tilelib.go

   1 package main
   2
   3 import (
   4         "bufio"
   5         "bytes"
   6         "context"
   7         "encoding/gob"
   8         "fmt"
   9         "io"
  10         "regexp"
  11         "runtime"
  12         "sort"
  13         "strings"
  14         "sync"
  15         "sync/atomic"
  16
  17         log "github.com/sirupsen/logrus"
  18         "golang.org/x/crypto/blake2b"
  19 )
  20
  21 type tileVariantID uint16 // 1-based
  22
  23 type tileLibRef struct {
  24         Tag     tagID
  25         Variant tileVariantID
  26 }
  27
  28 type tileSeq map[string][]tileLibRef
  29
  30 func (tseq tileSeq) Variants() ([]tileVariantID, int, int) {
  31         maxtag := 0
  32         for _, refs := range tseq {
  33                 for _, ref := range refs {
  34                         if maxtag < int(ref.Tag) {
  35                                 maxtag = int(ref.Tag)
  36                         }
  37                 }
  38         }
  39         vars := make([]tileVariantID, maxtag+1)
  40         var kept, dropped int
  41         for _, refs := range tseq {
  42                 for _, ref := range refs {
  43                         if vars[int(ref.Tag)] != 0 {
  44                                 dropped++
  45                         } else {
  46                                 kept++
  47                         }
  48                         vars[int(ref.Tag)] = ref.Variant
  49                 }
  50         }
  51         return vars, kept, dropped
  52 }
  53
  54 type tileLibrary struct {
  55         retainNoCalls       bool
  56         skipOOO             bool
  57         retainTileSequences bool
  58
  59         taglib         *tagLibrary
  60         variant        [][][blake2b.Size256]byte
  61         refseqs        map[string]map[string][]tileLibRef
  62         compactGenomes map[string][]tileVariantID
  63         // count [][]int
  64         seq      map[[blake2b.Size256]byte][]byte
  65         variants int64
  66         // if non-nil, write out any tile variants added while tiling
  67         encoder *gob.Encoder
  68
  69         mtx   sync.RWMutex
  70         vlock []sync.Locker
  71 }
  72
  73 func (tilelib *tileLibrary) loadTagSet(newtagset [][]byte) error {
  74         // Loading a tagset means either passing it through to the
  75         // output (if it's the first one we've seen), or just ensuring
  76         // it doesn't disagree with what we already have.
  77         if len(newtagset) == 0 {
  78                 return nil
  79         }
  80         tilelib.mtx.Lock()
  81         defer tilelib.mtx.Unlock()
  82         if tilelib.taglib == nil || tilelib.taglib.Len() == 0 {
  83                 tilelib.taglib = &tagLibrary{}
  84                 err := tilelib.taglib.setTags(newtagset)
  85                 if err != nil {
  86                         return err
  87                 }
  88                 if tilelib.encoder != nil {
  89                         err = tilelib.encoder.Encode(LibraryEntry{
  90                                 TagSet: newtagset,
  91                         })
  92                         if err != nil {
  93                                 return err
  94                         }
  95                 }
  96         } else if tilelib.taglib.Len() != len(newtagset) {
  97                 return fmt.Errorf("cannot merge libraries with differing tagsets")
  98         } else {
  99                 current := tilelib.taglib.Tags()
 100                 for i := range newtagset {
 101                         if !bytes.Equal(newtagset[i], current[i]) {
 102                                 return fmt.Errorf("cannot merge libraries with differing tagsets")
 103                         }
 104                 }
 105         }
 106         return nil
 107 }
 108
 109 func (tilelib *tileLibrary) loadTileVariants(tvs []TileVariant, variantmap map[tileLibRef]tileVariantID) error {
 110         for _, tv := range tvs {
 111                 // Assign a new variant ID (unique across all inputs)
 112                 // for each input variant.
