export.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         "errors"
  12         "flag"
  13         "fmt"
  14         "io"
  15         "net/http"
  16         _ "net/http/pprof"
  17         "os"
  18         "path/filepath"
  19         "runtime"
  20         "sort"
  21         "strconv"
  22         "strings"
  23         "sync"
  24         "time"
  25
  26         "git.arvados.org/arvados.git/sdk/go/arvados"
  27         "github.com/arvados/lightning/hgvs"
  28         "github.com/klauspost/pgzip"
  29         "github.com/kshedden/gonpy"
  30         "github.com/sirupsen/logrus"
  31         log "github.com/sirupsen/logrus"
  32 )
  33
  34 type tvVariant struct {
  35         hgvs.Variant
  36         librefs map[tileLibRef]bool
  37 }
  38
  39 type outputFormat interface {
  40         Filename() string
  41         PadLeft() bool
  42         Head(out io.Writer, cgs []CompactGenome) error
  43         Print(out io.Writer, seqname string, varslice []tvVariant) error
  44         Finish(outdir string, out io.Writer, seqname string) error
  45         MaxGoroutines() int
  46 }
  47
  48 var outputFormats = map[string]func() outputFormat{
  49         "hgvs-numpy": func() outputFormat {
  50                 return &formatHGVSNumpy{alleles: map[string][][]int8{}}
  51         },
  52         "hgvs-onehot": func() outputFormat { return formatHGVSOneHot{} },
  53         "hgvs":        func() outputFormat { return formatHGVS{} },
  54         "pvcf":        func() outputFormat { return formatPVCF{} },
  55         "vcf":         func() outputFormat { return formatVCF{} },
  56 }
  57
  58 type exporter struct {
  59         outputFormat   outputFormat
  60         outputPerChrom bool
  61         compress       bool
  62         maxTileSize    int
  63         filter         filter
  64 }
  65
  66 func (cmd *exporter) RunCommand(prog string, args []string, stdin io.Reader, stdout, stderr io.Writer) int {
  67         var err error
  68         defer func() {
  69                 if err != nil {
  70                         fmt.Fprintf(stderr, "%s\n", err)
  71                 }
  72         }()
  73         flags := flag.NewFlagSet("", flag.ContinueOnError)
  74         flags.SetOutput(stderr)
  75         pprof := flags.String("pprof", "", "serve Go profile data at http://`[addr]:port`")
  76         pprofdir := flags.String("pprof-dir", "", "write Go profile data to `directory` periodically")
  77         runlocal := flags.Bool("local", false, "run on local host (default: run in an arvados container)")
  78         projectUUID := flags.String("project", "", "project `UUID` for output data")
  79         priority := flags.Int("priority", 500, "container request priority")
  80         refname := flags.String("ref", "", "reference genome `name`")
  81         inputDir := flags.String("input-dir", ".", "input `directory`")
  82         outputDir := flags.String("output-dir", ".", "output `directory`")
  83         outputFormatStr := flags.String("output-format", "hgvs", "output `format`: hgvs, pvcf, or vcf")
  84         outputBed := flags.String("output-bed", "", "also output bed `file`")
  85         flags.BoolVar(&cmd.outputPerChrom, "output-per-chromosome", true, "output one file per chromosome")
  86         flags.BoolVar(&cmd.compress, "z", false, "write gzip-compressed output files")
  87         labelsFilename := flags.String("output-labels", "", "also output genome labels csv `file`")
  88         flags.IntVar(&cmd.maxTileSize, "max-tile-size", 50000, "don't try to make annotations for tiles bigger than given `size`")
  89         cmd.filter.Flags(flags)
  90         err = flags.Parse(args)
  91         if err == flag.ErrHelp {
  92                 err = nil
  93                 return 0
  94         } else if err != nil {
  95                 return 2
  96         }
  97         if flag.NArg() > 0 {
  98                 err = fmt.Errorf("extra unparsed command line arguments: %q", flag.Args())
  99                 return 2
 100         }
 101
 102         if f, ok := outputFormats[*outputFormatStr]; !ok {
 103                 err = fmt.Errorf("invalid output format %q", *outputFormatStr)
 104                 return 2
 105         } else {
 106                 cmd.outputFormat = f()
 107         }
 108
 109         if *pprof != "" {
 110                 go func() {
 111                         log.Println(http.ListenAndServe(*pprof, nil))
 112                 }()
 113         }
 114         if *pprofdir != "" {
 115                 go writeProfilesPeriodically(*pprofdir)
 116         }
 117
 118         if !*runlocal {
 119                 if *outputDir != "." {
 120                         err = errors.New("cannot specify output directory in container mode: not implemented")
 121                         return 1
 122                 }
 123                 runner := arvadosContainerRunner{
 124                         Name:        "lightning export",
 125                         Client:      arvados.NewClientFromEnv(),
