annotate.go

   1 package lightning
   2
   3 import (
   4         "bufio"
   5         "context"
   6         "errors"
   7         "flag"
   8         "fmt"
   9         "io"
  10         "io/ioutil"
  11         "net/http"
  12         _ "net/http/pprof"
  13         "os"
  14         "runtime"
  15         "sort"
  16         "strconv"
  17         "strings"
  18         "sync"
  19
  20         "git.arvados.org/arvados.git/sdk/go/arvados"
  21         "github.com/arvados/lightning/hgvs"
  22         log "github.com/sirupsen/logrus"
  23 )
  24
  25 type annotatecmd struct {
  26         dropTiles        []bool
  27         variantHash      bool
  28         maxTileSize      int
  29         tag2tagid        map[string]tagID
  30         reportAnnotation func(tag tagID, outcol int, variant tileVariantID, refname string, seqname string, pdi hgvs.Variant)
  31 }
  32
  33 func (cmd *annotatecmd) RunCommand(prog string, args []string, stdin io.Reader, stdout, stderr io.Writer) int {
  34         var err error
  35         defer func() {
  36                 if err != nil {
  37                         fmt.Fprintf(stderr, "%s\n", err)
  38                 }
  39         }()
  40         flags := flag.NewFlagSet("", flag.ContinueOnError)
  41         flags.SetOutput(stderr)
  42         pprof := flags.String("pprof", "", "serve Go profile data at http://`[addr]:port`")
  43         runlocal := flags.Bool("local", false, "run on local host (default: run in an arvados container)")
  44         projectUUID := flags.String("project", "", "project `UUID` for output data")
  45         priority := flags.Int("priority", 500, "container request priority")
  46         inputFilename := flags.String("i", "-", "input `file` (library)")
  47         outputFilename := flags.String("o", "-", "output `file`")
  48         flags.BoolVar(&cmd.variantHash, "variant-hash", false, "output variant hash instead of index")
  49         flags.IntVar(&cmd.maxTileSize, "max-tile-size", 50000, "don't try to make annotations for tiles bigger than given `size`")
  50         err = flags.Parse(args)
  51         if err == flag.ErrHelp {
  52                 err = nil
  53                 return 0
  54         } else if err != nil {
  55                 return 2
  56         }
  57
  58         if *pprof != "" {
  59                 go func() {
  60                         log.Println(http.ListenAndServe(*pprof, nil))
  61                 }()
  62         }
  63         if !*runlocal {
  64                 if *outputFilename != "-" {
  65                         err = errors.New("cannot specify output file in container mode: not implemented")
  66                         return 1
  67                 }
  68                 runner := arvadosContainerRunner{
  69                         Name:        "lightning annotate",
  70                         Client:      arvados.NewClientFromEnv(),
  71                         ProjectUUID: *projectUUID,
  72                         RAM:         80000000000,
  73                         VCPUs:       16,
  74                         Priority:    *priority,
  75                 }
  76                 err = runner.TranslatePaths(inputFilename)
  77                 if err != nil {
  78                         return 1
  79                 }
  80                 runner.Args = []string{"annotate", "-local=true", fmt.Sprintf("-variant-hash=%v", cmd.variantHash), "-max-tile-size", strconv.Itoa(cmd.maxTileSize), "-i", *inputFilename, "-o", "/mnt/output/tilevariants.csv"}
  81                 var output string
  82                 output, err = runner.Run()
  83                 if err != nil {
  84                         return 1
  85                 }
  86                 fmt.Fprintln(stdout, output+"/tilevariants.csv")
  87                 return 0
  88         }
  89
  90         var input io.ReadCloser
  91         if *inputFilename == "-" {
  92                 input = ioutil.NopCloser(stdin)
  93         } else {
  94                 input, err = os.Open(*inputFilename)
  95                 if err != nil {
  96                         return 1
  97                 }
  98                 defer input.Close()
  99         }
 100
 101         var output io.WriteCloser
 102         if *outputFilename == "-" {
 103                 output = nopCloser{stdout}
 104         } else {
 105                 output, err = os.OpenFile(*outputFilename, os.O_CREATE|os.O_WRONLY, 0666)
 106                 if err != nil {
 107                         return 1
 108                 }
 109                 defer output.Close()
 110         }
 111         bufw := bufio.NewWriterSize(output, 4*1024*1024)
 112
 113         tilelib := &tileLibrary{
 114                 retainNoCalls:       true,
 115                 retainTileSequences: true,
 116         }