 113                 variantmap[tileLibRef{Tag: tv.Tag, Variant: tv.Variant}] = tilelib.getRef(tv.Tag, tv.Sequence).Variant
 114         }
 115         return nil
 116 }
 117
 118 func (tilelib *tileLibrary) loadCompactGenomes(cgs []CompactGenome, variantmap map[tileLibRef]tileVariantID, onLoadGenome func(CompactGenome)) error {
 119         log.Debugf("loadCompactGenomes: %d", len(cgs))
 120         var wg sync.WaitGroup
 121         errs := make(chan error, 1)
 122         for _, cg := range cgs {
 123                 wg.Add(1)
 124                 cg := cg
 125                 go func() {
 126                         defer wg.Done()
 127                         for i, variant := range cg.Variants {
 128                                 if len(errs) > 0 {
 129                                         return
 130                                 }
 131                                 if variant == 0 {
 132                                         continue
 133                                 }
 134                                 tag := tagID(i / 2)
 135                                 newvariant, ok := variantmap[tileLibRef{Tag: tag, Variant: variant}]
 136                                 if !ok {
 137                                         err := fmt.Errorf("oops: genome %q has variant %d for tag %d, but that variant was not in its library", cg.Name, variant, tag)
 138                                         select {
 139                                         case errs <- err:
 140                                         default:
 141                                         }
 142                                         return
 143                                 }
 144                                 log.Tracef("loadCompactGenomes: cg %s tag %d variant %d => %d", cg.Name, tag, variant, newvariant)
 145                                 cg.Variants[i] = newvariant
 146                         }
 147                         if onLoadGenome != nil {
 148                                 onLoadGenome(cg)
 149                         }
 150                         if tilelib.encoder != nil {
 151                                 err := tilelib.encoder.Encode(LibraryEntry{
 152                                         CompactGenomes: []CompactGenome{cg},
 153                                 })
 154                                 if err != nil {
 155                                         select {
 156                                         case errs <- err:
 157                                         default:
 158                                         }
 159                                         return
 160                                 }
 161                         }
 162                         if tilelib.compactGenomes != nil {
 163                                 tilelib.mtx.Lock()
 164                                 defer tilelib.mtx.Unlock()
 165                                 tilelib.compactGenomes[cg.Name] = cg.Variants
 166                         }
 167                 }()
 168         }
 169         wg.Wait()
 170         go close(errs)
 171         return <-errs
 172 }
 173
 174 func (tilelib *tileLibrary) loadCompactSequences(cseqs []CompactSequence, variantmap map[tileLibRef]tileVariantID) error {
 175         log.Debugf("loadCompactSequences: %d", len(cseqs))
 176         for _, cseq := range cseqs {
 177                 for _, tseq := range cseq.TileSequences {
 178                         for i, libref := range tseq {
 179                                 if libref.Variant == 0 {
 180                                         // No variant (e.g., import
 181                                         // dropped tiles with
 182                                         // no-calls) = no translation.
 183                                         continue
 184                                 }
 185                                 v, ok := variantmap[libref]
 186                                 if !ok {
 187                                         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)
 188                                 }
 189                                 tseq[i].Variant = v
 190                         }
 191                 }
 192                 if tilelib.encoder != nil {
 193                         if err := tilelib.encoder.Encode(LibraryEntry{
 194                                 CompactSequences: []CompactSequence{cseq},
 195                         }); err != nil {
 196                                 return err
 197                         }
 198                 }
 199         }
 200         tilelib.mtx.Lock()
 201         defer tilelib.mtx.Unlock()
 202         if tilelib.refseqs == nil {
 203                 tilelib.refseqs = map[string]map[string][]tileLibRef{}
 204         }
 205         for _, cseq := range cseqs {
 206                 tilelib.refseqs[cseq.Name] = cseq.TileSequences
 207         }
 208         return nil
 209 }
 210
 211 // Load library data from rdr. Tile variants might be renumbered in
 212 // the process; in that case, genomes variants will be renumbered to
 213 // match.
 214 //
 215 // If onLoadGenome is non-nil, call it on each CompactGenome entry.