 126                         ProjectUUID: *projectUUID,
 127                         RAM:         750000000000,
 128                         VCPUs:       96,
 129                         Priority:    *priority,
 130                         APIAccess:   true,
 131                 }
 132                 err = runner.TranslatePaths(inputDir)
 133                 if err != nil {
 134                         return 1
 135                 }
 136                 if *outputBed != "" {
 137                         if strings.Contains(*outputBed, "/") {
 138                                 err = fmt.Errorf("cannot use -output-bed filename %q containing '/' char", *outputBed)
 139                                 return 1
 140                         }
 141                         *outputBed = "/mnt/output/" + *outputBed
 142                 }
 143                 runner.Args = []string{"export", "-local=true",
 144                         "-pprof", ":6000",
 145                         "-pprof-dir", "/mnt/output",
 146                         "-ref", *refname,
 147                         "-output-format", *outputFormatStr,
 148                         "-output-bed", *outputBed,
 149                         "-output-labels", "/mnt/output/labels.csv",
 150                         "-output-per-chromosome=" + fmt.Sprintf("%v", cmd.outputPerChrom),
 151                         "-max-tile-size", fmt.Sprintf("%d", cmd.maxTileSize),
 152                         "-input-dir", *inputDir,
 153                         "-output-dir", "/mnt/output",
 154                         "-z=" + fmt.Sprintf("%v", cmd.compress),
 155                 }
 156                 runner.Args = append(runner.Args, cmd.filter.Args()...)
 157                 var output string
 158                 output, err = runner.Run()
 159                 if err != nil {
 160                         return 1
 161                 }
 162                 fmt.Fprintln(stdout, output)
 163                 return 0
 164         }
 165
 166         var cgs []CompactGenome
 167         tilelib := &tileLibrary{
 168                 retainNoCalls:       true,
 169                 retainTileSequences: true,
 170                 compactGenomes:      map[string][]tileVariantID{},
 171         }
 172         err = tilelib.LoadDir(context.Background(), *inputDir, nil)
 173         if err != nil {
 174                 return 1
 175         }
 176
 177         refseq, ok := tilelib.refseqs[*refname]
 178         if !ok {
 179                 err = fmt.Errorf("reference name %q not found in input; have %v", *refname, func() (names []string) {
 180                         for name := range tilelib.refseqs {
 181                                 names = append(names, name)
 182                         }
 183                         return
 184                 }())
 185                 return 1
 186         }
 187
 188         log.Infof("filtering: %+v", cmd.filter)
 189         cmd.filter.Apply(tilelib)
 190
 191         names := cgnames(tilelib)
 192         for _, name := range names {
 193                 cgs = append(cgs, CompactGenome{Name: name, Variants: tilelib.compactGenomes[name]})
 194         }
 195         if *labelsFilename != "" {
 196                 log.Infof("writing labels to %s", *labelsFilename)
 197                 var f *os.File
 198                 f, err = os.OpenFile(*labelsFilename, os.O_CREATE|os.O_WRONLY, 0777)
 199                 if err != nil {
 200                         return 1
 201                 }
 202                 defer f.Close()
 203                 for i, name := range names {
 204                         _, err = fmt.Fprintf(f, "%d,%q,%q\n", i, trimFilenameForLabel(name), cmd.outputFormat.Filename())
 205                         if err != nil {
 206                                 err = fmt.Errorf("write %s: %w", *labelsFilename, err)
 207                                 return 1
 208                         }
 209                 }
 210                 err = f.Close()
 211                 if err != nil {
 212                         err = fmt.Errorf("close %s: %w", *labelsFilename, err)
 213                         return 1
 214                 }
 215         }
 216
 217         var bedout io.Writer
 218         var bedfile *os.File
 219         var bedbufw *bufio.Writer
 220         if *outputBed != "" {
 221                 bedfile, err = os.OpenFile(*outputBed, os.O_CREATE|os.O_WRONLY, 0666)
 222                 if err != nil {
 223                         return 1
 224                 }
 225                 defer bedfile.Close()
 226                 bedbufw = bufio.NewWriterSize(bedfile, 16*1024*1024)
 227                 bedout = bedbufw
 228         }
 229
 230         err = cmd.export(*outputDir, bedout, tilelib, refseq, cgs)
 231         if err != nil {
 232                 return 1
 233         }
 234         if bedout != nil {
 235                 err = bedbufw.Flush()
 236                 if err != nil {
 237                         return 1
