-package main
+package lightning
import (
"bufio"
"net/http"
_ "net/http/pprof"
"os"
+ "path"
+ "runtime"
"sort"
"strings"
"sync"
var (
outputFormats = map[string]outputFormat{
- "hgvs": outputFormatHGVS,
- "vcf": outputFormatVCF,
+ "hgvs-onehot": outputFormatHGVSOneHot,
+ "hgvs": outputFormatHGVS,
+ "vcf": outputFormatVCF,
}
- outputFormatHGVS = outputFormat{Print: printHGVS}
- outputFormatVCF = outputFormat{Print: printVCF, PadLeft: true}
+ outputFormatHGVS = outputFormat{Print: printHGVS}
+ outputFormatHGVSOneHot = outputFormat{Print: printHGVSOneHot}
+ outputFormatVCF = outputFormat{Print: printVCF, PadLeft: true}
)
type exporter struct {
outputFormat outputFormat
+ maxTileSize int
}
func (cmd *exporter) RunCommand(prog string, args []string, stdin io.Reader, stdout, stderr io.Writer) int {
flags := flag.NewFlagSet("", flag.ContinueOnError)
flags.SetOutput(stderr)
pprof := flags.String("pprof", "", "serve Go profile data at http://`[addr]:port`")
+ pprofdir := flags.String("pprof-dir", "", "write Go profile data to `directory` periodically")
runlocal := flags.Bool("local", false, "run on local host (default: run in an arvados container)")
projectUUID := flags.String("project", "", "project `UUID` for output data")
priority := flags.Int("priority", 500, "container request priority")
outputFilename := flags.String("o", "-", "output `file`")
outputFormatStr := flags.String("output-format", "hgvs", "output `format`: hgvs or vcf")
outputBed := flags.String("output-bed", "", "also output bed `file`")
- pick := flags.String("pick", "", "`name` of single genome to export")
+ labelsFilename := flags.String("output-labels", "", "also output genome labels csv `file`")
+ flags.IntVar(&cmd.maxTileSize, "max-tile-size", 50000, "don't try to make annotations for tiles bigger than given `size`")
err = flags.Parse(args)
if err == flag.ErrHelp {
err = nil
log.Println(http.ListenAndServe(*pprof, nil))
}()
}
+ if *pprofdir != "" {
+ go writeProfilesPeriodically(*pprofdir)
+ }
if !*runlocal {
if *outputFilename != "-" {
Name: "lightning export",
Client: arvados.NewClientFromEnv(),
ProjectUUID: *projectUUID,
- RAM: 240000000000,
- VCPUs: 32,
+ RAM: 700000000000,
+ VCPUs: 64,
Priority: *priority,
}
err = runner.TranslatePaths(inputFilename)
}
*outputBed = "/mnt/output/" + *outputBed
}
- runner.Args = []string{"export", "-local=true", "-pick", *pick, "-ref", *refname, "-output-format", *outputFormatStr, "-output-bed", *outputBed, "-i", *inputFilename, "-o", "/mnt/output/export.csv"}
+ runner.Args = []string{"export", "-local=true",
+ "-pprof", ":6000",
+ "-pprof-dir", "/mnt/output",
+ "-ref", *refname,
+ "-output-format", *outputFormatStr,
+ "-output-bed", *outputBed,
+ "-output-labels", "/mnt/output/labels.csv",
+ "-max-tile-size", fmt.Sprintf("%d", cmd.maxTileSize),
+ "-i", *inputFilename,
+ "-o", "/mnt/output/export.csv",
+ }
var output string
output, err = runner.Run()
if err != nil {
return 0
}
- input, err := os.Open(*inputFilename)
+ in, err := open(*inputFilename)
if err != nil {
return 1
}
- defer input.Close()
+ defer in.Close()
+ input, ok := in.(io.ReadSeeker)
+ if !ok {
+ err = fmt.Errorf("%s: %T cannot seek", *inputFilename, in)
+ return 1
+ }
// Error out early if seeking doesn't work on the input file.
