import (
"bufio"
"bytes"
+ "context"
+ "encoding/gob"
+ "fmt"
"io"
"strings"
"sync"
}
type tileLibrary struct {
- skipOOO bool
- taglib *tagLibrary
- variant [][][blake2b.Size256]byte
+ includeNoCalls bool
+ skipOOO bool
+ taglib *tagLibrary
+ variant [][][blake2b.Size256]byte
// count [][]int
// seq map[[blake2b.Size]byte][]byte
variants int
+ // if non-nil, write out any tile variants added while tiling
+ encoder *gob.Encoder
+ // if non-nil, call this func upon loading a genome
+ onLoadGenome func(CompactGenome)
mtx sync.Mutex
}
+func (tilelib *tileLibrary) loadTagSet(newtagset [][]byte) error {
+ // Loading a tagset means either passing it through to the
+ // output (if it's the first one we've seen), or just ensuring
+ // it doesn't disagree with what we already have.
+ if len(newtagset) == 0 {
+ return nil
+ }
+ tilelib.mtx.Lock()
+ defer tilelib.mtx.Unlock()
+ if tilelib.taglib == nil || tilelib.taglib.Len() == 0 {
+ tilelib.taglib = &tagLibrary{}
+ err := tilelib.taglib.setTags(newtagset)
+ if err != nil {
+ return err
+ }
+ if tilelib.encoder != nil {
+ err = tilelib.encoder.Encode(LibraryEntry{
+ TagSet: newtagset,
+ })
+ if err != nil {
+ return err
+ }
+ }
+ } else if tilelib.taglib.Len() != len(newtagset) {
+ return fmt.Errorf("cannot merge libraries with differing tagsets")
+ } else {
+ current := tilelib.taglib.Tags()
+ for i := range newtagset {
+ if !bytes.Equal(newtagset[i], current[i]) {
+ return fmt.Errorf("cannot merge libraries with differing tagsets")
+ }
+ }
+ }
+ return nil
+}
+
+func (tilelib *tileLibrary) loadTileVariants(tvs []TileVariant, variantmap map[tileLibRef]tileVariantID) error {
+ for _, tv := range tvs {
+ // Assign a new variant ID (unique across all inputs)
+ // for each input variant.
+ variantmap[tileLibRef{tag: tv.Tag, variant: tv.Variant}] = tilelib.getRef(tv.Tag, tv.Sequence).variant
+ }
+ return nil
+}
+
+func (tilelib *tileLibrary) loadGenomes(genomes map[string][]tileVariantID, variantmap map[tileLibRef]tileVariantID, onLoadGenome func(CompactGenome)) error {
+ var wg sync.WaitGroup
+ errs := make(chan error, 1)
+ for name, variants := range genomes {
+ name, variants := name, variants
+ wg.Add(1)
+ go func() {
+ defer wg.Done()
+ for i, variant := range variants {
+ if len(errs) > 0 {
+ return
+ }
+ if variant == 0 {
+ continue
+ }
+ tag := tagID(i / 2)
+ newvariant, ok := variantmap[tileLibRef{tag: tag, variant: variant}]
+ if !ok {
+ err := fmt.Errorf("oops: genome %q has variant %d for tag %d, but that variant was not in its library", name, variant, tag)
+ select {
+ case errs <- err:
+ default:
+ }
+ return
+ }
+ variants[i] = newvariant
+ }
+ if tilelib.encoder != nil {
+ for name, variants := range genomes {
+ cg := CompactGenome{
+ Name: name,
+ Variants: variants,
+ }
+ if onLoadGenome != nil {
+ onLoadGenome(cg)
+ }
+ err := tilelib.encoder.Encode(LibraryEntry{
+ CompactGenomes: []CompactGenome{cg},
+ })
+ if err != nil {
+ select {
+ case errs <- err:
+ default:
+ }
+ return
+ }
+ }
+ }
+ }()
+ }
+ wg.Wait()
+ go close(errs)
+ return <-errs
+}
+
+func (tilelib *tileLibrary) LoadGob(ctx context.Context, rdr io.Reader, onLoadGenome func(CompactGenome)) error {
+ genomes := map[string][]tileVariantID{}
