import.go

   1 package main
   2
   3 import (
   4         "bufio"
   5         "compress/gzip"
   6         "encoding/gob"
   7         "errors"
   8         "flag"
   9         "fmt"
  10         "io"
  11         "net/http"
  12         _ "net/http/pprof"
  13         "os"
  14         "os/exec"
  15         "path/filepath"
  16         "regexp"
  17         "runtime"
  18         "sort"
  19         "strings"
  20         "sync"
  21         "sync/atomic"
  22         "time"
  23
  24         "git.arvados.org/arvados.git/sdk/go/arvados"
  25         log "github.com/sirupsen/logrus"
  26 )
  27
  28 type importer struct {
  29         tagLibraryFile string
  30         refFile        string
  31         outputFile     string
  32         projectUUID    string
  33         runLocal       bool
  34         skipOOO        bool
  35         outputTiles    bool
  36         includeNoCalls bool
  37         encoder        *gob.Encoder
  38 }
  39
  40 func (cmd *importer) RunCommand(prog string, args []string, stdin io.Reader, stdout, stderr io.Writer) int {
  41         var err error
  42         defer func() {
  43                 if err != nil {
  44                         fmt.Fprintf(stderr, "%s\n", err)
  45                 }
  46         }()
  47         flags := flag.NewFlagSet("", flag.ContinueOnError)
  48         flags.SetOutput(stderr)
  49         flags.StringVar(&cmd.tagLibraryFile, "tag-library", "", "tag library fasta `file`")
  50         flags.StringVar(&cmd.refFile, "ref", "", "reference fasta `file`")
  51         flags.StringVar(&cmd.outputFile, "o", "-", "output `file`")
  52         flags.StringVar(&cmd.projectUUID, "project", "", "project `UUID` for output data")
  53         flags.BoolVar(&cmd.runLocal, "local", false, "run on local host (default: run in an arvados container)")
  54         flags.BoolVar(&cmd.skipOOO, "skip-ooo", false, "skip out-of-order tags")
  55         flags.BoolVar(&cmd.outputTiles, "output-tiles", false, "include tile variant sequences in output file")
  56         flags.BoolVar(&cmd.includeNoCalls, "include-no-calls", false, "treat tiles with no-calls as regular tiles")
  57         priority := flags.Int("priority", 500, "container request priority")
  58         pprof := flags.String("pprof", "", "serve Go profile data at http://`[addr]:port`")
  59         loglevel := flags.String("loglevel", "info", "logging threshold (trace, debug, info, warn, error, fatal, or panic)")
  60         err = flags.Parse(args)
  61         if err == flag.ErrHelp {
  62                 err = nil
  63                 return 0
  64         } else if err != nil {
  65                 return 2
  66         } else if cmd.tagLibraryFile == "" {
  67                 fmt.Fprintln(os.Stderr, "cannot import without -tag-library argument")
  68                 return 2
  69         } else if flags.NArg() == 0 {
  70                 flags.Usage()
  71                 return 2
  72         }
  73
  74         if *pprof != "" {
  75                 go func() {
  76                         log.Println(http.ListenAndServe(*pprof, nil))
  77                 }()
  78         }
  79
  80         lvl, err := log.ParseLevel(*loglevel)
  81         if err != nil {
  82                 return 2
  83         }
  84         log.SetLevel(lvl)
  85
  86         if !cmd.runLocal {
  87                 runner := arvadosContainerRunner{
  88                         Name:        "lightning import",
  89                         Client:      arvados.NewClientFromEnv(),
  90                         ProjectUUID: cmd.projectUUID,
  91                         RAM:         60000000000,
  92                         VCPUs:       16,
  93                         Priority:    *priority,
  94                 }
  95                 err = runner.TranslatePaths(&cmd.tagLibraryFile, &cmd.refFile, &cmd.outputFile)
  96                 if err != nil {
  97                         return 1
  98                 }
  99                 inputs := flags.Args()
 100                 for i := range inputs {
 101                         err = runner.TranslatePaths(&inputs[i])
 102                         if err != nil {
 103                                 return 1
 104                         }
 105                 }
 106                 if cmd.outputFile == "-" {
 107                         cmd.outputFile = "/mnt/output/library.gob"
 108                 } else {
 109                         // Not yet implemented, but this should write
 110                         // the collection to an existing collection,
 111                         // possibly even an in-place update.
