tilelib.go

   1 package main
   2
   3 import (
   4         "bufio"
   5         "bytes"
   6         "io"
   7         "log"
   8         "strings"
   9         "sync"
  10
  11         "golang.org/x/crypto/blake2b"
  12 )
  13
  14 type tileVariantID int32 // 1-based
  15
  16 type tileLibRef struct {
  17         tag     tagID
  18         variant tileVariantID
  19 }
  20
  21 type tileSeq map[string][]tileLibRef
  22
  23 type tileLibrary struct {
  24         taglib  *tagLibrary
  25         variant [][][blake2b.Size256]byte
  26         // count [][]int
  27         // seq map[[blake2b.Size]byte][]byte
  28         variants int
  29
  30         mtx sync.Mutex
  31 }
  32
  33 func (tilelib *tileLibrary) TileFasta(filelabel string, rdr io.Reader) (tileSeq, error) {
  34         ret := tileSeq{}
  35         type jobT struct {
  36                 label string
  37                 fasta []byte
  38         }
  39         todo := make(chan jobT)
  40         scanner := bufio.NewScanner(rdr)
  41         go func() {
  42                 defer close(todo)
  43                 var fasta []byte
  44                 var seqlabel string
  45                 for scanner.Scan() {
  46                         buf := scanner.Bytes()
  47                         if len(buf) == 0 || buf[0] == '>' {
  48                                 todo <- jobT{seqlabel, fasta}
  49                                 seqlabel, fasta = string(buf[1:]), nil
  50                                 log.Printf("%s %s reading fasta", filelabel, seqlabel)
  51                         } else {
  52                                 fasta = append(fasta, bytes.ToLower(buf)...)
  53                         }
  54                 }
  55                 todo <- jobT{seqlabel, fasta}
  56         }()
  57         path := make([]tileLibRef, 2000000)
  58         for job := range todo {
  59                 if len(job.fasta) == 0 || strings.Contains(job.label, "_") {
  60                         continue
  61                 }
  62                 log.Printf("%s %s tiling", filelabel, job.label)
  63                 path = path[:0]
  64                 tilestart := -1        // position in fasta of tile that ends here
  65                 tiletagid := tagID(-1) // tag id starting tile that ends here
  66                 tilelib.taglib.FindAll(job.fasta, func(id tagID, pos, taglen int) {
  67                         if tilestart >= 0 {
  68                                 path = append(path, tilelib.getRef(tiletagid, job.fasta[tilestart:pos+taglen]))
  69                         }
  70                         tilestart = pos
  71                         tiletagid = id
  72                 })
  73                 if tiletagid >= 0 {
  74                         path = append(path, tilelib.getRef(tiletagid, job.fasta[tilestart:]))
  75                 }
  76                 pathcopy := make([]tileLibRef, len(path))
  77                 copy(pathcopy, path)
  78                 ret[job.label] = pathcopy
  79                 log.Printf("%s %s tiled with path len %d", filelabel, job.label, len(path))
  80         }
  81         return ret, scanner.Err()
  82 }
  83
  84 func (tilelib *tileLibrary) Len() int {
  85         tilelib.mtx.Lock()
  86         defer tilelib.mtx.Unlock()
  87         return tilelib.variants
  88 }
  89
  90 // Return a tileLibRef for a tile with the given tag and sequence,
  91 // adding the sequence to the library if needed.
  92 func (tilelib *tileLibrary) getRef(tag tagID, seq []byte) tileLibRef {
  93         for _, b := range seq {
  94                 if b != 'a' && b != 'c' && b != 'g' && b != 't' {
  95                         // return "tile not found" if seq has any
  96                         // no-calls
  97                         return tileLibRef{tag: tag}
  98                 }
  99         }
 100         tilelib.mtx.Lock()
 101         defer tilelib.mtx.Unlock()
 102         // if tilelib.seq == nil {
 103         //      tilelib.seq = map[[blake2b.Size]byte][]byte{}
 104         // }
 105         if tilelib.variant == nil {
 106                 tilelib.variant = make([][][blake2b.Size256]byte, tilelib.taglib.Len())
 107         }
 108         seqhash := blake2b.Sum256(seq)
 109         for i, varhash := range tilelib.variant[tag] {
 110                 if varhash == seqhash {
 111                         return tileLibRef{tag: tag, variant: tileVariantID(i + 1)}
 112                 }
 113         }
 114         tilelib.variants++
 115         tilelib.variant[tag] = append(tilelib.variant[tag], seqhash)
 116         // tilelib.seq[seqhash] = append([]byte(nil), seq...)
 117         return tileLibRef{tag: tag, variant: tileVariantID(len(tilelib.variant[tag]))}
 118 }