_, err = input.Seek(0, io.SeekEnd)
return 1
}
- var mtx sync.Mutex
var cgs []CompactGenome
- tilelib := tileLibrary{
- includeNoCalls: true,
+ tilelib := &tileLibrary{
+ retainNoCalls: true,
+ compactGenomes: map[string][]tileVariantID{},
}
- err = tilelib.LoadGob(context.Background(), input, func(cg CompactGenome) {
- if *pick != "" && *pick != cg.Name {
- return
- }
- log.Debugf("export: pick %q", cg.Name)
- mtx.Lock()
- defer mtx.Unlock()
- cgs = append(cgs, cg)
- })
+ err = tilelib.LoadGob(context.Background(), input, strings.HasSuffix(*inputFilename, ".gz"), nil)
if err != nil {
return 1
}
- sort.Slice(cgs, func(i, j int) bool { return cgs[i].Name < cgs[j].Name })
- log.Printf("export: pick %q => %d genomes", *pick, len(cgs))
refseq, ok := tilelib.refseqs[*refname]
if !ok {
return 1
}
+ names := cgnames(tilelib)
+ for _, name := range names {
+ cgs = append(cgs, CompactGenome{Name: name, Variants: tilelib.compactGenomes[name]})
+ }
+ if *labelsFilename != "" {
+ log.Infof("writing labels to %s", *labelsFilename)
+ var f *os.File
+ f, err = os.OpenFile(*labelsFilename, os.O_CREATE|os.O_WRONLY, 0777)
+ if err != nil {
+ return 1
+ }
+ defer f.Close()
+ _, outBasename := path.Split(*outputFilename)
+ for i, name := range names {
+ _, err = fmt.Fprintf(f, "%d,%q,%q\n", i, trimFilenameForLabel(name), outBasename)
+ if err != nil {
+ err = fmt.Errorf("write %s: %w", *labelsFilename, err)
+ return 1
+ }
+ }
+ err = f.Close()
+ if err != nil {
+ err = fmt.Errorf("close %s: %w", *labelsFilename, err)
+ return 1
+ }
+ }
+
_, err = input.Seek(0, io.SeekStart)
if err != nil {
return 1
bedbufw = bufio.NewWriter(bedout)
}
- err = cmd.export(bufw, bedout, input, tilelib.taglib.keylen, refseq, cgs)
+ err = cmd.export(bufw, bedout, input, strings.HasSuffix(*inputFilename, ".gz"), tilelib, refseq, cgs)
if err != nil {
return 1
}
return 1
}
}
- err = input.Close()
+ err = in.Close()
if err != nil {
return 1
}
return 0
}
-func (cmd *exporter) export(out, bedout io.Writer, librdr io.Reader, taglen int, refseq map[string][]tileLibRef, cgs []CompactGenome) error {
+func (cmd *exporter) export(out, bedout io.Writer, librdr io.Reader, gz bool, tilelib *tileLibrary, refseq map[string][]tileLibRef, cgs []CompactGenome) error {
need := map[tileLibRef]bool{}
var seqnames []string
for seqname, librefs := range refseq {
log.Infof("export: loading %d tile variants", len(need))
tileVariant := map[tileLibRef]TileVariant{}
- err := DecodeLibrary(librdr, func(ent *LibraryEntry) error {
+ err := DecodeLibrary(librdr, gz, func(ent *LibraryEntry) error {
for _, tv := range ent.TileVariants {
- libref := tileLibRef{Tag: tv.Tag, Variant: tv.Variant}
+ libref := tilelib.getRef(tv.Tag, tv.Sequence)
if need[libref] {
tileVariant[libref] = tv
}
return fmt.Errorf("%d needed tiles are missing from library", len(missing))
}
- log.Infof("assembling %d sequences concurrently", len(seqnames))
- var wg sync.WaitGroup
- outbuf := make([]bytes.Buffer, len(seqnames))
- bedbuf := make([]bytes.Buffer, len(seqnames))
- for seqidx, seqname := range seqnames {
- seqname := seqname
- outbuf := &outbuf[seqidx]
- bedbuf := &bedbuf[seqidx]
- if bedout == nil {
- bedbuf = nil
- }
- // TODO: limit number of goroutines and unflushed bufs to ncpus
- wg.Add(1)
- go func() {
- defer wg.Done()
- cmd.exportSeq(outbuf, bedbuf, taglen, seqname, refseq[seqname], tileVariant, cgs)
- log.Infof("assembled %q to outbuf %d bedbuf %d", seqname, outbuf.Len(), bedbuf.Len())
- }()
- }
- wg.Wait()
+ outw := make([]io.WriteCloser, len(seqnames))
+ bedw := make([]io.WriteCloser, len(seqnames))
- wg.Add(1)
- go func() {
- defer wg.Done()
