export.go

   1 package main
   2
   3 import (
   4         "bufio"
   5         "bytes"
   6         "context"
   7         "errors"
   8         "flag"
   9         "fmt"
  10         "io"
  11         "net/http"
  12         _ "net/http/pprof"
  13         "os"
  14         "sort"
  15         "strings"
  16         "sync"
  17         "time"
  18
  19         "git.arvados.org/arvados.git/sdk/go/arvados"
  20         "github.com/arvados/lightning/hgvs"
  21         log "github.com/sirupsen/logrus"
  22 )
  23
  24 type outputFormat struct {
  25         Print   func(out io.Writer, seqname string, varslice []hgvs.Variant)
  26         PadLeft bool
  27 }
  28
  29 var (
  30         outputFormats = map[string]outputFormat{
  31                 "hgvs": outputFormatHGVS,
  32                 "vcf":  outputFormatVCF,
  33         }
  34         outputFormatHGVS = outputFormat{Print: printHGVS}
  35         outputFormatVCF  = outputFormat{Print: printVCF, PadLeft: true}
  36 )
  37
  38 type exporter struct {
  39         outputFormat outputFormat
  40 }
  41
  42 func (cmd *exporter) RunCommand(prog string, args []string, stdin io.Reader, stdout, stderr io.Writer) int {
  43         var err error
  44         defer func() {
  45                 if err != nil {
  46                         fmt.Fprintf(stderr, "%s\n", err)
  47                 }
  48         }()
  49         flags := flag.NewFlagSet("", flag.ContinueOnError)
  50         flags.SetOutput(stderr)
  51         pprof := flags.String("pprof", "", "serve Go profile data at http://`[addr]:port`")
  52         runlocal := flags.Bool("local", false, "run on local host (default: run in an arvados container)")
  53         projectUUID := flags.String("project", "", "project `UUID` for output data")
  54         priority := flags.Int("priority", 500, "container request priority")
  55         refname := flags.String("ref", "", "reference genome `name`")
  56         inputFilename := flags.String("i", "-", "input `file` (library)")
  57         outputFilename := flags.String("o", "-", "output `file`")
  58         outputFormatStr := flags.String("output-format", "hgvs", "output `format`: hgvs or vcf")
  59         outputBed := flags.String("output-bed", "", "also output bed `file`")
  60         pick := flags.String("pick", "", "`name` of single genome to export")
  61         err = flags.Parse(args)
  62         if err == flag.ErrHelp {
  63                 err = nil
  64                 return 0
  65         } else if err != nil {
  66                 return 2
  67         }
  68
  69         if f, ok := outputFormats[*outputFormatStr]; !ok {
  70                 err = fmt.Errorf("invalid output format %q", *outputFormatStr)
  71                 return 2
  72         } else {
  73                 cmd.outputFormat = f
  74         }
  75
  76         if *pprof != "" {
  77                 go func() {
  78                         log.Println(http.ListenAndServe(*pprof, nil))
  79                 }()
  80         }
  81
  82         if !*runlocal {
  83                 if *outputFilename != "-" {
  84                         err = errors.New("cannot specify output file in container mode: not implemented")
  85                         return 1
  86                 }
  87                 runner := arvadosContainerRunner{
  88                         Name:        "lightning export",
  89                         Client:      arvados.NewClientFromEnv(),
  90                         ProjectUUID: *projectUUID,
  91                         RAM:         240000000000,
  92                         VCPUs:       32,
  93                         Priority:    *priority,
  94                 }
  95                 err = runner.TranslatePaths(inputFilename)
  96                 if err != nil {
  97                         return 1
  98                 }
  99                 if *outputBed != "" {
 100                         if strings.Contains(*outputBed, "/") {
 101                                 err = fmt.Errorf("cannot use -output-bed filename %q containing '/' char", *outputBed)
 102                                 return 1
 103                         }
 104                         *outputBed = "/mnt/output/" + *outputBed
 105                 }
 106                 runner.Args = []string{"export", "-local=true", "-pick", *pick, "-ref", *refname, "-output-format", *outputFormatStr, "-output-bed", *outputBed, "-i", *inputFilename, "-o", "/mnt/output/export.csv"}
 107                 var output string
 108                 output, err = runner.Run()
 109                 if err != nil {
 110                         return 1
 111                 }
 112                 fmt.Fprintln(stdout, output+"/export.csv")
 113                 return 0
 114         }
 115
 116         input, err := os.Open(*inputFilename)
 117         if err != nil {
 118                 return 1
 119         }
 120         defer input.Close()
 121
 122         // Error out early if seeking doesn't work on the input file.
