import.go

   1 package main
   2
   3 import (
   4         "bufio"
   5         "compress/gzip"
   6         "encoding/gob"
   7         "flag"
   8         "fmt"
   9         "io"
  10         "log"
  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
  25 type importer struct {
  26         tagLibraryFile string
  27         refFile        string
  28         encoder        *gob.Encoder
  29 }
  30
  31 func (cmd *importer) RunCommand(prog string, args []string, stdin io.Reader, stdout, stderr io.Writer) int {
  32         var err error
  33         defer func() {
  34                 if err != nil {
  35                         fmt.Fprintf(stderr, "%s\n", err)
  36                 }
  37         }()
  38         flags := flag.NewFlagSet("", flag.ContinueOnError)
  39         flags.SetOutput(stderr)
  40         flags.StringVar(&cmd.tagLibraryFile, "tag-library", "", "tag library fasta `file`")
  41         flags.StringVar(&cmd.refFile, "ref", "", "reference fasta `file`")
  42         pprof := flags.String("pprof", "", "serve Go profile data at http://`[addr]:port`")
  43         err = flags.Parse(args)
  44         if err == flag.ErrHelp {
  45                 err = nil
  46                 return 0
  47         } else if err != nil {
  48                 return 2
  49         } else if cmd.refFile == "" || cmd.tagLibraryFile == "" {
  50                 fmt.Fprintln(os.Stderr, "cannot run without -tag-library and -ref arguments")
  51                 return 2
  52         } else if flags.NArg() == 0 {
  53                 flags.Usage()
  54                 return 2
  55         }
  56
  57         if *pprof != "" {
  58                 go func() {
  59                         log.Println(http.ListenAndServe(*pprof, nil))
  60                 }()
  61         }
  62
  63         infiles, err := listInputFiles(flags.Args())
  64         if err != nil {
  65                 return 1
  66         }
  67
  68         tilelib, err := cmd.loadTileLibrary()
  69         if err != nil {
  70                 return 1
  71         }
  72         go func() {
  73                 for range time.Tick(10 * time.Second) {
  74                         log.Printf("tilelib.Len() == %d", tilelib.Len())
  75                 }
  76         }()
  77         w := bufio.NewWriter(stdout)
  78         cmd.encoder = gob.NewEncoder(w)
  79         err = cmd.tileGVCFs(tilelib, infiles)
  80         if err != nil {
  81                 return 1
  82         }
  83         err = w.Flush()
  84         if err != nil {
  85                 return 1
  86         }
  87         return 0
  88 }
  89
  90 func (cmd *importer) tileFasta(tilelib *tileLibrary, infile string) (tileSeq, error) {
  91         var input io.ReadCloser
  92         input, err := os.Open(infile)
  93         if err != nil {
  94                 return nil, err
  95         }
  96         defer input.Close()
  97         if strings.HasSuffix(infile, ".gz") {
  98                 input, err = gzip.NewReader(input)
  99                 if err != nil {
 100                         return nil, err
 101                 }
 102                 defer input.Close()
 103         }
 104         return tilelib.TileFasta(infile, input)
 105 }
 106
 107 func (cmd *importer) loadTileLibrary() (*tileLibrary, error) {
 108         log.Printf("tag library %s load starting", cmd.tagLibraryFile)
 109         f, err := os.Open(cmd.tagLibraryFile)
 110         if err != nil {
 111                 return nil, err
 112         }
 113         defer f.Close()
 114         var rdr io.ReadCloser = f
 115         if strings.HasSuffix(cmd.tagLibraryFile, ".gz") {
 116                 rdr, err = gzip.NewReader(f)
 117                 if err != nil {
 118                         return nil, fmt.Errorf("%s: gzip: %s", cmd.tagLibraryFile, err)
 119                 }
 120                 defer rdr.Close()
 121         }
 122         var taglib tagLibrary
 123         err = taglib.Load(rdr)
 124         if err != nil {
 125                 return nil, err
 126         }
 127         if taglib.Len() < 1 {
 128                 return nil, fmt.Errorf("cannot tile: tag library is empty")
 129         }
 130         log.Printf("tag library %s load done", cmd.tagLibraryFile)
 131         return &tileLibrary{taglib: &taglib}, nil
 132 }
 133
 134 func listInputFiles(paths []string) (files []string, err error) {
 135         for _, path := range paths {
 136                 if fi, err := os.Stat(path); err != nil {
 137                         return nil, fmt.Errorf("%s: stat failed: %s", path, err)