 117         err = tilelib.LoadGob(context.Background(), input, strings.HasSuffix(*inputFilename, ".gz"), nil)
 118         if err != nil {
 119                 return 1
 120         }
 121         err = cmd.exportTileDiffs(bufw, tilelib)
 122         if err != nil {
 123                 return 1
 124         }
 125         err = bufw.Flush()
 126         if err != nil {
 127                 return 1
 128         }
 129         err = output.Close()
 130         if err != nil {
 131                 return 1
 132         }
 133         err = input.Close()
 134         if err != nil {
 135                 return 1
 136         }
 137         return 0
 138 }
 139
 140 func (cmd *annotatecmd) exportTileDiffs(outw io.Writer, tilelib *tileLibrary) error {
 141         tagset := tilelib.taglib.Tags()
 142         if len(tagset) == 0 {
 143                 return errors.New("cannot annotate library without tags")
 144         }
 145         taglen := len(tagset[0])
 146         var refs []string
 147         for name := range tilelib.refseqs {
 148                 refs = append(refs, name)
 149         }
 150         cmd.tag2tagid = make(map[string]tagID, len(tagset))
 151         for tagid, tagseq := range tagset {
 152                 cmd.tag2tagid[string(tagseq)] = tagID(tagid)
 153         }
 154         sort.Strings(refs)
 155         log.Infof("len(refs) %d", len(refs))
 156
 157         outch := make(chan string, runtime.NumCPU()*2)
 158         var outwg sync.WaitGroup
 159         defer outwg.Wait()
 160         outwg.Add(1)
 161         go func() {
 162                 defer outwg.Done()
 163                 for s := range outch {
 164                         io.WriteString(outw, s)
 165                 }
 166         }()
 167         defer close(outch)
 168
 169         nseqs := 0
 170         for _, refcs := range tilelib.refseqs {
 171                 nseqs += len(refcs)
 172         }
 173
 174         throttle := &throttle{Max: runtime.NumCPU()*2 + nseqs*2 + 1}
 175         defer throttle.Wait()
 176
 177         for _, refname := range refs {
 178                 refname := refname
 179                 refcs := tilelib.refseqs[refname]
 180                 var seqnames []string
 181                 for seqname := range refcs {
 182                         seqnames = append(seqnames, seqname)
 183                 }
 184                 sort.Strings(seqnames)
 185                 for _, seqname := range seqnames {
 186                         seqname := seqname
 187                         throttle.Acquire()
 188                         if throttle.Err() != nil {
 189                                 break
 190                         }
 191                         go func() {
 192                                 defer throttle.Release()
 193                                 throttle.Report(cmd.annotateSequence(throttle, outch, tilelib, taglen, refname, seqname, refcs[seqname], len(refs) > 1))
 194                         }()
 195                 }
 196         }
 197         throttle.Wait()
 198         return throttle.Err()
 199 }
 200
 201 func (cmd *annotatecmd) annotateSequence(throttle *throttle, outch chan<- string, tilelib *tileLibrary, taglen int, refname, seqname string, reftiles []tileLibRef, refnamecol bool) error {
 202         refnamefield := ""
 203         if refnamecol {
 204                 refnamefield = "," + trimFilenameForLabel(refname)
 205         }
 206         var refseq []byte
 207         // tilestart[123] is the index into refseq
 208         // where the tile for tag 123 was placed.
 209         tilestart := map[tagID]int{}
 210         tileend := map[tagID]int{}
 211         for _, libref := range reftiles {
 212                 if libref.Variant < 1 {
 213                         return fmt.Errorf("reference %q seq %q uses variant zero at tag %d", refname, seqname, libref.Tag)
 214                 }
 215                 seq := tilelib.TileVariantSequence(libref)
 216                 if len(seq) < taglen {
 217                         return fmt.Errorf("reference %q seq %q uses tile %d variant %d with sequence len %d < taglen %d", refname, seqname, libref.Tag, libref.Variant, len(seq), taglen)
 218                 }
 219                 overlap := taglen
 220                 if len(refseq) == 0 {
 221                         overlap = 0
 222                 }
 223                 tilestart[libref.Tag] = len(refseq) - overlap
 224                 refseq = append(refseq, seq[overlap:]...)
 225                 tileend[libref.Tag] = len(refseq)
 226         }
 227         log.Infof("seq %s len(refseq) %d len(tilestart) %d", seqname, len(refseq), len(tilestart))
 228         // outtag is tag's index in the subset of tags that aren't
 229         // dropped. If there are 10M tags and half are dropped by
 230         // dropTiles, tag ranges from 0 to 10M-1 and outtag ranges
 231         // from 0 to 5M-1.