 123         _, err = input.Seek(0, io.SeekEnd)
 124         if err != nil {
 125                 return 1
 126         }
 127         _, err = input.Seek(0, io.SeekStart)
 128         if err != nil {
 129                 return 1
 130         }
 131
 132         var mtx sync.Mutex
 133         var cgs []CompactGenome
 134         tilelib := tileLibrary{
 135                 retainNoCalls: true,
 136         }
 137         err = tilelib.LoadGob(context.Background(), input, strings.HasSuffix(*inputFilename, ".gz"), func(cg CompactGenome) {
 138                 if *pick != "" && *pick != cg.Name {
 139                         return
 140                 }
 141                 log.Debugf("export: pick %q", cg.Name)
 142                 mtx.Lock()
 143                 defer mtx.Unlock()
 144                 cgs = append(cgs, cg)
 145         })
 146         if err != nil {
 147                 return 1
 148         }
 149         sort.Slice(cgs, func(i, j int) bool { return cgs[i].Name < cgs[j].Name })
 150         log.Printf("export: pick %q => %d genomes", *pick, len(cgs))
 151
 152         refseq, ok := tilelib.refseqs[*refname]
 153         if !ok {
 154                 err = fmt.Errorf("reference name %q not found in input; have %v", *refname, func() (names []string) {
 155                         for name := range tilelib.refseqs {
 156                                 names = append(names, name)
 157                         }
 158                         return
 159                 }())
 160                 return 1
 161         }
 162
 163         _, err = input.Seek(0, io.SeekStart)
 164         if err != nil {
 165                 return 1
 166         }
 167
 168         var output io.WriteCloser
 169         if *outputFilename == "-" {
 170                 output = nopCloser{stdout}
 171         } else {
 172                 output, err = os.OpenFile(*outputFilename, os.O_CREATE|os.O_WRONLY, 0666)
 173                 if err != nil {
 174                         return 1
 175                 }
 176                 defer output.Close()
 177         }
 178         bufw := bufio.NewWriter(output)
 179
 180         var bedout *os.File
 181         var bedbufw *bufio.Writer
 182         if *outputBed != "" {
 183                 bedout, err = os.OpenFile(*outputBed, os.O_CREATE|os.O_WRONLY, 0666)
 184                 if err != nil {
 185                         return 1
 186                 }
 187                 defer bedout.Close()
 188                 bedbufw = bufio.NewWriter(bedout)
 189         }
 190
 191         err = cmd.export(bufw, bedout, input, strings.HasSuffix(*inputFilename, ".gz"), tilelib.taglib.keylen, refseq, cgs)
 192         if err != nil {
 193                 return 1
 194         }
 195         err = bufw.Flush()
 196         if err != nil {
 197                 return 1
 198         }
 199         err = output.Close()
 200         if err != nil {
 201                 return 1
 202         }
 203         if bedout != nil {
 204                 err = bedbufw.Flush()
 205                 if err != nil {
 206                         return 1
 207                 }
 208                 err = bedout.Close()
 209                 if err != nil {
 210                         return 1
 211                 }
 212         }
 213         err = input.Close()
 214         if err != nil {
 215                 return 1
 216         }
 217         return 0
 218 }
 219
 220 func (cmd *exporter) export(out, bedout io.Writer, librdr io.Reader, gz bool, taglen int, refseq map[string][]tileLibRef, cgs []CompactGenome) error {
 221         need := map[tileLibRef]bool{}
 222         var seqnames []string
 223         for seqname, librefs := range refseq {
 224                 seqnames = append(seqnames, seqname)
 225                 for _, libref := range librefs {
 226                         if libref.Variant != 0 {
 227                                 need[libref] = true
 228                         }
 229                 }
 230         }
 231         sort.Strings(seqnames)
 232
 233         for _, cg := range cgs {
 234                 for i, variant := range cg.Variants {
 235                         if variant == 0 {
 236                                 continue
 237                         }
 238                         libref := tileLibRef{Tag: tagID(i / 2), Variant: variant}
 239                         need[libref] = true
 240                 }
 241         }
 242
 243         log.Infof("export: loading %d tile variants", len(need))
 244         tileVariant := map[tileLibRef]TileVariant{}
 245         err := DecodeLibrary(librdr, gz, func(ent *LibraryEntry) error {
 246                 for _, tv := range ent.TileVariants {
 247                         libref := tileLibRef{Tag: tv.Tag, Variant: tv.Variant}
 248                         if need[libref] {
 249                                 tileVariant[libref] = tv
 250                         }
 251                 }
 252                 return nil
 253         })
 254         if err != nil {
 255                 return err
 256         }
 257
 258         log.Infof("export: loaded %d tile variants", len(tileVariant))
 259         var missing []tileLibRef