+ variantmap := map[tileLibRef]tileVariantID{}
+ err := DecodeLibrary(rdr, func(ent *LibraryEntry) error {
+ if ctx.Err() != nil {
+ return ctx.Err()
+ }
+ if err := tilelib.loadTagSet(ent.TagSet); err != nil {
+ return err
+ } else if err = tilelib.loadTileVariants(ent.TileVariants, variantmap); err != nil {
+ return err
+ } else {
+ for _, cg := range ent.CompactGenomes {
+ genomes[cg.Name] = cg.Variants
+ }
+ return nil
+ }
+ })
+ if err != nil {
+ return err
+ }
+ if ctx.Err() != nil {
+ return ctx.Err()
+ }
+ err = tilelib.loadGenomes(genomes, variantmap, onLoadGenome)
+ if err != nil {
+ return err
+ }
+ return nil
+}
+
func (tilelib *tileLibrary) TileFasta(filelabel string, rdr io.Reader) (tileSeq, error) {
ret := tileSeq{}
type jobT struct {
var seqlabel string
for scanner.Scan() {
buf := scanner.Bytes()
- if len(buf) == 0 || buf[0] == '>' {
+ if len(buf) > 0 && buf[0] == '>' {
todo <- jobT{seqlabel, fasta}
seqlabel, fasta = string(buf[1:]), nil
log.Debugf("%s %s reading fasta", filelabel, seqlabel)
// Return a tileLibRef for a tile with the given tag and sequence,
// adding the sequence to the library if needed.
func (tilelib *tileLibrary) getRef(tag tagID, seq []byte) tileLibRef {
- for _, b := range seq {
- if b != 'a' && b != 'c' && b != 'g' && b != 't' {
- // return "tile not found" if seq has any
- // no-calls
- return tileLibRef{tag: tag}
+ if !tilelib.includeNoCalls {
+ for _, b := range seq {
+ if b != 'a' && b != 'c' && b != 'g' && b != 't' {
+ // return "tile not found" if seq has any
+ // no-calls
+ return tileLibRef{tag: tag}
+ }
}
}
tilelib.mtx.Lock()
- defer tilelib.mtx.Unlock()
// if tilelib.seq == nil {
// tilelib.seq = map[[blake2b.Size]byte][]byte{}
// }
- if tilelib.variant == nil {
+ if tilelib.variant == nil && tilelib.taglib != nil {
tilelib.variant = make([][][blake2b.Size256]byte, tilelib.taglib.Len())
}
+ if int(tag) >= len(tilelib.variant) {
+ // If we haven't seen the tag library yet (as in a
+ // merge), tilelib.taglib.Len() is zero. We can still
+ // behave correctly, we just need to expand the
+ // tilelib.variant slice as needed.
+ if int(tag) >= cap(tilelib.variant) {
+ // Allocate 2x capacity.
+ newslice := make([][][blake2b.Size256]byte, int(tag)+1, (int(tag)+1)*2)
+ copy(newslice, tilelib.variant)
+ tilelib.variant = newslice[:int(tag)+1]
+ } else {
+ // Use previously allocated capacity, avoiding
+ // copy.
+ tilelib.variant = tilelib.variant[:int(tag)+1]
+ }
+ }
seqhash := blake2b.Sum256(seq)
for i, varhash := range tilelib.variant[tag] {
if varhash == seqhash {
+ tilelib.mtx.Unlock()
return tileLibRef{tag: tag, variant: tileVariantID(i + 1)}
}
}
tilelib.variants++
tilelib.variant[tag] = append(tilelib.variant[tag], seqhash)
// tilelib.seq[seqhash] = append([]byte(nil), seq...)
- return tileLibRef{tag: tag, variant: tileVariantID(len(tilelib.variant[tag]))}
+ variant := tileVariantID(len(tilelib.variant[tag]))
+ tilelib.mtx.Unlock()
+
+ if tilelib.encoder != nil {
+ tilelib.encoder.Encode(LibraryEntry{
+ TileVariants: []TileVariant{{
+ Tag: tag,
+ Variant: variant,
+ Blake2b: seqhash,
+ Sequence: seq,
+ }},
+ })
+ }
+ return tileLibRef{tag: tag, variant: variant}
}
projects / lightning.git / blobdiff