 216 func (tilelib *tileLibrary) LoadGob(ctx context.Context, rdr io.Reader, gz bool, onLoadGenome func(CompactGenome)) error {
 217         cgs := []CompactGenome{}
 218         cseqs := []CompactSequence{}
 219         variantmap := map[tileLibRef]tileVariantID{}
 220         err := DecodeLibrary(rdr, gz, func(ent *LibraryEntry) error {
 221                 if ctx.Err() != nil {
 222                         return ctx.Err()
 223                 }
 224                 if err := tilelib.loadTagSet(ent.TagSet); err != nil {
 225                         return err
 226                 }
 227                 if err := tilelib.loadTileVariants(ent.TileVariants, variantmap); err != nil {
 228                         return err
 229                 }
 230                 cgs = append(cgs, ent.CompactGenomes...)
 231                 cseqs = append(cseqs, ent.CompactSequences...)
 232                 return nil
 233         })
 234         if err != nil {
 235                 return err
 236         }
 237         if ctx.Err() != nil {
 238                 return ctx.Err()
 239         }
 240         err = tilelib.loadCompactGenomes(cgs, variantmap, onLoadGenome)
 241         if err != nil {
 242                 return err
 243         }
 244         err = tilelib.loadCompactSequences(cseqs, variantmap)
 245         if err != nil {
 246                 return err
 247         }
 248         return nil
 249 }
 250
 251 type importStats struct {
 252         InputFile              string
 253         InputLabel             string
 254         InputLength            int
 255         InputCoverage          int
 256         TileCoverage           int
 257         PathLength             int
 258         DroppedOutOfOrderTiles int
 259 }
 260
 261 func (tilelib *tileLibrary) TileFasta(filelabel string, rdr io.Reader, matchChromosome *regexp.Regexp) (tileSeq, []importStats, error) {
 262         ret := tileSeq{}
 263         type jobT struct {
 264                 label string
 265                 fasta []byte
 266         }
 267         todo := make(chan jobT, 1)
 268         scanner := bufio.NewScanner(rdr)
 269         go func() {
 270                 defer close(todo)
 271                 var fasta []byte
 272                 var seqlabel string
 273                 for scanner.Scan() {
 274                         buf := scanner.Bytes()
 275                         if len(buf) > 0 && buf[0] == '>' {
 276                                 todo <- jobT{seqlabel, append([]byte(nil), fasta...)}
 277                                 seqlabel, fasta = strings.SplitN(string(buf[1:]), " ", 2)[0], fasta[:0]
 278                                 log.Debugf("%s %s reading fasta", filelabel, seqlabel)
 279                         } else {
 280                                 fasta = append(fasta, bytes.ToLower(buf)...)
 281                         }
 282                 }
 283                 todo <- jobT{seqlabel, fasta}
 284         }()
 285         type foundtag struct {
 286                 pos    int
 287                 tagid  tagID
 288                 taglen int
 289         }
 290         found := make([]foundtag, 2000000)
 291         path := make([]tileLibRef, 2000000)
 292         totalFoundTags := 0
 293         totalPathLen := 0
 294         skippedSequences := 0
 295         var stats []importStats
 296         for job := range todo {
 297                 if len(job.fasta) == 0 {
 298                         continue
 299                 } else if !matchChromosome.MatchString(job.label) {
 300                         skippedSequences++
 301                         continue
 302                 }
 303                 log.Debugf("%s %s tiling", filelabel, job.label)
 304
 305                 found = found[:0]
 306                 tilelib.taglib.FindAll(job.fasta, func(tagid tagID, pos, taglen int) {
 307                         found = append(found, foundtag{pos: pos, tagid: tagid, taglen: taglen})
 308                 })
 309                 totalFoundTags += len(found)
 310
 311                 basesOut := 0
 312                 skipped := 0
 313                 path = path[:0]
 314                 last := foundtag{tagid: -1}
 315                 if tilelib.skipOOO {
 316                         keep := longestIncreasingSubsequence(len(found), func(i int) int { return int(found[i].tagid) })
 317                         for i, x := range keep {
 318                                 found[i] = found[x]
 319                         }
 320                         skipped = len(found) - len(keep)
 321                         found = found[:len(keep)]
 322                 }
 323                 for i, f := range found {
 324                         log.Tracef("%s %s found[%d] == %#v", filelabel, job.label, i, f)
 325                         if last.tagid < 0 {
 326                                 // first tag in sequence
 327                                 last = foundtag{tagid: f.tagid}
 328                                 continue
 329                         }
 330                         libref := tilelib.getRef(last.tagid, job.fasta[last.pos:f.pos+f.taglen])
 331                         path = append(path, libref)
 332                         if libref.Variant > 0 {
 333                                 // Count output coverage from
 334                                 // the end of the previous tag
 335                                 // (if any) to the end of the
 336                                 // current tag, IOW don't
 337                                 // double-count coverage for
 338                                 // the previous tag.