 260         for libref := range need {
 261                 if _, ok := tileVariant[libref]; !ok {
 262                         missing = append(missing, libref)
 263                 }
 264         }
 265         if len(missing) > 0 {
 266                 if limit := 100; len(missing) > limit {
 267                         log.Warnf("first %d missing tiles: %v", limit, missing[:limit])
 268                 } else {
 269                         log.Warnf("missing tiles: %v", missing)
 270                 }
 271                 return fmt.Errorf("%d needed tiles are missing from library", len(missing))
 272         }
 273
 274         log.Infof("assembling %d sequences concurrently", len(seqnames))
 275         var wg sync.WaitGroup
 276         outbuf := make([]bytes.Buffer, len(seqnames))
 277         bedbuf := make([]bytes.Buffer, len(seqnames))
 278         for seqidx, seqname := range seqnames {
 279                 seqname := seqname
 280                 outbuf := &outbuf[seqidx]
 281                 bedbuf := &bedbuf[seqidx]
 282                 if bedout == nil {
 283                         bedbuf = nil
 284                 }
 285                 // TODO: limit number of goroutines and unflushed bufs to ncpus
 286                 wg.Add(1)
 287                 go func() {
 288                         defer wg.Done()
 289                         cmd.exportSeq(outbuf, bedbuf, taglen, seqname, refseq[seqname], tileVariant, cgs)
 290                         log.Infof("assembled %q to outbuf %d bedbuf %d", seqname, outbuf.Len(), bedbuf.Len())
 291                 }()
 292         }
 293         wg.Wait()
 294
 295         wg.Add(1)
 296         go func() {
 297                 defer wg.Done()
 298                 for i, seqname := range seqnames {
 299                         log.Infof("writing outbuf %s", seqname)
 300                         io.Copy(out, &outbuf[i])
 301                 }
 302         }()
 303         if bedout != nil {
 304                 wg.Add(1)
 305                 go func() {
 306                         defer wg.Done()
 307                         for i, seqname := range seqnames {
 308                                 log.Infof("writing bedbuf %s", seqname)
 309                                 io.Copy(bedout, &bedbuf[i])
 310                         }
 311                 }()
 312         }
 313         wg.Wait()
 314         return nil
 315 }
 316
 317 // Align genome tiles to reference tiles, write diffs to outw, and (if
 318 // bedw is not nil) write tile coverage to bedw.
 319 func (cmd *exporter) exportSeq(outw, bedw io.Writer, taglen int, seqname string, reftiles []tileLibRef, tileVariant map[tileLibRef]TileVariant, cgs []CompactGenome) {
 320         refpos := 0
 321         variantAt := map[int][]hgvs.Variant{} // variantAt[chromOffset][genomeIndex*2+phase]
 322         for refstep, libref := range reftiles {
 323                 reftile := tileVariant[libref]
 324                 tagcoverage := 0 // number of times the start tag was found in genomes -- max is len(cgs)*2
 325                 for cgidx, cg := range cgs {
 326                         for phase := 0; phase < 2; phase++ {
 327                                 if len(cg.Variants) <= int(libref.Tag)*2+phase {
 328                                         continue
 329                                 }
 330                                 variant := cg.Variants[int(libref.Tag)*2+phase]
 331                                 if variant == 0 {
 332                                         continue
 333                                 }
 334                                 tagcoverage++
 335                                 if variant == libref.Variant {
 336                                         continue
 337                                 }
 338                                 genometile := tileVariant[tileLibRef{Tag: libref.Tag, Variant: variant}]
 339                                 if len(genometile.Sequence) == 0 {
 340                                         // Hash is known but sequence
 341                                         // is not, e.g., retainNoCalls
 342                                         // was false during import
 343                                         continue
 344                                 }
 345                                 refSequence := reftile.Sequence
 346                                 // If needed, extend the reference
 347                                 // sequence up to the tag at the end
 348                                 // of the genometile sequence.
 349                                 refstepend := refstep + 1
 350                                 for refstepend < len(reftiles) && len(refSequence) >= taglen && !bytes.EqualFold(refSequence[len(refSequence)-taglen:], genometile.Sequence[len(genometile.Sequence)-taglen:]) {
 351                                         if &refSequence[0] == &reftile.Sequence[0] {
 352                                                 refSequence = append([]byte(nil), refSequence...)
 353                                         }
 354                                         refSequence = append(refSequence, tileVariant[reftiles[refstepend]].Sequence...)