 232         //
 233         // IOW, in the matrix built by cgs2array(), {tag} is
 234         // represented by columns {outtag}*2 and {outtag}*2+1.
 235         outcol := -1
 236         for tag, tvs := range tilelib.variant {
 237                 if len(cmd.dropTiles) > tag && cmd.dropTiles[tag] {
 238                         continue
 239                 }
 240                 tag := tagID(tag)
 241                 outcol++
 242                 // Must shadow outcol var to use safely in goroutine below.
 243                 outcol := outcol
 244                 refstart, ok := tilestart[tag]
 245                 if !ok {
 246                         // Tag didn't place on this reference
 247                         // sequence. (It might place on the same
 248                         // chromosome in a genome anyway, but we don't
 249                         // output the annotations that would result.)
 250                         // outch <- fmt.Sprintf("%d,%d,-1%s\n", tag, outcol, refnamefield)
 251                         continue
 252                 }
 253                 for variant := 1; variant <= len(tvs); variant++ {
 254                         variant, hash := tileVariantID(variant), tvs[variant-1]
 255                         tileseq := tilelib.TileVariantSequence(tileLibRef{Tag: tag, Variant: variant})
 256                         if len(tileseq) == 0 {
 257                                 continue
 258                         } else if len(tileseq) < taglen {
 259                                 return fmt.Errorf("tilevar %d,%d has sequence len %d < taglen %d", tag, variant, len(tileseq), taglen)
 260                         }
 261                         var refpart []byte
 262                         endtag := string(tileseq[len(tileseq)-taglen:])
 263                         if endtagid, ok := cmd.tag2tagid[endtag]; !ok {
 264                                 // Tile variant doesn't end on a tag, so it can only place at the end of a chromosome.
 265                                 refpart = refseq[refstart:]
 266                                 log.Warnf("%x tilevar %d,%d endtag not in ref: %s", hash[:13], tag, variant, endtag)
 267                         } else if refendtagstart, ok := tilestart[endtagid]; !ok {
 268                                 // Ref ends a chromsome with a (possibly very large) variant of this tile, but genomes with this tile don't.
 269                                 // Give up. (TODO: something smarter)
 270                                 log.Debugf("%x not annotating tilevar %d,%d because end tag %d is not in ref", hash[:13], tag, variant, endtagid)
 271                                 continue
 272                         } else {
 273                                 // Non-terminal tile vs. non-terminal reference.
 274                                 refpart = refseq[refstart : refendtagstart+taglen]
 275                                 log.Tracef("\n%x tilevar %d,%d endtag %s endtagid %d refendtagstart %d", hash[:13], tag, variant, endtag, endtagid, refendtagstart)
 276                         }
 277                         if len(refpart) > cmd.maxTileSize {
 278                                 log.Warnf("%x tilevar %d,%d skipping long diff, ref %s seq %s pos %d ref len %d", hash[:13], tag, variant, refname, seqname, refstart, len(refpart))
 279                                 continue
 280                         }
 281                         if len(tileseq) > cmd.maxTileSize {
 282                                 log.Warnf("%x tilevar %d,%d skipping long diff, ref %s seq %s variant len %d", hash[:13], tag, variant, refname, seqname, len(tileseq))
 283                                 continue
 284                         }
 285                         // log.Printf("\n%x @ refstart %d \n< %s\n> %s\n", tv.Blake2b, refstart, refpart, tileseq)
 286
 287                         throttle.Acquire()
 288                         go func() {
 289                                 defer throttle.Release()
 290                                 diffs, _ := hgvs.Diff(strings.ToUpper(string(refpart)), strings.ToUpper(string(tileseq)), 0)
 291                                 for _, diff := range diffs {
 292                                         diff.Position += refstart
 293                                         var varid string
 294                                         if cmd.variantHash {
 295                                                 varid = fmt.Sprintf("%x", hash)[:13]
 296                                         } else {
 297                                                 varid = fmt.Sprintf("%d", variant)
 298                                         }
 299                                         outch <- fmt.Sprintf("%d,%d,%s%s,%s:g.%s\n", tag, outcol, varid, refnamefield, seqname, diff.String())
 300                                         if cmd.reportAnnotation != nil {
 301                                                 cmd.reportAnnotation(tag, outcol, variant, refname, seqname, diff)
 302                                         }
 303                                 }
 304                         }()
 305                 }
 306         }
 307         return nil
 308 }