 138                 } else if !fi.IsDir() {
 139                         if !strings.HasSuffix(path, ".2.fasta") || strings.HasSuffix(path, ".2.fasta.gz") {
 140                                 files = append(files, path)
 141                         }
 142                         continue
 143                 }
 144                 d, err := os.Open(path)
 145                 if err != nil {
 146                         return nil, fmt.Errorf("%s: open failed: %s", path, err)
 147                 }
 148                 defer d.Close()
 149                 names, err := d.Readdirnames(0)
 150                 if err != nil {
 151                         return nil, fmt.Errorf("%s: readdir failed: %s", path, err)
 152                 }
 153                 sort.Strings(names)
 154                 for _, name := range names {
 155                         if strings.HasSuffix(name, ".vcf") || strings.HasSuffix(name, ".vcf.gz") {
 156                                 files = append(files, filepath.Join(path, name))
 157                         } else if strings.HasSuffix(name, ".1.fasta") || strings.HasSuffix(name, ".1.fasta.gz") {
 158                                 files = append(files, filepath.Join(path, name))
 159                         }
 160                 }
 161                 d.Close()
 162         }
 163         for _, file := range files {
 164                 if strings.HasSuffix(file, ".1.fasta") || strings.HasSuffix(file, ".1.fasta.gz") {
 165                         continue
 166                 } else if _, err := os.Stat(file + ".csi"); err == nil {
 167                         continue
 168                 } else if _, err = os.Stat(file + ".tbi"); err == nil {
 169                         continue
 170                 } else {
 171                         return nil, fmt.Errorf("%s: cannot read without .tbi or .csi index file", file)
 172                 }
 173         }
 174         return
 175 }
 176
 177 func (cmd *importer) tileGVCFs(tilelib *tileLibrary, infiles []string) error {
 178         starttime := time.Now()
 179         errs := make(chan error, 1)
 180         todo := make(chan func() error, len(infiles)*2)
 181         var encodeJobs sync.WaitGroup
 182         for _, infile := range infiles {
 183                 infile := infile
 184                 var phases sync.WaitGroup
 185                 phases.Add(2)
 186                 variants := make([][]tileVariantID, 2)
 187                 if strings.HasSuffix(infile, ".1.fasta") || strings.HasSuffix(infile, ".1.fasta.gz") {
 188                         todo <- func() error {
 189                                 defer phases.Done()
 190                                 log.Printf("%s starting", infile)
 191                                 defer log.Printf("%s done", infile)
 192                                 tseqs, err := cmd.tileFasta(tilelib, infile)
 193                                 variants[0] = tseqs.Variants()
 194                                 return err
 195                         }
 196                         infile2 := regexp.MustCompile(`\.1\.fasta(\.gz)?$`).ReplaceAllString(infile, `.2.fasta$1`)
 197                         todo <- func() error {
 198                                 defer phases.Done()
 199                                 log.Printf("%s starting", infile2)
 200                                 defer log.Printf("%s done", infile2)
 201                                 tseqs, err := cmd.tileFasta(tilelib, infile2)
 202                                 variants[1] = tseqs.Variants()
 203                                 return err
 204                         }
 205                 } else {
 206                         for phase := 0; phase < 2; phase++ {
 207                                 phase := phase
 208                                 todo <- func() error {
 209                                         defer phases.Done()
 210                                         log.Printf("%s phase %d starting", infile, phase+1)
 211                                         defer log.Printf("%s phase %d done", infile, phase+1)
 212                                         tseqs, err := cmd.tileGVCF(tilelib, infile, phase)
 213                                         variants[phase] = tseqs.Variants()
 214                                         return err
 215                                 }
 216                         }
 217                 }
 218                 encodeJobs.Add(1)
 219                 go func() {
 220                         defer encodeJobs.Done()
 221                         phases.Wait()
