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