 238                 }
 239                 err = bedfile.Close()
 240                 if err != nil {
 241                         return 1
 242                 }
 243         }
 244         return 0
 245 }
 246
 247 func (cmd *exporter) export(outdir string, bedout io.Writer, tilelib *tileLibrary, refseq map[string][]tileLibRef, cgs []CompactGenome) error {
 248         var seqnames []string
 249         var missing []tileLibRef
 250         for seqname, librefs := range refseq {
 251                 seqnames = append(seqnames, seqname)
 252                 for _, libref := range librefs {
 253                         if libref.Variant != 0 && tilelib.TileVariantSequence(libref) == nil {
 254                                 missing = append(missing, libref)
 255                         }
 256                 }
 257         }
 258         sort.Strings(seqnames)
 259
 260         if len(missing) > 0 {
 261                 if limit := 100; len(missing) > limit {
 262                         log.Warnf("first %d missing tiles: %v", limit, missing[:limit])
 263                 } else {
 264                         log.Warnf("missing tiles: %v", missing)
 265                 }
 266                 return fmt.Errorf("%d needed tiles are missing from library", len(missing))
 267         }
 268
 269         outw := make([]io.WriteCloser, len(seqnames))
 270         bedw := make([]io.WriteCloser, len(seqnames))
 271
 272         var merges sync.WaitGroup
 273         merge := func(dst io.Writer, src []io.WriteCloser, label string) {
 274                 var mtx sync.Mutex
 275                 for i, seqname := range seqnames {
 276                         pr, pw := io.Pipe()
 277                         src[i] = pw
 278                         merges.Add(1)
 279                         seqname := seqname
 280                         go func() {
 281                                 defer merges.Done()
 282                                 log.Infof("writing %s %s", seqname, label)
 283                                 scanner := bufio.NewScanner(pr)
 284                                 for scanner.Scan() {
 285                                         mtx.Lock()
 286                                         dst.Write(scanner.Bytes())
 287                                         dst.Write([]byte{'\n'})
 288                                         mtx.Unlock()
 289                                 }
 290                                 log.Infof("writing %s %s done", seqname, label)
 291                         }()
 292                 }
 293         }
 294         if cmd.outputPerChrom {
 295                 for i, seqname := range seqnames {
 296                         fnm := filepath.Join(outdir, strings.Replace(cmd.outputFormat.Filename(), ".", "."+seqname+".", 1))
 297                         if cmd.compress {
 298                                 fnm += ".gz"
 299                         }
 300                         f, err := os.OpenFile(fnm, os.O_CREATE|os.O_WRONLY|os.O_TRUNC, 0666)
 301                         if err != nil {
 302                                 return err
 303                         }
 304                         defer f.Close()
 305                         log.Infof("writing %q", f.Name())
 306                         outw[i] = f
 307                         if cmd.compress {
 308                                 z := pgzip.NewWriter(f)
 309                                 defer z.Close()
 310                                 outw[i] = z
 311                         }
 312                         err = cmd.outputFormat.Head(outw[i], cgs)
 313                         if err != nil {
 314                                 return err
 315                         }
 316                 }
 317         } else {
 318                 fnm := filepath.Join(outdir, cmd.outputFormat.Filename())
 319                 if cmd.compress {
 320                         fnm += ".gz"
 321                 }
 322                 f, err := os.OpenFile(fnm, os.O_CREATE|os.O_WRONLY|os.O_TRUNC, 0666)
 323                 if err != nil {
 324                         return err
 325                 }
 326                 defer f.Close()
 327                 log.Infof("writing %q", fnm)
 328                 var out io.Writer = f
 329                 if cmd.compress {
 330                         z := pgzip.NewWriter(out)
 331                         defer z.Close()
 332                         out = z
 333                 }
 334                 cmd.outputFormat.Head(out, cgs)
 335                 merge(out, outw, "output")
 336         }
 337         if bedout != nil {
 338                 merge(bedout, bedw, "bed")
 339         }
 340
 341         throttle := throttle{Max: runtime.NumCPU()}
 342         if max := cmd.outputFormat.MaxGoroutines(); max > 0 {
 343                 throttle.Max = max
 344         }
 345         log.Infof("assembling %d sequences in %d goroutines", len(seqnames), throttle.Max)
 346         for seqidx, seqname := range seqnames {