 112                         err = errors.New("cannot specify output file in container mode: not implemented")
 113                         return 1
 114                 }
 115                 runner.Args = append([]string{"import",
 116                         "-local=true",
 117                         "-loglevel=" + *loglevel,
 118                         fmt.Sprintf("-skip-ooo=%v", cmd.skipOOO),
 119                         fmt.Sprintf("-output-tiles=%v", cmd.outputTiles),
 120                         fmt.Sprintf("-include-no-calls=%v", cmd.includeNoCalls),
 121                         "-tag-library", cmd.tagLibraryFile,
 122                         "-ref", cmd.refFile,
 123                         "-o", cmd.outputFile,
 124                 }, inputs...)
 125                 var output string
 126                 output, err = runner.Run()
 127                 if err != nil {
 128                         return 1
 129                 }
 130                 fmt.Fprintln(stdout, output+"/library.gob")
 131                 return 0
 132         }
 133
 134         infiles, err := listInputFiles(flags.Args())
 135         if err != nil {
 136                 return 1
 137         }
 138
 139         taglib, err := cmd.loadTagLibrary()
 140         if err != nil {
 141                 return 1
 142         }
 143
 144         var output io.WriteCloser
 145         if cmd.outputFile == "-" {
 146                 output = nopCloser{stdout}
 147         } else {
 148                 output, err = os.OpenFile(cmd.outputFile, os.O_CREATE|os.O_WRONLY, 0777)
 149                 if err != nil {
 150                         return 1
 151                 }
 152                 defer output.Close()
 153         }
 154         bufw := bufio.NewWriter(output)
 155         cmd.encoder = gob.NewEncoder(bufw)
 156
 157         tilelib := &tileLibrary{taglib: taglib, includeNoCalls: cmd.includeNoCalls, skipOOO: cmd.skipOOO}
 158         if cmd.outputTiles {
 159                 tilelib.encoder = cmd.encoder
 160         }
 161         go func() {
 162                 for range time.Tick(10 * time.Minute) {
 163                         log.Printf("tilelib.Len() == %d", tilelib.Len())
 164                 }
 165         }()
 166
 167         err = cmd.tileInputs(tilelib, infiles)
 168         if err != nil {
 169                 return 1
 170         }
 171         err = bufw.Flush()
 172         if err != nil {
 173                 return 1
 174         }
 175         err = output.Close()
 176         if err != nil {
 177                 return 1
 178         }
 179         return 0
 180 }
 181
 182 func (cmd *importer) tileFasta(tilelib *tileLibrary, infile string) (tileSeq, error) {
 183         var input io.ReadCloser
 184         input, err := os.Open(infile)
 185         if err != nil {
 186                 return nil, err
 187         }
 188         defer input.Close()
 189         if strings.HasSuffix(infile, ".gz") {
 190                 input, err = gzip.NewReader(input)
 191                 if err != nil {
 192                         return nil, err
 193                 }
 194                 defer input.Close()
 195         }
 196         return tilelib.TileFasta(infile, input)
 197 }
 198
 199 func (cmd *importer) loadTagLibrary() (*tagLibrary, error) {
 200         log.Printf("tag library %s load starting", cmd.tagLibraryFile)
 201         f, err := os.Open(cmd.tagLibraryFile)
 202         if err != nil {
 203                 return nil, err
 204         }
 205         defer f.Close()
 206         var rdr io.ReadCloser = f
 207         if strings.HasSuffix(cmd.tagLibraryFile, ".gz") {
 208                 rdr, err = gzip.NewReader(f)
 209                 if err != nil {
 210                         return nil, fmt.Errorf("%s: gzip: %s", cmd.tagLibraryFile, err)
 211                 }
 212                 defer rdr.Close()
 213         }
 214         var taglib tagLibrary
 215         err = taglib.Load(rdr)
 216         if err != nil {
 217                 return nil, err
 218         }
 219         if taglib.Len() < 1 {
 220                 return nil, fmt.Errorf("cannot tile: tag library is empty")
 221         }
 222         log.Printf("tag library %s load done", cmd.tagLibraryFile)
 223         return &taglib, nil
 224 }
 225
 226 func listInputFiles(paths []string) (files []string, err error) {
 227         for _, path := range paths {
 228                 if fi, err := os.Stat(path); err != nil {
 229                         return nil, fmt.Errorf("%s: stat failed: %s", path, err)
 230                 } else if !fi.IsDir() {
 231                         if !strings.HasSuffix(path, ".2.fasta") || strings.HasSuffix(path, ".2.fasta.gz") {
 232                                 files = append(files, path)
 233                         }
 234                         continue
 235                 }
 236                 d, err := os.Open(path)
