == 0 {
 609                         log.Warnf("%s %s no tags found", filelabel, job.label)
 610                 }
 611
 612                 droppedDup := 0
 613                 if !tilelib.useDups {
 614                         // Remove any tags that appeared more than once
 615                         dup := map[tagID]bool{}
 616                         for _, ft := range found {
 617                                 _, dup[ft.tagid] = dup[ft.tagid]
 618                         }
 619                         dst := 0
 620                         for _, ft := range found {
 621                                 if !dup[ft.tagid] {
 622                                         found[dst] = ft
 623                                         dst++
 624                                 }
 625                         }
 626                         droppedDup = len(found) - dst
 627                         log.Infof("%s %s dropping %d non-unique tags", filelabel, job.label, droppedDup)
 628                         found = found[:dst]
 629                 }
 630
 631                 droppedOOO := 0
 632                 if tilelib.skipOOO {
 633                         keep := longestIncreasingSubsequence(len(found), func(i int) int { return int(found[i].tagid) })
 634                         for i, x := range keep {
 635                                 found[i] = found[x]
 636                         }
 637                         droppedOOO = len(found) - len(keep)
 638                         log.Infof("%s %s dropping %d out-of-order tags", filelabel, job.label, droppedOOO)
 639                         found = found[:len(keep)]
 640                 }
 641
 642                 log.Infof("%s %s getting %d librefs", filelabel, job.label, len(found))
 643                 throttle := &throttle{Max: runtime.NumCPU()}
 644                 path = path[:len(found)]
 645                 var lowquality int64
 646                 for i, f := range found {
 647                         i, f := i, f
 648                         throttle.Acquire()
 649                         go func() {
 650                                 defer throttle.Release()
 651                                 var startpos, endpos int
 652                                 if i == 0 {
 653                                         startpos = 0
 654                                 } else {
 655                                         startpos = f.pos
 656                                 }
 657                                 if i == len(found)-1 {
 658                                         endpos = len(job.fasta)
 659                                 } else {
 660                                         endpos = found[i+1].pos + taglen
 661                                 }
 662                                 path[i] = tilelib.getRef(f.tagid, job.fasta[startpos:endpos], isRef)
 663                                 if countBases(job.fasta[startpos:endpos]) != endpos-startpos {
 664                                         atomic.AddInt64(&lowquality, 1)
 665                                 }
 666                         }()
 667                 }
 668                 throttle.Wait()
 669
 670                 log.Infof("%s %s copying path", filelabel, job.label)
 671
 672                 pathcopy := make([]tileLibRef, len(path))
 673                 copy(pathcopy, path)
 674                 ret[job.label] = pathcopy
 675
 676                 basesIn := countBases(job.fasta)
 677                 log.Infof("%s %s fasta in %d coverage in %d path len %d low-quality %d", filelabel, job.label, len(job.fasta), basesIn, len(path), lowquality)
 678                 stats = append(stats, importStats{
 679                         InputFile:             filelabel,
 680                         InputLabel:            job.label,
 681                         InputLength:           len(job.fasta),
 682                         InputCoverage:         basesIn,
 683                         PathLength:            len(path),
 684                         DroppedOutOfOrderTags: droppedOOO,
 685                         DroppedRepeatedTags:   droppedDup,
 686                 })
 687
 688                 totalPathLen += len(path)
 689         }
 690         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)
 691         return ret, stats, scanner.Err()
 692 }
 693
 694 func (tilelib *tileLibrary) Len() int64 {
 695         return atomic.LoadInt64(&tilelib.variants)
 696 }
 697
 698 // Return a tileLibRef for a tile with the given tag and sequence,
 699 // adding the sequence to the library if needed.
 700 func (tilelib *tileLibrary) getRef(tag tagID, seq []byte, usedByRef bool) tileLibRef {
 701         dropSeq := false
 702         if !tilelib.retainNoCalls {
 703                 for _, b := range seq {
 704                         if b != 'a' && b != 'c' && b != 'g' && b != 't' {
 705                                 dropSeq = true
 706                                 break
 707                         }
 708                 }
 709         }
 710         seqhash := blake2b.Sum256(seq)
 711         var vlock sync.Locker
 712
 713         tilelib.mtx.RLock()
 714         if len(tilelib.vlock) > int(tag) {
 715                 vlock = tilelib.vlock[tag]
 716         }
 717         tilelib.mtx.RUnlock()
 718
 719         if vlock != nil {
 720                 vlock.Lock()
 721                 for i, varhash := range tilelib.variant[tag] {
 722                         if varhash == seqhash {
 723                                 vlock.Unlock()
 724                                 return tileLibRef{Tag: tag, Variant: tileVariantID(i + 1)}
 725                         }
 726                 }
 727                 vlock.Unlock()
 728         } else {
 729                 tilelib.mtx.Lock()
 730                 if tilelib.variant == nil && tilelib.taglib != nil {
 731                         tilelib.variant = make([][][blake2b.Size256]byte, tilelib.taglib.Len())
 732                         tilelib.vlock = make([]sync.Locker, tilelib.taglib.Len())
 733                         for i := range tilelib.vlock {
 734                                 tilelib.vlock[i] = new(sync.Mutex)
 735                         }
 736                 }
 737                 if int(tag) >= len(tilelib.variant) {
 738                         oldlen := len(tilelib.vlock)
 739                         for i := 0; i < oldlen; i++ {
 740                                 tilelib.vlock[i].Lock()
 741                         }
 742                         // If we haven't seen the tag library yet (as
 743                         // in a merge), tilelib.taglib.Len() is
 744                         // zero. We can still behave correctly, we
 745                         // just need to expand the tilelib.variant and
 746                         // tilelib.vlock slices as needed.