 339                                 basesOut += countBases(job.fasta[last.pos+last.taglen : f.pos+f.taglen])
 340                         } else {
 341                                 // If we dropped this tile
 342                                 // (because !retainNoCalls),
 343                                 // set taglen=0 so the
 344                                 // overlapping tag is counted
 345                                 // toward coverage on the
 346                                 // following tile.
 347                                 f.taglen = 0
 348                         }
 349                         last = f
 350                 }
 351                 if last.tagid < 0 {
 352                         log.Warnf("%s %s no tags found", filelabel, job.label)
 353                 } else {
 354                         libref := tilelib.getRef(last.tagid, job.fasta[last.pos:])
 355                         path = append(path, libref)
 356                         if libref.Variant > 0 {
 357                                 basesOut += countBases(job.fasta[last.pos+last.taglen:])
 358                         }
 359                 }
 360
 361                 pathcopy := make([]tileLibRef, len(path))
 362                 copy(pathcopy, path)
 363                 ret[job.label] = pathcopy
 364
 365                 basesIn := countBases(job.fasta)
 366                 log.Infof("%s %s fasta in %d coverage in %d coverage out %d path len %d skipped %d", filelabel, job.label, len(job.fasta), basesIn, basesOut, len(path), skipped)
 367                 stats = append(stats, importStats{
 368                         InputFile:              filelabel,
 369                         InputLabel:             job.label,
 370                         InputLength:            len(job.fasta),
 371                         InputCoverage:          basesIn,
 372                         TileCoverage:           basesOut,
 373                         PathLength:             len(path),
 374                         DroppedOutOfOrderTiles: skipped,
 375                 })
 376
 377                 totalPathLen += len(path)
 378         }
 379         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)
 380         return ret, stats, scanner.Err()
 381 }
 382
 383 func (tilelib *tileLibrary) Len() int64 {
 384         return atomic.LoadInt64(&tilelib.variants)
 385 }
 386
 387 // Return a tileLibRef for a tile with the given tag and sequence,
 388 // adding the sequence to the library if needed.