 222                         if len(errs) > 0 {
 223                                 return
 224                         }
 225                         ntags := len(variants[0])
 226                         if ntags < len(variants[1]) {
 227                                 ntags = len(variants[1])
 228                         }
 229                         flat := make([]tileVariantID, ntags*2)
 230                         for i := 0; i < ntags; i++ {
 231                                 flat[i*2] = variants[0][i]
 232                                 flat[i*2+1] = variants[1][i]
 233                         }
 234                         err := cmd.encoder.Encode(LibraryEntry{
 235                                 CompactGenomes: []CompactGenome{{Name: infile, Variants: flat}},
 236                         })
 237                         if err != nil {
 238                                 select {
 239                                 case errs <- err:
 240                                 default:
 241                                 }
 242                         }
 243                 }()
 244         }
 245         go close(todo)
 246         var tileJobs sync.WaitGroup
 247         running := int64(runtime.NumCPU())
 248         for i := 0; i < runtime.NumCPU(); i++ {
 249                 tileJobs.Add(1)
 250                 go func() {
 251                         defer tileJobs.Done()
 252                         defer atomic.AddInt64(&running, -1)
 253                         for fn := range todo {
 254                                 if len(errs) > 0 {
 255                                         return
 256                                 }
 257                                 err := fn()
 258                                 if err != nil {
 259                                         select {
 260                                         case errs <- err:
 261                                         default:
 262                                         }
 263                                 }
 264                                 remain := len(todo) + int(atomic.LoadInt64(&running)) - 1
 265                                 ttl := time.Now().Sub(starttime) * time.Duration(remain) / time.Duration(cap(todo)-remain)
 266                                 eta := time.Now().Add(ttl)
 267                                 log.Printf("progress %d/%d, eta %v (%v)", cap(todo)-remain, cap(todo), eta, ttl)
 268                         }
 269                 }()
 270         }
 271         tileJobs.Wait()
 272         encodeJobs.Wait()
 273         go close(errs)
 274         return <-errs
 275 }
 276
 277 func (cmd *importer) tileGVCF(tilelib *tileLibrary, infile string, phase int) (tileseq tileSeq, err error) {
 278         args := []string{"bcftools", "consensus", "--fasta-ref", cmd.refFile, "-H", fmt.Sprint(phase + 1), infile}
 279         indexsuffix := ".tbi"
 280         if _, err := os.Stat(infile + ".csi"); err == nil {
 281                 indexsuffix = ".csi"
 282         }
 283         if out, err := exec.Command("docker", "image", "ls", "-q", "lightning-runtime").Output(); err == nil && len(out) > 0 {
 284                 args = append([]string{
 285                         "docker", "run", "--rm",
 286                         "--log-driver=none",
 287                         "--volume=" + infile + ":" + infile + ":ro",
 288                         "--volume=" + infile + indexsuffix + ":" + infile + indexsuffix + ":ro",
 289                         "--volume=" + cmd.refFile + ":" + cmd.refFile + ":ro",
 290                         "lightning-runtime",
 291                 }, args...)
 292         }
 293         consensus := exec.Command(args[0], args[1:]...)
 294         consensus.Stderr = os.Stderr
 295         stdout, err := consensus.StdoutPipe()
 296         defer stdout.Close()
 297         if err != nil {
 298                 return
 299         }
 300         err = consensus.Start()
 301         if err != nil {
 302                 return
 303         }
 304         defer consensus.Wait()
 305         tileseq, err = tilelib.TileFasta(fmt.Sprintf("%s phase %d", infile, phase+1), stdout)
 306         if err != nil {
 307                 return
 308         }
 309         err = stdout.Close()
 310         if err != nil {
 311                 return
 312         }
 313         err = consensus.Wait()
 314         if err != nil {
 315                 err = fmt.Errorf("%s phase %d: bcftools: %s", infile, phase, err)
 316                 return
 317         }
 318         return
 319 }