 347                 seqidx, seqname := seqidx, seqname
 348                 outw := outw[seqidx]
 349                 bedw := bedw[seqidx]
 350                 throttle.Acquire()
 351                 go func() {
 352                         defer throttle.Release()
 353                         if bedw != nil {
 354                                 defer bedw.Close()
 355                         }
 356                         outwb := bufio.NewWriterSize(outw, 8*1024*1024)
 357                         eachVariant(bedw, tilelib.taglib.keylen, seqname, refseq[seqname], tilelib, cgs, cmd.outputFormat.PadLeft(), cmd.maxTileSize, func(varslice []tvVariant) {
 358                                 err := cmd.outputFormat.Print(outwb, seqname, varslice)
 359                                 throttle.Report(err)
 360                         })
 361                         err := cmd.outputFormat.Finish(outdir, outwb, seqname)
 362                         throttle.Report(err)
 363                         err = outwb.Flush()
 364                         throttle.Report(err)
 365                         err = outw.Close()
 366                         throttle.Report(err)
 367                 }()
 368         }
 369
 370         merges.Wait()
 371         throttle.Wait()
 372         return throttle.Err()
 373 }
 374
 375 // Align genome tiles to reference tiles, call callback func on each
 376 // variant, and (if bedw is not nil) write tile coverage to bedw.
 377 func eachVariant(bedw io.Writer, taglen int, seqname string, reftiles []tileLibRef, tilelib *tileLibrary, cgs []CompactGenome, padLeft bool, maxTileSize int, callback func(varslice []tvVariant)) {
 378         t0 := time.Now()
 379         progressbar := time.NewTicker(time.Minute)
 380         defer progressbar.Stop()
 381         var outmtx sync.Mutex
 382         defer outmtx.Lock()
 383         refpos := 0
 384         variantAt := map[int][]tvVariant{} // variantAt[chromOffset][genomeIndex*2+phase]
 385         for refstep, libref := range reftiles {
 386                 select {
 387                 case <-progressbar.C:
 388                         var eta interface{}
 389                         if refstep > 0 {
 390                                 fin := t0.Add(time.Duration(float64(time.Now().Sub(t0)) * float64(len(reftiles)) / float64(refstep)))
 391                                 eta = fmt.Sprintf("%v (%v)", fin.Format(time.RFC3339), fin.Sub(time.Now()))
 392                         } else {
 393                                 eta = "N/A"
 394                         }
 395                         log.Printf("exportSeq: %s: refstep %d of %d, %.0f/s, ETA %v", seqname, refstep, len(reftiles), float64(refstep)/time.Now().Sub(t0).Seconds(), eta)
 396                 default:
 397                 }
 398                 diffs := map[tileLibRef][]hgvs.Variant{}
 399                 refseq := tilelib.TileVariantSequence(libref)
 400                 tagcoverage := 0 // number of times the start tag was found in genomes -- max is len(cgs)*2
 401                 for cgidx, cg := range cgs {
 402                         for phase := 0; phase < 2; phase++ {
 403                                 var variant tileVariantID
 404                                 if i := int(libref.Tag)*2 + phase; len(cg.Variants) > i {
 405                                         variant = cg.Variants[i]
 406                                 }
 407                                 if variant > 0 {
 408                                         tagcoverage++
 409                                 }
 410                                 if variant == libref.Variant || variant == 0 {
 411                                         continue
 412                                 }
 413                                 glibref := tileLibRef{Tag: libref.Tag, Variant: variant}
 414                                 vars, ok := diffs[glibref]
 415                                 if !ok {
 416                                         genomeseq := tilelib.TileVariantSequence(glibref)
 417                                         if len(genomeseq) == 0 {
 418                                                 // Hash is known but sequence
 419                                                 // is not, e.g., retainNoCalls
 420                                                 // was false during import
 421                                                 continue
 422                                         }
 423                                         if len(genomeseq) > maxTileSize {
 424                                                 continue
 425                                         }
 426                                         refSequence := refseq
 427                                         // If needed, extend the
 428                                         // reference sequence up to
 429                                         // the tag at the end of the
 430                                         // genomeseq sequence.