 389 func (tilelib *tileLibrary) getRef(tag tagID, seq []byte) tileLibRef {
 390         dropSeq := false
 391         if !tilelib.retainNoCalls {
 392                 for _, b := range seq {
 393                         if b != 'a' && b != 'c' && b != 'g' && b != 't' {
 394                                 dropSeq = true
 395                                 break
 396                         }
 397                 }
 398         }
 399         seqhash := blake2b.Sum256(seq)
 400         tilelib.mtx.RLock()
 401         if int(tag) < len(tilelib.variant) {
 402                 for i, varhash := range tilelib.variant[tag] {
 403                         if varhash == seqhash {
 404                                 tilelib.mtx.RUnlock()
 405                                 return tileLibRef{Tag: tag, Variant: tileVariantID(i + 1)}
 406                         }
 407                 }
 408         }
 409         var vlock sync.Locker
 410         if len(tilelib.vlock) > int(tag) {
 411                 vlock = tilelib.vlock[tag]
 412         }
 413         tilelib.mtx.RUnlock()
 414         if vlock == nil {
 415                 tilelib.mtx.Lock()
 416                 if tilelib.variant == nil && tilelib.taglib != nil {
 417                         tilelib.variant = make([][][blake2b.Size256]byte, tilelib.taglib.Len())
 418                         tilelib.vlock = make([]sync.Locker, tilelib.taglib.Len())
 419                         for i := range tilelib.vlock {
 420                                 tilelib.vlock[i] = &sync.Mutex{}
 421                         }
 422                 }
 423                 if int(tag) >= len(tilelib.variant) {
 424                         oldlen := len(tilelib.variant)
 425                         // If we haven't seen the tag library yet (as
 426                         // in a merge), tilelib.taglib.Len() is
 427                         // zero. We can still behave correctly, we
 428                         // just need to expand the tilelib.variant
 429                         // slice as needed.
 430                         if int(tag) >= cap(tilelib.variant) {
 431                                 // Allocate 2x capacity.
 432                                 newslice := make([][][blake2b.Size256]byte, int(tag)+1, (int(tag)+1)*2)
 433                                 copy(newslice, tilelib.variant)
 434                                 tilelib.variant = newslice[:int(tag)+1]
 435                                 newvlock := make([]sync.Locker, int(tag)+1, (int(tag)+1)*2)
 436                                 copy(newvlock, tilelib.vlock)
 437                                 tilelib.vlock = newvlock[:int(tag)+1]
 438                         } else {
 439                                 // Use previously allocated capacity,
 440                                 // avoiding copy.
 441                                 tilelib.variant = tilelib.variant[:int(tag)+1]
 442                                 tilelib.vlock = tilelib.vlock[:int(tag)+1]
 443                         }
 444                         for i := oldlen; i < len(tilelib.variant); i++ {
 445                                 tilelib.vlock[i] = &sync.Mutex{}
 446                         }
 447                 }
 448                 vlock = tilelib.vlock[tag]
 449                 tilelib.mtx.Unlock()
 450         }
 451
 452         tilelib.mtx.RLock()
 453         vlock.Lock()
 454         for i, varhash := range tilelib.variant[tag] {
 455                 if varhash == seqhash {
 456                         vlock.Unlock()
 457                         tilelib.mtx.RUnlock()
 458                         return tileLibRef{Tag: tag, Variant: tileVariantID(i + 1)}
 459                 }
 460         }
 461         atomic.AddInt64(&tilelib.variants, 1)
 462         tilelib.variant[tag] = append(tilelib.variant[tag], seqhash)
 463         if tilelib.retainTileSequences && !dropSeq {
 464                 if tilelib.seq == nil {
 465                         tilelib.seq = map[[blake2b.Size256]byte][]byte{}
 466                 }
 467                 tilelib.seq[seqhash] = append([]byte(nil), seq...)
 468         }
 469         variant := tileVariantID(len(tilelib.variant[tag]))
 470         vlock.Unlock()
 471         tilelib.mtx.RUnlock()
 472
 473         if tilelib.encoder != nil {
 474                 saveSeq := seq
 475                 if dropSeq {
 476                         // Save the hash, but not the sequence
 477                         saveSeq = nil
 478                 }
 479                 tilelib.encoder.Encode(LibraryEntry{
 480                         TileVariants: []TileVariant{{
 481                                 Tag:      tag,
 482                                 Variant:  variant,
 483                                 Blake2b:  seqhash,
 484                                 Sequence: saveSeq,
 485                         }},
 486                 })
 487         }
 488         return tileLibRef{Tag: tag, Variant: variant}
 489 }
 490
 491 func (tilelib *tileLibrary) TileVariantSequence(libref tileLibRef) []byte {
 492         if libref.Variant == 0 || len(tilelib.variant) <= int(libref.Tag) || len(tilelib.variant[libref.Tag]) < int(libref.Variant) {
 493                 return nil
 494         }
 495         return tilelib.seq[tilelib.variant[libref.Tag][libref.Variant-1]]
 496 }
 497
 498 // Tidy deletes unreferenced tile variants and renumbers variants so
 499 // more common variants have smaller IDs.