 355                                         refstepend++
 356                                 }
 357                                 // (TODO: handle no-calls)
 358                                 vars, _ := hgvs.Diff(strings.ToUpper(string(refSequence)), strings.ToUpper(string(genometile.Sequence)), time.Second)
 359                                 for _, v := range vars {
 360                                         if cmd.outputFormat.PadLeft {
 361                                                 v = v.PadLeft()
 362                                         }
 363                                         v.Position += refpos
 364                                         log.Debugf("%s seq %s phase %d tag %d tile diff %s\n", cg.Name, seqname, phase, libref.Tag, v.String())
 365                                         varslice := variantAt[v.Position]
 366                                         if varslice == nil {
 367                                                 varslice = make([]hgvs.Variant, len(cgs)*2)
 368                                                 variantAt[v.Position] = varslice
 369                                         }
 370                                         varslice[cgidx*2+phase] = v
 371                                 }
 372                         }
 373                 }
 374                 refpos += len(reftile.Sequence) - taglen
 375
 376                 // Flush entries from variantAt that are behind
 377                 // refpos. Flush all entries if this is the last
 378                 // reftile of the path/chromosome.
 379                 var flushpos []int
 380                 lastrefstep := refstep == len(reftiles)-1
 381                 for pos := range variantAt {
 382                         if lastrefstep || pos <= refpos {
 383                                 flushpos = append(flushpos, pos)
 384                         }
 385                 }
 386                 sort.Slice(flushpos, func(i, j int) bool { return flushpos[i] < flushpos[j] })
 387                 for _, pos := range flushpos {
 388                         varslice := variantAt[pos]
 389                         delete(variantAt, pos)
 390                         for i := range varslice {
 391                                 if varslice[i].Position == 0 {
 392                                         varslice[i].Position = pos
 393                                 }
 394                         }
 395                         cmd.outputFormat.Print(outw, seqname, varslice)
 396                 }
 397                 if bedw != nil && len(reftile.Sequence) > 0 {
 398                         tilestart := refpos - len(reftile.Sequence) + taglen
 399                         tileend := refpos
 400                         if !lastrefstep {
 401                                 tileend += taglen
 402                         }
 403                         thickstart := tilestart + taglen
 404                         if refstep == 0 {
 405                                 thickstart = 0
 406                         }
 407                         thickend := refpos
 408
 409                         // coverage score, 0 to 1000
 410                         score := 1000
 411                         if len(cgs) > 0 {
 412                                 score = 1000 * tagcoverage / len(cgs) / 2
 413                         }
 414
 415                         fmt.Fprintf(bedw, "%s %d %d %d %d . %d %d\n",
 416                                 seqname, tilestart, tileend,
 417                                 libref.Tag,
 418                                 score,
 419                                 thickstart, thickend)
 420                 }
 421         }
 422 }
 423
 424 func printVCF(out io.Writer, seqname string, varslice []hgvs.Variant) {
 425         refs := map[string]map[string]int{}
 426         for _, v := range varslice {
 427                 if v.Ref == "" && v.New == "" {
 428                         continue
 429                 }
 430                 alts := refs[v.Ref]
 431                 if alts == nil {
 432                         alts = map[string]int{}
 433                         refs[v.Ref] = alts
 434                 }
 435                 alts[v.New] = 0
 436         }
 437         for ref, alts := range refs {
 438                 var altslice []string
 439                 for alt := range alts {
 440                         altslice = append(altslice, alt)
 441                 }
 442                 sort.Strings(altslice)
 443                 for i, a := range altslice {
 444                         alts[a] = i + 1
 445                 }
 446                 fmt.Fprintf(out, "%s\t%d\t%s\t%s", seqname, varslice[0].Position, ref, strings.Join(altslice, ","))
 447                 for i := 0; i < len(varslice); i += 2 {
 448                         v1, v2 := varslice[i], varslice[i+1]
 449                         a1, a2 := alts[v1.New], alts[v2.New]
 450                         if v1.Ref != ref {
 451                                 a1 = 0
 452                         }
 453                         if v2.Ref != ref {
 454                                 a2 = 0
 455                         }
 456                         fmt.Fprintf(out, "\t%d/%d", a1, a2)
 457                 }
 458                 out.Write([]byte{'\n'})
 459         }
 460 }
 461
 462 func printHGVS(out io.Writer, seqname string, varslice []hgvs.Variant) {
 463         for i := 0; i < len(varslice)/2; i++ {
 464                 if i > 0 {
 465                         out.Write([]byte{'\t'})
 466                 }
 467                 var1, var2 := varslice[i*2], varslice[i*2+1]
 468                 if var1 == var2 {
 469                         if var1.Ref == var1.New {
 470                                 out.Write([]byte{'.'})
 471                         } else {
 472                                 fmt.Fprintf(out, "%s:g.%s", seqname, var1.String())
 473                         }
 474                 } else {
 475                         fmt.Fprintf(out, "%s:g.[%s];[%s]", seqname, var1.String(), var2.String())
 476                 }
 477         }
 478         out.Write([]byte{'\n'})
 479 }