 431                                         refstepend := refstep + 1
 432                                         for refstepend < len(reftiles) && len(refSequence) >= taglen && !bytes.EqualFold(refSequence[len(refSequence)-taglen:], genomeseq[len(genomeseq)-taglen:]) && len(refSequence) <= maxTileSize {
 433                                                 if &refSequence[0] == &refseq[0] {
 434                                                         refSequence = append([]byte(nil), refSequence...)
 435                                                 }
 436                                                 refSequence = append(refSequence, tilelib.TileVariantSequence(reftiles[refstepend])...)
 437                                                 refstepend++
 438                                         }
 439                                         // (TODO: handle no-calls)
 440                                         if len(refSequence) <= maxTileSize {
 441                                                 refstr := strings.ToUpper(string(refSequence))
 442                                                 genomestr := strings.ToUpper(string(genomeseq))
 443                                                 vars, _ = hgvs.Diff(refstr, genomestr, time.Second)
 444                                         }
 445                                         diffs[glibref] = vars
 446                                 }
 447                                 for _, v := range vars {
 448                                         if padLeft {
 449                                                 v = v.PadLeft()
 450                                         }
 451                                         v.Position += refpos
 452                                         varslice := variantAt[v.Position]
 453                                         if varslice == nil {
 454                                                 varslice = make([]tvVariant, len(cgs)*2)
 455                                                 variantAt[v.Position] = varslice
 456                                         }
 457                                         varslice[cgidx*2+phase].Variant = v
 458                                         if varslice[cgidx*2+phase].librefs == nil {
 459                                                 varslice[cgidx*2+phase].librefs = map[tileLibRef]bool{glibref: true}
 460                                         } else {
 461                                                 varslice[cgidx*2+phase].librefs[glibref] = true
 462                                         }
 463                                 }
 464                         }
 465                 }
 466                 refpos += len(refseq) - taglen
 467
 468                 // Flush entries from variantAt that are behind
 469                 // refpos. Flush all entries if this is the last
 470                 // reftile of the path/chromosome.
 471                 flushpos := make([]int, 0, len(variantAt))
 472                 lastrefstep := refstep == len(reftiles)-1
 473                 for pos := range variantAt {
 474                         if lastrefstep || pos <= refpos {
 475                                 flushpos = append(flushpos, pos)
 476                         }
 477                 }
 478                 sort.Slice(flushpos, func(i, j int) bool { return flushpos[i] < flushpos[j] })
 479                 flushvariants := make([][]tvVariant, len(flushpos))
 480                 for i, pos := range flushpos {
 481                         varslice := variantAt[pos]
 482                         delete(variantAt, pos)
 483                         // Check for uninitialized (zero-value)
 484                         // elements in varslice
 485                         for i := range varslice {
 486                                 if varslice[i].Position != 0 {
 487                                         // Not a zero-value element
 488                                         continue
 489                                 }
 490                                 // Set the position so
 491                                 // varslice[*].Position are all equal
 492                                 varslice[i].Position = pos
 493                                 // This could be either =ref or a
 494                                 // missing/low-quality tile. Figure
 495                                 // out which.
 496                                 v := cgs[i/2].Variants[int(libref.Tag)*2+i%2]
 497                                 if v < 1 || len(tilelib.TileVariantSequence(tileLibRef{Tag: libref.Tag, Variant: v})) == 0 {
 498                                         // Missing/low-quality tile.
