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         PathLength             int
 257         DroppedOutOfOrderTiles int
 258 }
 259
 260 func (tilelib *tileLibrary) TileFasta(filelabel string, rdr io.Reader, matchChromosome *regexp.Regexp) (tileSeq, []importStats, error) {
 261         ret := tileSeq{}
 262         type jobT struct {
 263                 label string
 264                 fasta []byte
 265         }
 266         todo := make(chan jobT, 1)
 267         scanner := bufio.NewScanner(rdr)
 268         go func() {
 269                 defer close(todo)
 270                 var fasta []byte
 271                 var seqlabel string
 272                 for scanner.Scan() {
 273                         buf := scanner.Bytes()
 274                         if len(buf) > 0 && buf[0] == '>' {
 275                                 todo <- jobT{seqlabel, append([]byte(nil), fasta...)}
 276                                 seqlabel, fasta = strings.SplitN(string(buf[1:]), " ", 2)[0], fasta[:0]
 277                                 log.Debugf("%s %s reading fasta", filelabel, seqlabel)
 278                         } else {
 279                                 fasta = append(fasta, bytes.ToLower(buf)...)
 280                         }
 281                 }
 282                 todo <- jobT{seqlabel, fasta}
 283         }()
 284         type foundtag struct {
 285                 pos   int
 286                 tagid tagID
 287         }
 288         found := make([]foundtag, 2000000)
 289         path := make([]tileLibRef, 2000000)
 290         totalFoundTags := 0
 291         totalPathLen := 0
 292         skippedSequences := 0
 293         taglen := tilelib.taglib.TagLen()
 294         var stats []importStats
 295         for job := range todo {
 296                 if len(job.fasta) == 0 {
 297                         continue
 298                 } else if !matchChromosome.MatchString(job.label) {
 299                         skippedSequences++
 300                         continue
 301                 }
 302                 log.Debugf("%s %s tiling", filelabel, job.label)
 303
 304                 found = found[:0]
 305                 tilelib.taglib.FindAll(job.fasta, func(tagid tagID, pos, taglen int) {
 306                         found = append(found, foundtag{pos: pos, tagid: tagid})
 307                 })
 308                 totalFoundTags += len(found)
 309                 if len(found) == 0 {
 310                         log.Warnf("%s %s no tags found", filelabel, job.label)
 311                 }
 312
 313                 skipped := 0
 314                 if tilelib.skipOOO {
 315                         log.Infof("%s %s keeping longest increasing subsequence", filelabel, job.label)
 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
 324                 log.Infof("%s %s getting %d librefs", filelabel, job.label, len(found))
 325                 throttle := &throttle{Max: runtime.NumCPU()}
 326                 path = path[:len(found)]
 327                 for i, f := range found {
 328                         i, f := i, f
 329                         throttle.Acquire()
 330                         go func() {
 331                                 defer throttle.Release()
 332                                 var startpos, endpos int
 333                                 if i == 0 {
 334                                         startpos = 0
 335                                 } else {
 336                                         startpos = f.pos
 337                                 }
 338                                 if i == len(found)-1 {
 339                                         endpos = len(job.fasta)
 340                                 } else {
 341                                         endpos = found[i+1].pos + taglen
 342                                 }
 343                                 path[i] = tilelib.getRef(f.tagid, job.fasta[startpos:endpos])
 344                         }()
 345                 }
 346                 throttle.Wait()
 347
 348                 log.Infof("%s %s copying path", filelabel, job.label)
 349
 350                 pathcopy := make([]tileLibRef, len(path))
 351                 copy(pathcopy, path)
 352                 ret[job.label] = pathcopy
 353
 354                 basesIn := countBases(job.fasta)
 355                 log.Infof("%s %s fasta in %d coverage in %d path len %d skipped-out-of-order %d", filelabel, job.label, len(job.fasta), basesIn, len(path), skipped)
 356                 stats = append(stats, importStats{
 357                         InputFile:              filelabel,
 358                         InputLabel:             job.label,
 359                         InputLength:            len(job.fasta),
 360                         InputCoverage:          basesIn,
 361                         PathLength:             len(path),
 362                         DroppedOutOfOrderTiles: skipped,
 363                 })
 364
 365                 totalPathLen += len(path)
 366         }
 367         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)
 368         return ret, stats, scanner.Err()
 369 }
 370
 371 func (tilelib *tileLibrary) Len() int64 {
 372         return atomic.LoadInt64(&tilelib.variants)
 373 }
 374
 375 // Return a tileLibRef for a tile with the given tag and sequence,
 376 // adding the sequence to the library if needed.
