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         encoder        *gob.Encoder
  35 }
  36
  37 func (cmd *importer) RunCommand(prog string, args []string, stdin io.Reader, stdout, stderr io.Writer) int {
  38         var err error
  39         defer func() {
  40                 if err != nil {
  41                         fmt.Fprintf(stderr, "%s\n", err)
  42                 }
  43         }()
  44         flags := flag.NewFlagSet("", flag.ContinueOnError)
  45         flags.SetOutput(stderr)
  46         flags.StringVar(&cmd.tagLibraryFile, "tag-library", "", "tag library fasta `file`")
  47         flags.StringVar(&cmd.refFile, "ref", "", "reference fasta `file`")
  48         flags.StringVar(&cmd.outputFile, "o", "-", "output `file`")
  49         flags.StringVar(&cmd.projectUUID, "project", "", "project `UUID` for output data")
  50         flags.BoolVar(&cmd.runLocal, "local", false, "run on local host (default: run in an arvados container)")
  51         priority := flags.Int("priority", 500, "container request priority")
  52         pprof := flags.String("pprof", "", "serve Go profile data at http://`[addr]:port`")
  53         err = flags.Parse(args)
  54         if err == flag.ErrHelp {
  55                 err = nil
  56                 return 0
  57         } else if err != nil {
  58                 return 2
  59         } else if cmd.tagLibraryFile == "" {
  60                 fmt.Fprintln(os.Stderr, "cannot import without -tag-library argument")
  61                 return 2
  62         } else if flags.NArg() == 0 {
  63                 flags.Usage()
  64                 return 2
  65         }
  66
  67         if *pprof != "" {
  68                 go func() {
  69                         log.Println(http.ListenAndServe(*pprof, nil))
  70                 }()
  71         }
  72
  73         if !cmd.runLocal {
  74                 runner := arvadosContainerRunner{
  75                         Name:        "lightning import",
  76                         Client:      arvados.NewClientFromEnv(),
  77                         ProjectUUID: cmd.projectUUID,
  78                         RAM:         30000000000,
  79                         VCPUs:       16,
  80                         Priority:    *priority,
  81                 }
  82                 err = runner.TranslatePaths(&cmd.tagLibraryFile, &cmd.refFile, &cmd.outputFile)
  83                 if err != nil {
  84                         return 1
  85                 }
  86                 inputs := flags.Args()
  87                 for i := range inputs {
  88                         err = runner.TranslatePaths(&inputs[i])
  89                         if err != nil {
  90                                 return 1
  91                         }
  92                 }
  93                 if cmd.outputFile == "-" {
  94                         cmd.outputFile = "/mnt/output/library.gob"
  95                 } else {
  96                         // Not yet implemented, but this should write
  97                         // the collection to an existing collection,
  98                         // possibly even an in-place update.
  99                         err = errors.New("cannot specify output file in container mode: not implemented")
 100                         return 1
 101                 }
 102                 runner.Args = append([]string{"import", "-local=true", "-tag-library", cmd.tagLibraryFile, "-ref", cmd.refFile, "-o", cmd.outputFile}, inputs...)
 103                 var output string
 104                 output, err = runner.Run()
 105                 if err != nil {
 106                         return 1
 107                 }
 108                 fmt.Fprintln(stdout, output+"/library.gob")
 109                 return 0
 110         }
 111
 112         infiles, err := listInputFiles(flags.Args())
 113         if err != nil {
 114                 return 1
 115         }
 116
 117         tilelib, err := cmd.loadTileLibrary()
 118         if err != nil {
 119                 return 1
 120         }
 121         go func() {
 122                 for range time.Tick(10 * time.Minute) {
 123                         log.Printf("tilelib.Len() == %d", tilelib.Len())
 124                 }
 125         }()
 126
 127         var output io.WriteCloser
 128         if cmd.outputFile == "-" {
 129                 output = nopCloser{stdout}
 130         } else {
 131                 output, err = os.OpenFile(cmd.outputFile, os.O_CREATE|os.O_WRONLY, 0777)
 132                 if err != nil {
 133                         return 1
 134                 }
 135                 defer output.Close()
 136         }
 137         bufw := bufio.NewWriter(output)
 138         cmd.encoder = gob.NewEncoder(bufw)
 139
 140         err = cmd.tileInputs(tilelib, infiles)
 141         if err != nil {
 142                 return 1
 143         }
 144         err = bufw.Flush()
 145         if err != nil {
 146                 return 1
 147         }
 148         err = output.Close()
 149         if err != nil {
 150                 return 1
 151         }
 152         return 0
 153 }
 154
 155 func (cmd *importer) tileFasta(tilelib *tileLibrary, infile string) (tileSeq, error) {
 156         var input io.ReadCloser
 157         input, err := os.Open(infile)
 158         if err != nil {
 159                 return nil, err
 160         }
 161         defer input.Close()
 162         if strings.HasSuffix(infile, ".gz") {