 747                         if int(tag) >= cap(tilelib.variant) {
 748                                 // Allocate 2x capacity.
 749                                 newslice := make([][][blake2b.Size256]byte, int(tag)+1, (int(tag)+1)*2)
 750                                 copy(newslice, tilelib.variant)
 751                                 tilelib.variant = newslice[:int(tag)+1]
 752                                 newvlock := make([]sync.Locker, int(tag)+1, (int(tag)+1)*2)
 753                                 copy(newvlock, tilelib.vlock)
 754                                 tilelib.vlock = newvlock[:int(tag)+1]
 755                         } else {
 756                                 // Use previously allocated capacity,
 757                                 // avoiding copy.
 758                                 tilelib.variant = tilelib.variant[:int(tag)+1]
 759                                 tilelib.vlock = tilelib.vlock[:int(tag)+1]
 760                         }
 761                         for i := oldlen; i < len(tilelib.vlock); i++ {
 762                                 tilelib.vlock[i] = new(sync.Mutex)
 763                         }
 764                         for i := 0; i < oldlen; i++ {
 765                                 tilelib.vlock[i].Unlock()
 766                         }
 767                 }
 768                 vlock = tilelib.vlock[tag]
 769                 tilelib.mtx.Unlock()
 770         }
 771
 772         vlock.Lock()
 773         for i, varhash := range tilelib.variant[tag] {
 774                 if varhash == seqhash {
 775                         vlock.Unlock()
 776                         return tileLibRef{Tag: tag, Variant: tileVariantID(i + 1)}
 777                 }
 778         }
 779         atomic.AddInt64(&tilelib.variants, 1)
 780         tilelib.variant[tag] = append(tilelib.variant[tag], seqhash)
 781         variant := tileVariantID(len(tilelib.variant[tag]))
 782         vlock.Unlock()
 783
 784         if tilelib.retainTileSequences && !dropSeq {
 785                 seqCopy := append([]byte(nil), seq...)
 786                 if tilelib.seq2 == nil {
 787                         tilelib.mtx.Lock()
 788                         if tilelib.seq2 == nil {
 789                                 tilelib.seq2lock = map[[2]byte]sync.Locker{}
 790                                 m := map[[2]byte]map[[blake2b.Size256]byte][]byte{}
 791                                 var k [2]byte
 792                                 for i := 0; i < 256; i++ {
 793                                         k[0] = byte(i)
 794                                         for j := 0; j < 256; j++ {
 795                                                 k[1] = byte(j)
 796                                                 m[k] = map[[blake2b.Size256]byte][]byte{}
 797                                                 tilelib.seq2lock[k] = &sync.Mutex{}
 798                                         }
 799                                 }
 800                                 tilelib.seq2 = m
 801                         }
 802                         tilelib.mtx.Unlock()
 803                 }
 804                 var k [2]byte
 805                 copy(k[:], seqhash[:])
 806                 locker := tilelib.seq2lock[k]
 807                 locker.Lock()
 808                 tilelib.seq2[k][seqhash] = seqCopy
 809                 locker.Unlock()
 810         }
 811
 812         saveSeq := seq
 813         if dropSeq {
 814                 // Save the hash, but not the sequence
 815                 saveSeq = nil
 816         }
 817         if tilelib.encoder != nil {
 818                 tilelib.encoder.Encode(LibraryEntry{
 819                         TileVariants: []TileVariant{{
 820                                 Tag:      tag,
 821                                 Ref:      usedByRef,
 822                                 Variant:  variant,
 823                                 Blake2b:  seqhash,
 824                                 Sequence: saveSeq,
 825                         }},
 826                 })
 827         }
 828         if tilelib.onAddTileVariant != nil {
 829                 tilelib.onAddTileVariant(tileLibRef{tag, variant}, seqhash, saveSeq)
 830         }
 831         return tileLibRef{Tag: tag, Variant: variant}
 832 }
 833
 834 func (tilelib *tileLibrary) hashSequence(hash [blake2b.Size256]byte) []byte {
 835         var partition [2]byte
 836         copy(partition[:], hash[:])
 837         return tilelib.seq2[partition][hash]
 838 }
 839
 840 func (tilelib *tileLibrary) TileVariantSequence(libref tileLibRef) []byte {
 841         if libref.Variant == 0 || len(tilelib.variant) <= int(libref.Tag) || len(tilelib.variant[libref.Tag]) < int(libref.Variant) {
 842                 return nil
 843         }
 844         return tilelib.hashSequence(tilelib.variant[libref.Tag][libref.Variant-1])
 845 }
 846
 847 // Tidy deletes unreferenced tile variants and renumbers variants so
 848 // more common variants have smaller IDs.