 237                 if err != nil {
 238                         return nil, fmt.Errorf("%s: open failed: %s", path, err)
 239                 }
 240                 defer d.Close()
 241                 names, err := d.Readdirnames(0)
 242                 if err != nil {
 243                         return nil, fmt.Errorf("%s: readdir failed: %s", path, err)
 244                 }
 245                 sort.Strings(names)
 246                 for _, name := range names {
 247                         if strings.HasSuffix(name, ".vcf") || strings.HasSuffix(name, ".vcf.gz") {
 248                                 files = append(files, filepath.Join(path, name))
 249                         } else if strings.HasSuffix(name, ".1.fasta") || strings.HasSuffix(name, ".1.fasta.gz") {
 250                                 files = append(files, filepath.Join(path, name))
 251                         }
 252                 }
 253                 d.Close()
 254         }
 255         for _, file := range files {
 256                 if strings.HasSuffix(file, ".1.fasta") || strings.HasSuffix(file, ".1.fasta.gz") {
 257                         continue
 258                 } else if _, err := os.Stat(file + ".csi"); err == nil {
 259                         continue
 260                 } else if _, err = os.Stat(file + ".tbi"); err == nil {
 261                         continue
 262                 } else {
 263                         return nil, fmt.Errorf("%s: cannot read without .tbi or .csi index file", file)
 264                 }
 265         }
 266         return
 267 }
 268
 269 func (cmd *importer) tileInputs(tilelib *tileLibrary, infiles []string) error {
 270         starttime := time.Now()
 271         errs := make(chan error, 1)
 272         todo := make(chan func() error, len(infiles)*2)
 273         var encodeJobs sync.WaitGroup
 274         for _, infile := range infiles {
 275                 infile := infile
 276                 var phases sync.WaitGroup
 277                 phases.Add(2)
 278                 variants := make([][]tileVariantID, 2)
 279                 if strings.HasSuffix(infile, ".1.fasta") || strings.HasSuffix(infile, ".1.fasta.gz") {
 280                         todo <- func() error {
 281                                 defer phases.Done()
 282                                 log.Printf("%s starting", infile)
 283                                 defer log.Printf("%s done", infile)
 284                                 tseqs, err := cmd.tileFasta(tilelib, infile)
 285                                 var kept, dropped int
 286                                 variants[0], kept, dropped = tseqs.Variants()
 287                                 log.Printf("%s found %d unique tags plus %d repeats", infile, kept, dropped)
 288                                 return err
 289                         }
 290                         infile2 := regexp.MustCompile(`\.1\.fasta(\.gz)?$`).ReplaceAllString(infile, `.2.fasta$1`)
 291                         todo <- func() error {
 292                                 defer phases.Done()
 293                                 log.Printf("%s starting", infile2)
 294                                 defer log.Printf("%s done", infile2)
 295                                 tseqs, err := cmd.tileFasta(tilelib, infile2)
 296                                 var kept, dropped int
 297                                 variants[1], kept, dropped = tseqs.Variants()
 298                                 log.Printf("%s found %d unique tags plus %d repeats", infile, kept, dropped)
 299                                 return err
 300                         }
 301                 } else {
 302                         for phase := 0; phase < 2; phase++ {
 303                                 phase := phase
 304                                 todo <- func() error {
 305                                         defer phases.Done()
 306                                         log.Printf("%s phase %d starting", infile, phase+1)
 307                                         defer log.Printf("%s phase %d done", infile, phase+1)
 308                                         tseqs, err := cmd.tileGVCF(tilelib, infile, phase)
 309                                         var kept, dropped int
 310                                         variants[phase], kept, dropped = tseqs.Variants()
 311                                         log.Printf("%s phase %d found %d unique tags plus %d repeats", infile, phase+1, kept, dropped)
 312                                         return err
 313                                 }
 314                         }
 315                 }
 316                 encodeJobs.Add(1)
 317                 go func() {
 318                         defer encodeJobs.Done()
 319                         phases.Wait()
 320                         if len(errs) > 0 {