 500 func (tilelib *tileLibrary) Tidy() {
 501         log.Print("Tidy: compute inref")
 502         inref := map[tileLibRef]bool{}
 503         for _, refseq := range tilelib.refseqs {
 504                 for _, librefs := range refseq {
 505                         for _, libref := range librefs {
 506                                 inref[libref] = true
 507                         }
 508                 }
 509         }
 510         log.Print("Tidy: compute remap")
 511         remap := make([][]tileVariantID, len(tilelib.variant))
 512         throttle := throttle{Max: runtime.NumCPU() + 1}
 513         for tag, oldvariants := range tilelib.variant {
 514                 tag, oldvariants := tagID(tag), oldvariants
 515                 if tag%1000000 == 0 {
 516                         log.Printf("Tidy: tag %d", tag)
 517                 }
 518                 throttle.Acquire()
 519                 go func() {
 520                         defer throttle.Release()
 521                         uses := make([]int, len(oldvariants))
 522                         for _, cg := range tilelib.compactGenomes {
 523                                 for phase := 0; phase < 2; phase++ {
 524                                         cgi := int(tag)*2 + phase
 525                                         if cgi < len(cg) && cg[cgi] > 0 {
 526                                                 uses[cg[cgi]-1]++
 527                                         }
 528                                 }
 529                         }
 530
 531                         // Compute desired order of variants:
 532                         // neworder[x] == index in oldvariants that
 533                         // should move to position x.
 534                         neworder := make([]int, len(oldvariants))
 535                         for i := range neworder {
 536                                 neworder[i] = i
 537                         }
 538                         sort.Slice(neworder, func(i, j int) bool {
 539                                 if cmp := uses[neworder[i]] - uses[neworder[j]]; cmp != 0 {
 540                                         return cmp > 0
 541                                 } else {
 542                                         return bytes.Compare(oldvariants[neworder[i]][:], oldvariants[neworder[j]][:]) < 0
 543                                 }
 544                         })
 545
 546                         // Replace tilelib.variant[tag] with a new
 547                         // re-ordered slice of hashes, and make a
 548                         // mapping from old to new variant IDs.
 549                         remaptag := make([]tileVariantID, len(oldvariants)+1)
 550                         newvariants := make([][blake2b.Size256]byte, 0, len(neworder))
 551                         for _, oldi := range neworder {
 552                                 if uses[oldi] > 0 || inref[tileLibRef{Tag: tag, Variant: tileVariantID(oldi + 1)}] {
 553                                         newvariants = append(newvariants, oldvariants[oldi])
 554                                         remaptag[oldi+1] = tileVariantID(len(newvariants))
 555                                 }
 556                         }
 557                         tilelib.variant[tag] = newvariants
 558                         remap[tag] = remaptag
 559                 }()
 560         }
 561         throttle.Wait()
 562
 563         // Apply remap to genomes and reference sequences, so they
 564         // refer to the same tile variants using the changed IDs.
 565         log.Print("Tidy: apply remap")
 566         for _, cg := range tilelib.compactGenomes {
 567                 for idx, variant := range cg {
 568                         cg[idx] = remap[tagID(idx/2)][variant]
 569                 }
 570         }
 571         for _, refcs := range tilelib.refseqs {
 572                 for _, refseq := range refcs {
 573                         for i, tv := range refseq {
 574                                 refseq[i].Variant = remap[tv.Tag][tv.Variant]
 575                         }
 576                 }
 577         }
 578         log.Print("Tidy: done")
 579 }
 580
 581 func countBases(seq []byte) int {
 582         n := 0
 583         for _, c := range seq {
 584                 if isbase[c] {
 585                         n++
 586                 }
 587         }
 588         return n
 589 }