 163                 input, err = gzip.NewReader(input)
 164                 if err != nil {
 165                         return nil, err
 166                 }
 167                 defer input.Close()
 168         }
 169         return tilelib.TileFasta(infile, input)
 170 }
 171
 172 func (cmd *importer) loadTileLibrary() (*tileLibrary, error) {
 173         log.Printf("tag library %s load starting", cmd.tagLibraryFile)
 174         f, err := os.Open(cmd.tagLibraryFile)
 175         if err != nil {
 176                 return nil, err
 177         }
 178         defer f.Close()
 179         var rdr io.ReadCloser = f
 180         if strings.HasSuffix(cmd.tagLibraryFile, ".gz") {
 181                 rdr, err = gzip.NewReader(f)
 182                 if err != nil {
 183                         return nil, fmt.Errorf("%s: gzip: %s", cmd.tagLibraryFile, err)
 184                 }
 185                 defer rdr.Close()
 186         }
 187         var taglib tagLibrary
 188         err = taglib.Load(rdr)
 189         if err != nil {
 190                 return nil, err
 191         }
 192         if taglib.Len() < 1 {
 193                 return nil, fmt.Errorf("cannot tile: tag library is empty")
 194         }
 195         log.Printf("tag library %s load done", cmd.tagLibraryFile)
 196         return &tileLibrary{taglib: &taglib}, nil
 197 }
 198
 199 func listInputFiles(paths []string) (files []string, err error) {
 200         for _, path := range paths {
 201                 if fi, err := os.Stat(path); err != nil {
 202                         return nil, fmt.Errorf("%s: stat failed: %s", path, err)
 203                 } else if !fi.IsDir() {
 204                         if !strings.HasSuffix(path, ".2.fasta") || strings.HasSuffix(path, ".2.fasta.gz") {
 205                                 files = append(files, path)
 206                         }
 207                         continue
 208                 }
 209                 d, err := os.Open(path)
 210                 if err != nil {
 211                         return nil, fmt.Errorf("%s: open failed: %s", path, err)
 212                 }
 213                 defer d.Close()
 214                 names, err := d.Readdirnames(0)
 215                 if err != nil {
 216                         return nil, fmt.Errorf("%s: readdir failed: %s", path, err)
 217                 }
 218                 sort.Strings(names)
 219                 for _, name := range names {
 220                         if strings.HasSuffix(name, ".vcf") || strings.HasSuffix(name, ".vcf.gz") {
 221                                 files = append(files, filepath.Join(path, name))
 222                         } else if strings.HasSuffix(name, ".1.fasta") || strings.HasSuffix(name, ".1.fasta.gz") {
 223                                 files = append(files, filepath.Join(path, name))
 224                         }
 225                 }
 226                 d.Close()
 227         }
 228         for _, file := range files {
 229                 if strings.HasSuffix(file, ".1.fasta") || strings.HasSuffix(file, ".1.fasta.gz") {
 230                         continue
 231                 } else if _, err := os.Stat(file + ".csi"); err == nil {
 232                         continue
 233                 } else if _, err = os.Stat(file + ".tbi"); err == nil {
 234                         continue
 235                 } else {
 236                         return nil, fmt.Errorf("%s: cannot read without .tbi or .csi index file", file)
 237                 }
 238         }
 239         return
 240 }
 241
 242 func (cmd *importer) tileInputs(tilelib *tileLibrary, infiles []string) error {
 243         starttime := time.Now()
 244         errs := make(chan error, 1)
 245         todo := make(chan func() error, len(infiles)*2)
 246         var encodeJobs sync.WaitGroup
 247         for _, infile := range infiles {
 248                 infile := infile
 249                 var phases sync.WaitGroup
 250                 phases.Add(2)
 251                 variants := make([][]tileVariantID, 2)
 252                 if strings.HasSuffix(infile, ".1.fasta") || strings.HasSuffix(infile, ".1.fasta.gz") {
 253                         todo <- func() error {
 254                                 defer phases.Done()
 255                                 log.Printf("%s starting", infile)
 256                                 defer log.Printf("%s done", infile)
 257                                 tseqs, err := cmd.tileFasta(tilelib, infile)
 258                                 variants[0] = tseqs.Variants()
 259                                 return err
 260                         }
 261                         infile2 := regexp.MustCompile(`\.1\.fasta(\.gz)?$`).ReplaceAllString(infile, `.2.fasta$1`)
 262                         todo <- func() error {
 263                                 defer phases.Done()
 264                                 log.Printf("%s starting", infile2)
 265                                 defer log.Printf("%s done", infile2)
 266                                 tseqs, err := cmd.tileFasta(tilelib, infile2)
 267                                 variants[1] = tseqs.Variants()
 268                                 return err
 269                         }
 270                 } else {
 271                         for phase := 0; phase < 2; phase++ {
 272                                 phase := phase
 273                                 todo <- func() error {
 274                                         defer phases.Done()