 499                                         varslice[i].New = "-" // fasta "gap of indeterminate length"
 500                                 }
 501                         }
 502                         flushvariants[i] = varslice
 503                 }
 504                 outmtx.Lock()
 505                 go func() {
 506                         defer outmtx.Unlock()
 507                         for _, varslice := range flushvariants {
 508                                 callback(varslice)
 509                         }
 510                 }()
 511                 if bedw != nil && len(refseq) > 0 {
 512                         tilestart := refpos - len(refseq) + taglen
 513                         tileend := refpos
 514                         if !lastrefstep {
 515                                 tileend += taglen
 516                         }
 517                         thickstart := tilestart + taglen
 518                         if refstep == 0 {
 519                                 thickstart = 0
 520                         }
 521                         thickend := refpos
 522
 523                         // coverage score, 0 to 1000
 524                         score := 1000
 525                         if len(cgs) > 0 {
 526                                 score = 1000 * tagcoverage / len(cgs) / 2
 527                         }
 528
 529                         fmt.Fprintf(bedw, "%s %d %d %d %d . %d %d\n",
 530                                 seqname, tilestart, tileend,
 531                                 libref.Tag,
 532                                 score,
 533                                 thickstart, thickend)
 534                 }
 535         }
 536 }
 537
 538 func bucketVarsliceByRef(varslice []tvVariant) map[string]map[string]int {
 539         byref := map[string]map[string]int{}
 540         for _, v := range varslice {
 541                 if v.Ref == "" && v.New == "" {
 542                         // =ref
 543                         continue
 544                 }
 545                 if v.New == "-" {
 546                         // no-call
 547                         continue
 548                 }
 549                 alts := byref[v.Ref]
 550                 if alts == nil {
 551                         alts = map[string]int{}
 552                         byref[v.Ref] = alts
 553                 }
 554                 alts[v.New]++
 555         }
 556         return byref
 557 }
 558
 559 type formatVCF struct{}
 560
 561 func (formatVCF) MaxGoroutines() int                     { return 0 }
 562 func (formatVCF) Filename() string                       { return "out.vcf" }
 563 func (formatVCF) PadLeft() bool                          { return true }
 564 func (formatVCF) Finish(string, io.Writer, string) error { return nil }
 565 func (formatVCF) Head(out io.Writer, cgs []CompactGenome) error {
 566         _, err := fmt.Fprint(out, "#CHROM\tPOS\tID\tREF\tALT\tQUAL\tFILTER\tINFO\n")
 567         return err
 568 }
 569 func (formatVCF) Print(out io.Writer, seqname string, varslice []tvVariant) error {
 570         for ref, alts := range bucketVarsliceByRef(varslice) {
 571                 altslice := make([]string, 0, len(alts))
 572                 for alt := range alts {
 573                         altslice = append(altslice, alt)
 574                 }
 575                 sort.Strings(altslice)
 576
 577                 info := "AC="
 578                 for i, a := range altslice {
 579                         if i > 0 {
 580                                 info += ","
 581                         }
 582                         info += strconv.Itoa(alts[a])
 583                 }
 584                 _, err := fmt.Fprintf(out, "%s\t%d\t.\t%s\t%s\t.\t.\t%s\n", seqname, varslice[0].Position, ref, strings.Join(altslice, ","), info)
 585                 if err != nil {
 586                         return err
 587                 }
 588         }
 589         return nil
 590 }
 591
 592 type formatPVCF struct{}
 593
 594 func (formatPVCF) MaxGoroutines() int                     { return 0 }
 595 func (formatPVCF) Filename() string                       { return "out.vcf" }
 596 func (formatPVCF) PadLeft() bool                          { return true }
 597 func (formatPVCF) Finish(string, io.Writer, string) error { return nil }
 598 func (formatPVCF) Head(out io.Writer, cgs []CompactGenome) error {
 599         fmt.Fprintln(out, `##FORMAT=<ID=GT,Number=1,Type=String,Description="Genotype">`)
 600         fmt.Fprintf(out, "#CHROM\tPOS\tID\tREF\tALT\tQUAL\tFILTER\tINFO\tFORMAT")
 601         for _, cg := range cgs {
 602                 fmt.Fprintf(out, "\t%s", cg.Name)
 603         }
 604         _, err := fmt.Fprintf(out, "\n")
 605         return err
 606 }
 607
 608 func (formatPVCF) Print(out io.Writer, seqname string, varslice []tvVariant) error {
 609         for ref, alts := range bucketVarsliceByRef(varslice) {
 610                 altslice := make([]string, 0, len(alts))
 611                 for alt := range alts {
 612                         altslice = append(altslice, alt)
 613                 }
 614                 sort.Strings(altslice)
 615                 for i, a := range altslice {
 616                         alts[a] = i + 1
 617                 }
 618                 _, err := fmt.Fprintf(out, "%s\t%d\t.\t%s\t%s\t.\t.\t.\tGT", seqname, varslice[0].Position, ref, strings.Join(altslice, ","))
 619                 if err != nil {
 620                         return err
 621                 }
 622                 for i := 0; i < len(varslice); i += 2 {
 623                         v1, v2 := varslice[i], varslice[i+1]
 624                         a1, a2 := alts[v1.New], alts[v2.New]
 625                         if v1.Ref != ref {
 626                                 // variant on allele 0 belongs on a
 627                                 // different output line -- same