 321                                 return
 322                         }
 323                         ntags := len(variants[0])
 324                         if ntags < len(variants[1]) {
 325                                 ntags = len(variants[1])
 326                         }
 327                         flat := make([]tileVariantID, ntags*2)
 328                         for i := 0; i < ntags; i++ {
 329                                 for hap := 0; hap < 2; hap++ {
 330                                         if i < len(variants[hap]) {
 331                                                 flat[i*2+hap] = variants[hap][i]
 332                                         }
 333                                 }
 334                         }
 335                         err := cmd.encoder.Encode(LibraryEntry{
 336                                 CompactGenomes: []CompactGenome{{Name: infile, Variants: flat}},
 337                         })
 338                         if err != nil {
 339                                 select {
 340                                 case errs <- err:
 341                                 default:
 342                                 }
 343                         }
 344                 }()
 345         }
 346         go close(todo)
 347         var tileJobs sync.WaitGroup
 348         var running int64
 349         for i := 0; i < runtime.NumCPU()*9/8+1; i++ {
 350                 tileJobs.Add(1)
 351                 atomic.AddInt64(&running, 1)
 352                 go func() {
 353                         defer tileJobs.Done()
 354                         defer atomic.AddInt64(&running, -1)
 355                         for fn := range todo {
 356                                 if len(errs) > 0 {
 357                                         return
 358                                 }
 359                                 err := fn()
 360                                 if err != nil {
 361                                         select {
 362                                         case errs <- err:
 363                                         default:
 364                                         }
 365                                 }
 366                                 remain := len(todo) + int(atomic.LoadInt64(&running)) - 1
 367                                 ttl := time.Now().Sub(starttime) * time.Duration(remain) / time.Duration(cap(todo)-remain)
 368                                 eta := time.Now().Add(ttl)
 369                                 log.Printf("progress %d/%d, eta %v (%v)", cap(todo)-remain, cap(todo), eta, ttl)
 370                         }
 371                 }()
 372         }
 373         tileJobs.Wait()
 374         encodeJobs.Wait()
 375         go close(errs)
 376         return <-errs
 377 }
 378
 379 func (cmd *importer) tileGVCF(tilelib *tileLibrary, infile string, phase int) (tileseq tileSeq, err error) {
 380         if cmd.refFile == "" {
 381                 err = errors.New("cannot import vcf: reference data (-ref) not specified")
 382                 return
 383         }
 384         args := []string{"bcftools", "consensus", "--fasta-ref", cmd.refFile, "-H", fmt.Sprint(phase + 1), infile}
 385         indexsuffix := ".tbi"
 386         if _, err := os.Stat(infile + ".csi"); err == nil {
 387                 indexsuffix = ".csi"
 388         }
 389         if out, err := exec.Command("docker", "image", "ls", "-q", "lightning-runtime").Output(); err == nil && len(out) > 0 {
 390                 args = append([]string{
 391                         "docker", "run", "--rm",
 392                         "--log-driver=none",
 393                         "--volume=" + infile + ":" + infile + ":ro",
 394                         "--volume=" + infile + indexsuffix + ":" + infile + indexsuffix + ":ro",
 395                         "--volume=" + cmd.refFile + ":" + cmd.refFile + ":ro",
 396                         "lightning-runtime",
 397                 }, args...)
 398         }
 399         consensus := exec.Command(args[0], args[1:]...)
 400         consensus.Stderr = os.Stderr
 401         stdout, err := consensus.StdoutPipe()
 402         defer stdout.Close()
 403         if err != nil {
 404                 return
 405         }
 406         err = consensus.Start()
 407         if err != nil {
 408                 return
 409         }
 410         defer consensus.Wait()
 411         tileseq, err = tilelib.TileFasta(fmt.Sprintf("%s phase %d", infile, phase+1), stdout)
 412         if err != nil {
 413                 return
 414         }
 415         err = stdout.Close()
 416         if err != nil {
 417                 return
 418         }
 419         err = consensus.Wait()
 420         if err != nil {
 421                 err = fmt.Errorf("%s phase %d: bcftools: %s", infile, phase, err)
 422                 return
 423         }
 424         return
 425 }