+ var merges sync.WaitGroup
+ merge := func(dst io.Writer, src []io.WriteCloser, label string) {
+ var mtx sync.Mutex
for i, seqname := range seqnames {
- log.Infof("writing outbuf %s", seqname)
- io.Copy(out, &outbuf[i])
+ pr, pw := io.Pipe()
+ src[i] = pw
+ merges.Add(1)
+ seqname := seqname
+ go func() {
+ defer merges.Done()
+ log.Infof("writing %s %s", seqname, label)
+ scanner := bufio.NewScanner(pr)
+ for scanner.Scan() {
+ mtx.Lock()
+ dst.Write(scanner.Bytes())
+ dst.Write([]byte{'\n'})
+ mtx.Unlock()
+ }
+ log.Infof("writing %s %s done", seqname, label)
+ }()
}
- }()
+ }
+ merge(out, outw, "output")
if bedout != nil {
- wg.Add(1)
+ merge(bedout, bedw, "bed")
+ }
+
+ throttle := throttle{Max: runtime.NumCPU()}
+ log.Infof("assembling %d sequences in %d goroutines", len(seqnames), throttle.Max)
+ for seqidx, seqname := range seqnames {
+ seqidx, seqname := seqidx, seqname
+ outw := outw[seqidx]
+ bedw := bedw[seqidx]
+ throttle.Acquire()
go func() {
- defer wg.Done()
- for i, seqname := range seqnames {
- log.Infof("writing bedbuf %s", seqname)
- io.Copy(bedout, &bedbuf[i])
+ defer throttle.Release()
+ if bedw != nil {
+ defer bedw.Close()
}
+ defer outw.Close()
+ outwb := bufio.NewWriter(outw)
+ defer outwb.Flush()
+ cmd.exportSeq(outwb, bedw, tilelib.taglib.keylen, seqname, refseq[seqname], tileVariant, cgs)
}()
}
- wg.Wait()
+
+ merges.Wait()
return nil
}
continue
}
genometile := tileVariant[tileLibRef{Tag: libref.Tag, Variant: variant}]
+ if len(genometile.Sequence) == 0 {
+ // Hash is known but sequence
+ // is not, e.g., retainNoCalls
+ // was false during import
+ continue
+ }
+ if len(genometile.Sequence) > cmd.maxTileSize {
+ continue
+ }
refSequence := reftile.Sequence
// If needed, extend the reference
// sequence up to the tag at the end
// of the genometile sequence.
refstepend := refstep + 1
- for refstepend < len(reftiles) && len(refSequence) >= taglen && !bytes.EqualFold(refSequence[len(refSequence)-taglen:], genometile.Sequence[len(genometile.Sequence)-taglen:]) {
+ for refstepend < len(reftiles) && len(refSequence) >= taglen && !bytes.EqualFold(refSequence[len(refSequence)-taglen:], genometile.Sequence[len(genometile.Sequence)-taglen:]) && len(refSequence) <= cmd.maxTileSize {
if &refSequence[0] == &reftile.Sequence[0] {
refSequence = append([]byte(nil), refSequence...)
}
v = v.PadLeft()
}
v.Position += refpos
- log.Debugf("%s seq %s phase %d tag %d tile diff %s\n", cg.Name, seqname, phase, libref.Tag, v.String())
varslice := variantAt[v.Position]
if varslice == nil {
varslice = make([]hgvs.Variant, len(cgs)*2)
thickstart = 0
}
thickend := refpos
+
// coverage score, 0 to 1000
- score := 1000 * tagcoverage / len(cgs) / 2
+ score := 1000
+ if len(cgs) > 0 {
+ score = 1000 * tagcoverage / len(cgs) / 2
+ }
+
fmt.Fprintf(bedw, "%s %d %d %d %d . %d %d\n",
seqname, tilestart, tileend,
libref.Tag,
}
out.Write([]byte{'\n'})
}
+
+func printHGVSOneHot(out io.Writer, seqname string, varslice []hgvs.Variant) {
+ vars := map[hgvs.Variant]bool{}
+ for _, v := range varslice {
+ if v.Ref != v.New {
+ vars[v] = true
+ }
+ }
+
+ // sort variants to ensure output is deterministic
+ sorted := make([]hgvs.Variant, 0, len(vars))
+ for v := range vars {
+ sorted = append(sorted, v)
+ }
+ sort.Slice(sorted, func(a, b int) bool { return hgvs.Less(sorted[a], sorted[b]) })
+
+ for _, v := range sorted {
+ fmt.Fprintf(out, "%s.%s", seqname, v.String())
+ for i := 0; i < len(varslice); i += 2 {
+ if varslice[i] == v || varslice[i+1] == v {
+ out.Write([]byte("\t1"))
+ } else {
+ out.Write([]byte("\t0"))
+ }
+ }
+ out.Write([]byte{'\n'})
+ }
+}
projects / lightning.git / blobdiff