 628                                 // chr,pos but different "ref" length
 629                                 a1 = 0
 630                         }
 631                         if v2.Ref != ref {
 632                                 a2 = 0
 633                         }
 634                         _, err := fmt.Fprintf(out, "\t%d/%d", a1, a2)
 635                         if err != nil {
 636                                 return err
 637                         }
 638                 }
 639                 _, err = out.Write([]byte{'\n'})
 640                 if err != nil {
 641                         return err
 642                 }
 643         }
 644         return nil
 645 }
 646
 647 type formatHGVS struct{}
 648
 649 func (formatHGVS) MaxGoroutines() int                            { return 0 }
 650 func (formatHGVS) Filename() string                              { return "out.tsv" }
 651 func (formatHGVS) PadLeft() bool                                 { return false }
 652 func (formatHGVS) Head(out io.Writer, cgs []CompactGenome) error { return nil }
 653 func (formatHGVS) Finish(string, io.Writer, string) error        { return nil }
 654 func (formatHGVS) Print(out io.Writer, seqname string, varslice []tvVariant) error {
 655         for i := 0; i < len(varslice)/2; i++ {
 656                 if i > 0 {
 657                         out.Write([]byte{'\t'})
 658                 }
 659                 var1, var2 := varslice[i*2], varslice[i*2+1]
 660                 if var1.New == "-" || var2.New == "-" {
 661                         _, err := out.Write([]byte{'N'})
 662                         if err != nil {
 663                                 return err
 664                         }
 665                         continue
 666                 }
 667                 if var1.Variant == var2.Variant {
 668                         if var1.Ref == var1.New {
 669                                 _, err := out.Write([]byte{'.'})
 670                                 if err != nil {
 671                                         return err
 672                                 }
 673                         } else {
 674                                 _, err := fmt.Fprintf(out, "%s:g.%s", seqname, var1.String())
 675                                 if err != nil {
 676                                         return err
 677                                 }
 678                         }
 679                 } else {
 680                         _, err := fmt.Fprintf(out, "%s:g.[%s];[%s]", seqname, var1.String(), var2.String())
 681                         if err != nil {
 682                                 return err
 683                         }
 684                 }
 685         }
 686         _, err := out.Write([]byte{'\n'})
 687         return err
 688 }
 689
 690 type formatHGVSOneHot struct{}
 691
 692 func (formatHGVSOneHot) MaxGoroutines() int                            { return 0 }
 693 func (formatHGVSOneHot) Filename() string                              { return "out.tsv" }
 694 func (formatHGVSOneHot) PadLeft() bool                                 { return false }
 695 func (formatHGVSOneHot) Head(out io.Writer, cgs []CompactGenome) error { return nil }
 696 func (formatHGVSOneHot) Finish(string, io.Writer, string) error        { return nil }
 697 func (formatHGVSOneHot) Print(out io.Writer, seqname string, varslice []tvVariant) error {
 698         vars := map[hgvs.Variant]bool{}
 699         for _, v := range varslice {
 700                 if v.Ref != v.New {
 701                         vars[v.Variant] = true
 702                 }
 703         }
 704
 705         // sort variants to ensure output is deterministic
 706         sorted := make([]hgvs.Variant, 0, len(vars))
 707         for v := range vars {
 708                 sorted = append(sorted, v)
 709         }
 710         sort.Slice(sorted, func(a, b int) bool { return hgvs.Less(sorted[a], sorted[b]) })
 711
 712         for _, v := range sorted {
 713                 if v.New == "-" {
 714                         continue
 715                 }
 716                 fmt.Fprintf(out, "%s.%s", seqname, v.String())
 717                 for i := 0; i < len(varslice); i += 2 {
 718                         if varslice[i].Variant == v || varslice[i+1].Variant == v {
 719                                 out.Write([]byte("\t1"))
 720                         } else {
 721                                 out.Write([]byte("\t0"))
 722                         }
 723                 }
 724                 _, err := out.Write([]byte{'\n'})
 725                 if err != nil {
 726                         return err
 727                 }
 728         }
 729         return nil
 730 }
 731
 732 type formatHGVSNumpy struct {
 733         sync.Mutex
 734         writelock sync.Mutex
 735         alleles   map[string][][]int8 // alleles[seqname][variantidx][genomeidx*2+phase]
 736 }
 737
 738 func (*formatHGVSNumpy) MaxGoroutines() int                            { return 4 }
 739 func (*formatHGVSNumpy) Filename() string                              { return "annotations.csv" }
 740 func (*formatHGVSNumpy) PadLeft() bool                                 { return false }
 741 func (*formatHGVSNumpy) Head(out io.Writer, cgs []CompactGenome) error { return nil }
 742 func (f *formatHGVSNumpy) Print(outw io.Writer, seqname string, varslice []tvVariant) error {
 743         // sort variants to ensure output is deterministic
 744         sorted := make([]hgvs.Variant, 0, len(varslice))
 745         for _, v := range varslice {
 746                 sorted = append(sorted, v.Variant)
 747         }
 748         sort.Slice(sorted, func(a, b int) bool { return hgvs.Less(sorted[a], sorted[b]) })
 749
 750         f.Lock()
 751         seqalleles := f.alleles[seqname]
 752         f.Unlock()
 753
 754         // append a row to seqalleles for each unique non-ref variant
 755         // in varslice.