 849 func (tilelib *tileLibrary) Tidy() {
 850         log.Print("Tidy: compute inref")
 851         inref := map[tileLibRef]bool{}
 852         for _, refseq := range tilelib.refseqs {
 853                 for _, librefs := range refseq {
 854                         for _, libref := range librefs {
 855                                 inref[libref] = true
 856                         }
 857                 }
 858         }
 859         log.Print("Tidy: compute remap")
 860         remap := make([][]tileVariantID, len(tilelib.variant))
 861         throttle := throttle{Max: runtime.NumCPU() + 1}
 862         for tag, oldvariants := range tilelib.variant {
 863                 tag, oldvariants := tagID(tag), oldvariants
 864                 if tag%1000000 == 0 {
 865                         log.Printf("Tidy: tag %d", tag)
 866                 }
 867                 throttle.Acquire()
 868                 go func() {
 869                         defer throttle.Release()
 870                         uses := make([]int, len(oldvariants))
 871                         for _, cg := range tilelib.compactGenomes {
 872                                 for phase := 0; phase < 2; phase++ {
 873                                         cgi := int(tag)*2 + phase
 874                                         if cgi < len(cg) && cg[cgi] > 0 {
 875                                                 uses[cg[cgi]-1]++
 876                                         }
 877                                 }
 878                         }
 879
 880                         // Compute desired order of variants:
 881                         // neworder[x] == index in oldvariants that
 882                         // should move to position x.
 883                         neworder := make([]int, len(oldvariants))
 884                         for i := range neworder {
 885                                 neworder[i] = i
 886                         }
 887                         sort.Slice(neworder, func(i, j int) bool {
 888                                 if cmp := uses[neworder[i]] - uses[neworder[j]]; cmp != 0 {
 889                                         return cmp > 0
 890                                 } else {
 891                                         return bytes.Compare(oldvariants[neworder[i]][:], oldvariants[neworder[j]][:]) < 0
 892                                 }
 893                         })
 894
 895                         // Replace tilelib.variant[tag] with a new
 896                         // re-ordered slice of hashes, and make a
 897                         // mapping from old to new variant IDs.
 898                         remaptag := make([]tileVariantID, len(oldvariants)+1)
 899                         newvariants := make([][blake2b.Size256]byte, 0, len(neworder))
 900                         for _, oldi := range neworder {
 901                                 if uses[oldi] > 0 || inref[tileLibRef{Tag: tag, Variant: tileVariantID(oldi + 1)}] {
 902                                         newvariants = append(newvariants, oldvariants[oldi])
 903                                         remaptag[oldi+1] = tileVariantID(len(newvariants))
 904                                 }
 905                         }
 906                         tilelib.variant[tag] = newvariants
 907                         remap[tag] = remaptag
 908                 }()
 909         }
 910         throttle.Wait()
 911
 912         // Apply remap to genomes and reference sequences, so they
 913         // refer to the same tile variants using the changed IDs.
 914         log.Print("Tidy: apply remap")
 915         var wg sync.WaitGroup
 916         for _, cg := range tilelib.compactGenomes {
 917                 cg := cg
 918                 wg.Add(1)
 919                 go func() {
 920                         defer wg.Done()
 921                         for idx, variant := range cg {
 922                                 cg[idx] = remap[tagID(idx/2)][variant]
 923                         }
 924                 }()
 925         }
 926         for _, refcs := range tilelib.refseqs {
 927                 for _, refseq := range refcs {
 928                         refseq := refseq
 929                         wg.Add(1)
 930                         go func() {
 931                                 defer wg.Done()
 932                                 for i, tv := range refseq {
 933                                         refseq[i].Variant = remap[tv.Tag][tv.Variant]
 934                                 }
 935                         }()
 936                 }
 937         }
 938         wg.Wait()
 939         log.Print("Tidy: done")
 940 }
 941
 942 func countBases(seq []byte) int {
 943         n := 0
 944         for _, c := range seq {
 945                 if isbase[c] {
 946                         n++
 947                 }
 948         }
 949         return n
 950 }