11 log "github.com/sirupsen/logrus"
12 "golang.org/x/crypto/blake2b"
15 type tileVariantID uint16 // 1-based
17 type tileLibRef struct {
22 type tileSeq map[string][]tileLibRef
24 func (tseq tileSeq) Variants() ([]tileVariantID, int, int) {
26 for _, refs := range tseq {
27 for _, ref := range refs {
28 if maxtag < int(ref.tag) {
33 vars := make([]tileVariantID, maxtag+1)
35 for _, refs := range tseq {
36 for _, ref := range refs {
37 if vars[int(ref.tag)] != 0 {
42 vars[int(ref.tag)] = ref.variant
45 return vars, kept, dropped
48 type tileLibrary struct {
51 variant [][][blake2b.Size256]byte
53 // seq map[[blake2b.Size]byte][]byte
55 // if non-nil, write out any tile variants added while tiling
61 func (tilelib *tileLibrary) TileFasta(filelabel string, rdr io.Reader) (tileSeq, error) {
67 todo := make(chan jobT)
68 scanner := bufio.NewScanner(rdr)
74 buf := scanner.Bytes()
75 if len(buf) > 0 && buf[0] == '>' {
76 todo <- jobT{seqlabel, fasta}
77 seqlabel, fasta = string(buf[1:]), nil
78 log.Debugf("%s %s reading fasta", filelabel, seqlabel)
80 fasta = append(fasta, bytes.ToLower(buf)...)
83 todo <- jobT{seqlabel, fasta}
85 type foundtag struct {
90 found := make([]foundtag, 2000000)
91 path := make([]tileLibRef, 2000000)
95 for job := range todo {
96 if len(job.fasta) == 0 {
98 } else if strings.Contains(job.label, "_") {
102 log.Debugf("%s %s tiling", filelabel, job.label)
105 tilelib.taglib.FindAll(job.fasta, func(tagid tagID, pos, taglen int) {
106 found = append(found, foundtag{pos: pos, tagid: tagid, taglen: taglen})
108 totalFoundTags += len(found)
112 last := foundtag{tagid: -1}
114 keep := longestIncreasingSubsequence(len(found), func(i int) int { return int(found[i].tagid) })
115 for i, x := range keep {
118 skipped = len(found) - len(keep)
119 found = found[:len(keep)]
121 for i, f := range found {
122 log.Tracef("%s %s found[%d] == %#v", filelabel, job.label, i, f)
124 path = append(path, tilelib.getRef(last.tagid, job.fasta[last.pos:f.pos+f.taglen]))
129 path = append(path, tilelib.getRef(last.tagid, job.fasta[last.pos:]))
132 pathcopy := make([]tileLibRef, len(path))
134 ret[job.label] = pathcopy
135 log.Debugf("%s %s tiled with path len %d, skipped %d", filelabel, job.label, len(path), skipped)
136 totalPathLen += len(path)
138 log.Printf("%s tiled with total path len %d in %d sequences (skipped %d sequences with '_' in name, skipped %d out-of-order tags)", filelabel, totalPathLen, len(ret), skippedSequences, totalFoundTags-totalPathLen)
139 return ret, scanner.Err()
142 func (tilelib *tileLibrary) Len() int {
144 defer tilelib.mtx.Unlock()
145 return tilelib.variants
148 // Return a tileLibRef for a tile with the given tag and sequence,
149 // adding the sequence to the library if needed.
150 func (tilelib *tileLibrary) getRef(tag tagID, seq []byte) tileLibRef {
151 for _, b := range seq {
152 if b != 'a' && b != 'c' && b != 'g' && b != 't' {
153 // return "tile not found" if seq has any
155 return tileLibRef{tag: tag}
159 // if tilelib.seq == nil {
160 // tilelib.seq = map[[blake2b.Size]byte][]byte{}
162 if tilelib.variant == nil {
163 tilelib.variant = make([][][blake2b.Size256]byte, tilelib.taglib.Len())
165 seqhash := blake2b.Sum256(seq)
166 for i, varhash := range tilelib.variant[tag] {
167 if varhash == seqhash {
169 return tileLibRef{tag: tag, variant: tileVariantID(i + 1)}
173 tilelib.variant[tag] = append(tilelib.variant[tag], seqhash)
174 // tilelib.seq[seqhash] = append([]byte(nil), seq...)
175 ret := tileLibRef{tag: tag, variant: tileVariantID(len(tilelib.variant[tag]))}
178 if tilelib.encoder != nil {
179 tilelib.encoder.Encode(LibraryEntry{
180 TileVariants: []TileVariant{{