 377 func (tilelib *tileLibrary) getRef(tag tagID, seq []byte) tileLibRef {
 378         dropSeq := false
 379         if !tilelib.retainNoCalls {
 380                 for _, b := range seq {
 381                         if b != 'a' && b != 'c' && b != 'g' && b != 't' {
 382                                 dropSeq = true
 383                                 break
 384                         }
 385                 }
 386         }
 387         seqhash := blake2b.Sum256(seq)
 388         var vlock sync.Locker
 389
 390         tilelib.mtx.RLock()
 391         if len(tilelib.vlock) > int(tag) {
 392                 vlock = tilelib.vlock[tag]
 393         }
 394         tilelib.mtx.RUnlock()
 395
 396         if vlock != nil {
 397                 vlock.Lock()
 398                 for i, varhash := range tilelib.variant[tag] {
 399                         if varhash == seqhash {
 400                                 vlock.Unlock()
 401                                 return tileLibRef{Tag: tag, Variant: tileVariantID(i + 1)}
 402                         }
 403                 }
 404                 vlock.Unlock()
 405         } else {
 406                 tilelib.mtx.Lock()
 407                 if tilelib.variant == nil && tilelib.taglib != nil {
 408                         tilelib.variant = make([][][blake2b.Size256]byte, tilelib.taglib.Len())
 409                         tilelib.vlock = make([]sync.Locker, tilelib.taglib.Len())
 410                         for i := range tilelib.vlock {
 411                                 tilelib.vlock[i] = new(sync.Mutex)
 412                         }
 413                 }
 414                 if int(tag) >= len(tilelib.variant) {
 415                         oldlen := len(tilelib.vlock)
 416                         for i := 0; i < oldlen; i++ {
 417                                 tilelib.vlock[i].Lock()
 418                         }
 419                         // If we haven't seen the tag library yet (as
 420                         // in a merge), tilelib.taglib.Len() is
 421                         // zero. We can still behave correctly, we
 422                         // just need to expand the tilelib.variant and
 423                         // tilelib.vlock slices as needed.
 424                         if int(tag) >= cap(tilelib.variant) {
 425                                 // Allocate 2x capacity.
 426                                 newslice := make([][][blake2b.Size256]byte, int(tag)+1, (int(tag)+1)*2)
 427                                 copy(newslice, tilelib.variant)
 428                                 tilelib.variant = newslice[:int(tag)+1]
 429                                 newvlock := make([]sync.Locker, int(tag)+1, (int(tag)+1)*2)
 430                                 copy(newvlock, tilelib.vlock)
 431                                 tilelib.vlock = newvlock[:int(tag)+1]
 432                         } else {
 433                                 // Use previously allocated capacity,
 434                                 // avoiding copy.
 435                                 tilelib.variant = tilelib.variant[:int(tag)+1]
 436                                 tilelib.vlock = tilelib.vlock[:int(tag)+1]
 437                         }
 438                         for i := oldlen; i < len(tilelib.vlock); i++ {
 439                                 tilelib.vlock[i] = new(sync.Mutex)
 440                         }
 441                         for i := 0; i < oldlen; i++ {
 442                                 tilelib.vlock[i].Unlock()
 443                         }
 444                 }
 445                 vlock = tilelib.vlock[tag]
 446                 tilelib.mtx.Unlock()
 447         }
 448
 449         vlock.Lock()
 450         for i, varhash := range tilelib.variant[tag] {
 451                 if varhash == seqhash {
 452                         vlock.Unlock()
 453                         return tileLibRef{Tag: tag, Variant: tileVariantID(i + 1)}
 454                 }
 455         }
 456         atomic.AddInt64(&tilelib.variants, 1)
 457         tilelib.variant[tag] = append(tilelib.variant[tag], seqhash)
 458         variant := tileVariantID(len(tilelib.variant[tag]))
 459         vlock.Unlock()
 460
 461         if tilelib.retainTileSequences && !dropSeq {
 462                 tilelib.mtx.Lock()
 463                 if tilelib.seq == nil {
 464                         tilelib.seq = map[[blake2b.Size256]byte][]byte{}
 465                 }
 466                 tilelib.seq[seqhash] = append([]byte(nil), seq...)