 756         var previous hgvs.Variant
 757         for _, v := range sorted {
 758                 if previous == v || v.Ref == v.New || v.New == "-" {
 759                         continue
 760                 }
 761                 previous = v
 762                 newrow := make([]int8, len(varslice))
 763                 for i, allele := range varslice {
 764                         if allele.Variant == v {
 765                                 newrow[i] = 1
 766                         } else if allele.Variant.New == "-" {
 767                                 newrow[i] = -1
 768                         }
 769                 }
 770                 seqalleles = append(seqalleles, newrow)
 771                 _, err := fmt.Fprintf(outw, "%d,%q\n", len(seqalleles)-1, seqname+"."+v.String())
 772                 if err != nil {
 773                         return err
 774                 }
 775         }
 776
 777         f.Lock()
 778         f.alleles[seqname] = seqalleles
 779         f.Unlock()
 780         return nil
 781 }
 782 func (f *formatHGVSNumpy) Finish(outdir string, _ io.Writer, seqname string) error {
 783         // Write seqname's data to a .npy matrix with one row per
 784         // genome and 2 columns per variant.
 785         f.Lock()
 786         seqalleles := f.alleles[seqname]
 787         delete(f.alleles, seqname)
 788         f.Unlock()
 789         if len(seqalleles) == 0 {
 790                 return nil
 791         }
 792         out := make([]int8, len(seqalleles)*len(seqalleles[0]))
 793         rows := len(seqalleles[0]) / 2
 794         cols := len(seqalleles) * 2
 795         // copy seqalleles[varidx][genome*2+phase] to
 796         // out[genome*nvars*2 + varidx*2 + phase]
 797         for varidx, alleles := range seqalleles {
 798                 for g := 0; g < len(alleles)/2; g++ {
 799                         aa, ab := alleles[g*2], alleles[g*2+1]
 800                         if aa < 0 || ab < 0 {
 801                                 // no-call
 802                                 out[g*cols+varidx*2] = -1
 803                                 out[g*cols+varidx*2+1] = -1
 804                         } else if aa > 0 && ab > 0 {
 805                                 // hom
 806                                 out[g*cols+varidx*2] = 1
 807                         } else if aa > 0 || ab > 0 {
 808                                 // het
 809                                 out[g*cols+varidx*2+1] = 1
 810                         }
 811                 }
 812         }
 813         outf, err := os.OpenFile(outdir+"/matrix."+seqname+".npy", os.O_CREATE|os.O_EXCL|os.O_WRONLY, 0777)
 814         if err != nil {
 815                 return err
 816         }
 817         defer outf.Close()
 818         bufw := bufio.NewWriter(outf)
 819         npw, err := gonpy.NewWriter(nopCloser{bufw})
 820         if err != nil {
 821                 return err
 822         }
 823         log.WithFields(logrus.Fields{
 824                 "seqname": seqname,
 825                 "rows":    rows,
 826                 "cols":    cols,
 827         }).Info("writing numpy")
 828         npw.Shape = []int{rows, cols}
 829         f.writelock.Lock() // serialize because WriteInt8 uses lots of memory
 830         npw.WriteInt8(out)
 831         f.writelock.Unlock()
 832         err = bufw.Flush()
 833         if err != nil {
 834                 return err
 835         }
 836         err = outf.Close()
 837         if err != nil {
 838                 return err
 839         }
 840         return nil
 841 }