 275                                         log.Printf("%s phase %d starting", infile, phase+1)
 276                                         defer log.Printf("%s phase %d done", infile, phase+1)
 277                                         tseqs, err := cmd.tileGVCF(tilelib, infile, phase)
 278                                         variants[phase] = tseqs.Variants()
 279                                         return err
 280                                 }
 281                         }
 282                 }
 283                 encodeJobs.Add(1)
 284                 go func() {
 285                         defer encodeJobs.Done()
 286                         phases.Wait()
 287                         if len(errs) > 0 {
 288                                 return
 289                         }
 290                         ntags := len(variants[0])
 291                         if ntags < len(variants[1]) {
 292                                 ntags = len(variants[1])
 293                         }
 294                         flat := make([]tileVariantID, ntags*2)
 295                         for i := 0; i < ntags; i++ {
 296                                 flat[i*2] = variants[0][i]
 297                                 flat[i*2+1] = variants[1][i]
 298                         }
 299                         err := cmd.encoder.Encode(LibraryEntry{
 300                                 CompactGenomes: []CompactGenome{{Name: infile, Variants: flat}},
 301                         })
 302                         if err != nil {
 303                                 select {
 304                                 case errs <- err:
 305                                 default:
 306                                 }
 307                         }
 308                 }()
 309         }
 310         go close(todo)
 311         var tileJobs sync.WaitGroup
 312         var running int64
 313         for i := 0; i < runtime.NumCPU()*9/8+1; i++ {
 314                 tileJobs.Add(1)
 315                 atomic.AddInt64(&running, 1)
 316                 go func() {
 317                         defer tileJobs.Done()
 318                         defer atomic.AddInt64(&running, -1)
 319                         for fn := range todo {
 320                                 if len(errs) > 0 {
 321                                         return
 322                                 }
 323                                 err := fn()
 324                                 if err != nil {
 325                                         select {
 326                                         case errs <- err:
 327                                         default:
 328                                         }
 329                                 }
 330                                 remain := len(todo) + int(atomic.LoadInt64(&running)) - 1
 331                                 ttl := time.Now().Sub(starttime) * time.Duration(remain) / time.Duration(cap(todo)-remain)
 332                                 eta := time.Now().Add(ttl)
 333                                 log.Printf("progress %d/%d, eta %v (%v)", cap(todo)-remain, cap(todo), eta, ttl)
 334                         }
 335                 }()
 336         }
 337         tileJobs.Wait()
 338         encodeJobs.Wait()
 339         go close(errs)
 340         return <-errs
 341 }
 342
 343 func (cmd *importer) tileGVCF(tilelib *tileLibrary, infile string, phase int) (tileseq tileSeq, err error) {
 344         if cmd.refFile == "" {
 345                 err = errors.New("cannot import vcf: reference data (-ref) not specified")
 346                 return
 347         }
 348         args := []string{"bcftools", "consensus", "--fasta-ref", cmd.refFile, "-H", fmt.Sprint(phase + 1), infile}
 349         indexsuffix := ".tbi"
 350         if _, err := os.Stat(infile + ".csi"); err == nil {
 351                 indexsuffix = ".csi"
 352         }
 353         if out, err := exec.Command("docker", "image", "ls", "-q", "lightning-runtime").Output(); err == nil && len(out) > 0 {
 354                 args = append([]string{
 355                         "docker", "run", "--rm",
 356                         "--log-driver=none",
 357                         "--volume=" + infile + ":" + infile + ":ro",
 358                         "--volume=" + infile + indexsuffix + ":" + infile + indexsuffix + ":ro",
 359                         "--volume=" + cmd.refFile + ":" + cmd.refFile + ":ro",
 360                         "lightning-runtime",
 361                 }, args...)
 362         }
 363         consensus := exec.Command(args[0], args[1:]...)
 364         consensus.Stderr = os.Stderr
 365         stdout, err := consensus.StdoutPipe()
 366         defer stdout.Close()
 367         if err != nil {
 368                 return
 369         }
 370         err = consensus.Start()
 371         if err != nil {
 372                 return
 373         }
 374         defer consensus.Wait()
 375         tileseq, err = tilelib.TileFasta(fmt.Sprintf("%s phase %d", infile, phase+1), stdout)
 376         if err != nil {
 377                 return
 378         }
 379         err = stdout.Close()
 380         if err != nil {
 381                 return
 382         }
 383         err = consensus.Wait()
 384         if err != nil {
 385                 err = fmt.Errorf("%s phase %d: bcftools: %s", infile, phase, err)
 386                 return
 387         }
 388         return
 389 }