 467                 tilelib.mtx.Unlock()
 468         }
 469
 470         if tilelib.encoder != nil {
 471                 saveSeq := seq
 472                 if dropSeq {
 473                         // Save the hash, but not the sequence
 474                         saveSeq = nil
 475                 }
 476                 tilelib.encoder.Encode(LibraryEntry{
 477                         TileVariants: []TileVariant{{
 478                                 Tag:      tag,
 479                                 Variant:  variant,
 480                                 Blake2b:  seqhash,
 481                                 Sequence: saveSeq,
 482                         }},
 483                 })
 484         }
 485         return tileLibRef{Tag: tag, Variant: variant}
 486 }
 487
 488 func (tilelib *tileLibrary) TileVariantSequence(libref tileLibRef) []byte {
 489         if libref.Variant == 0 || len(tilelib.variant) <= int(libref.Tag) || len(tilelib.variant[libref.Tag]) < int(libref.Variant) {
 490                 return nil
 491         }
 492         return tilelib.seq[tilelib.variant[libref.Tag][libref.Variant-1]]
 493 }
 494
 495 // Tidy deletes unreferenced tile variants and renumbers variants so
 496 // more common variants have smaller IDs.
 497 func (tilelib *tileLibrary) Tidy() {
 498         log.Print("Tidy: compute inref")
 499         inref := map[tileLibRef]bool{}
 500         for _, refseq := range tilelib.refseqs {
 501                 for _, librefs := range refseq {
 502                         for _, libref := range librefs {
 503                                 inref[libref] = true
 504                         }
 505                 }
 506         }
 507         log.Print("Tidy: compute remap")
 508         remap := make([][]tileVariantID, len(tilelib.variant))
 509         throttle := throttle{Max: runtime.NumCPU() + 1}
 510         for tag, oldvariants := range tilelib.variant {
 511                 tag, oldvariants := tagID(tag), oldvariants
 512                 if tag%1000000 == 0 {
 513                         log.Printf("Tidy: tag %d", tag)
 514                 }
 515                 throttle.Acquire()
 516                 go func() {
 517                         defer throttle.Release()
 518                         uses := make([]int, len(oldvariants))
 519                         for _, cg := range tilelib.compactGenomes {
 520                                 for phase := 0; phase < 2; phase++ {
 521                                         cgi := int(tag)*2 + phase
 522                                         if cgi < len(cg) && cg[cgi] > 0 {
 523                                                 uses[cg[cgi]-1]++
 524                                         }
 525                                 }
 526                         }
 527
 528                         // Compute desired order of variants:
 529                         // neworder[x] == index in oldvariants that
 530                         // should move to position x.
 531                         neworder := make([]int, len(oldvariants))
 532                         for i := range neworder {
 533                                 neworder[i] = i
 534                         }
 535                         sort.Slice(neworder, func(i, j int) bool {
 536                                 if cmp := uses[neworder[i]] - uses[neworder[j]]; cmp != 0 {
 537                                         return cmp > 0
 538                                 } else {
 539                                         return bytes.Compare(oldvariants[neworder[i]][:], oldvariants[neworder[j]][:]) < 0
 540                                 }
 541                         })
 542
 543                         // Replace tilelib.variant[tag] with a new
 544                         // re-ordered slice of hashes, and make a
 545                         // mapping from old to new variant IDs.
 546                         remaptag := make([]tileVariantID, len(oldvariants)+1)
 547                         newvariants := make([][blake2b.Size256]byte, 0, len(neworder))
 548                         for _, oldi := range neworder {
 549                                 if uses[oldi] > 0 || inref[tileLibRef{Tag: tag, Variant: tileVariantID(oldi + 1)}] {
 550                                         newvariants = append(newvariants, oldvariants[oldi])
 551                                         remaptag[oldi+1] = tileVariantID(len(newvariants))
 552                                 }
 553                         }
 554                         tilelib.variant[tag] = newvariants
 555                         remap[tag] = remaptag
 556                 }()
 557         }
 558         throttle.Wait()
 559
 560         // Apply remap to genomes and reference sequences, so they
 561         // refer to the same tile variants using the changed IDs.
 562         log.Print("Tidy: apply remap")
 563         for _, cg := range tilelib.compactGenomes {
 564                 for idx, variant := range cg {
 565                         cg[idx] = remap[tagID(idx/2)][variant]
 566                 }
 567         }
 568         for _, refcs := range tilelib.refseqs {
 569                 for _, refseq := range refcs {
 570                         for i, tv := range refseq {
 571                                 refseq[i].Variant = remap[tv.Tag][tv.Variant]
 572                         }
 573                 }
 574         }
 575         log.Print("Tidy: done")
 576 }
 577
 578 func countBases(seq []byte) int {
 579         n := 0
 580         for _, c := range seq {
 581                 if isbase[c] {
 582                         n++
 583                 }
 584         }
 585         return n
 586 }