X-Git-Url: https://git.arvados.org/lightning.git/blobdiff_plain/2fa08e11db49c9aa2dddeaaf7b8ff5b44d682a65..ff2b46988b3e6c076766283325bb5ac04ab753f3:/tilelib.go

diff --git a/tilelib.go b/tilelib.go
index 403be31c7b..8c88956b97 100644
--- a/tilelib.go
+++ b/tilelib.go
@@ -7,8 +7,12 @@ import (
 	"encoding/gob"
 	"fmt"
 	"io"
+	"regexp"
+	"runtime"
+	"sort"
 	"strings"
 	"sync"
+	"sync/atomic"
 
 	log "github.com/sirupsen/logrus"
 	"golang.org/x/crypto/blake2b"
@@ -48,18 +52,22 @@ func (tseq tileSeq) Variants() ([]tileVariantID, int, int) {
 }
 
 type tileLibrary struct {
-	includeNoCalls bool
-	skipOOO        bool
+	retainNoCalls       bool
+	skipOOO             bool
+	retainTileSequences bool
+
 	taglib         *tagLibrary
 	variant        [][][blake2b.Size256]byte
 	refseqs        map[string]map[string][]tileLibRef
+	compactGenomes map[string][]tileVariantID
 	// count [][]int
-	// seq map[[blake2b.Size]byte][]byte
-	variants int
+	seq      map[[blake2b.Size256]byte][]byte
+	variants int64
 	// if non-nil, write out any tile variants added while tiling
 	encoder *gob.Encoder
 
-	mtx sync.Mutex
+	mtx   sync.RWMutex
+	vlock []sync.Locker
 }
 
 func (tilelib *tileLibrary) loadTagSet(newtagset [][]byte) error {
@@ -112,11 +120,11 @@ func (tilelib *tileLibrary) loadCompactGenomes(cgs []CompactGenome, variantmap m
 	var wg sync.WaitGroup
 	errs := make(chan error, 1)
 	for _, cg := range cgs {
-		name, variants := cg.Name, cg.Variants
 		wg.Add(1)
+		cg := cg
 		go func() {
 			defer wg.Done()
-			for i, variant := range variants {
+			for i, variant := range cg.Variants {
 				if len(errs) > 0 {
 					return
 				}
@@ -126,15 +134,15 @@ func (tilelib *tileLibrary) loadCompactGenomes(cgs []CompactGenome, variantmap m
 				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)
+					err := fmt.Errorf("oops: genome %q has variant %d for tag %d, but that variant was not in its library", cg.Name, variant, tag)
 					select {
 					case errs <- err:
 					default:
 					}
 					return
 				}
-				log.Tracef("loadCompactGenomes: cg %s tag %d variant %d => %d", name, tag, variant, newvariant)
-				variants[i] = newvariant
+				log.Tracef("loadCompactGenomes: cg %s tag %d variant %d => %d", cg.Name, tag, variant, newvariant)
+				cg.Variants[i] = newvariant
 			}
 			if onLoadGenome != nil {
 				onLoadGenome(cg)
@@ -151,6 +159,11 @@ func (tilelib *tileLibrary) loadCompactGenomes(cgs []CompactGenome, variantmap m
 					return
 				}
 			}
+			if tilelib.compactGenomes != nil {
+				tilelib.mtx.Lock()
+				defer tilelib.mtx.Unlock()
+				tilelib.compactGenomes[cg.Name] = cg.Variants
+			}
 		}()
 	}
 	wg.Wait()
@@ -200,11 +213,11 @@ func (tilelib *tileLibrary) loadCompactSequences(cseqs []CompactSequence, varian
 // match.
 //
 // If onLoadGenome is non-nil, call it on each CompactGenome entry.
-func (tilelib *tileLibrary) LoadGob(ctx context.Context, rdr io.Reader, onLoadGenome func(CompactGenome)) error {
+func (tilelib *tileLibrary) LoadGob(ctx context.Context, rdr io.Reader, gz bool, onLoadGenome func(CompactGenome)) error {
 	cgs := []CompactGenome{}
 	cseqs := []CompactSequence{}
 	variantmap := map[tileLibRef]tileVariantID{}
-	err := DecodeLibrary(rdr, func(ent *LibraryEntry) error {
+	err := DecodeLibrary(rdr, gz, func(ent *LibraryEntry) error {
 		if ctx.Err() != nil {
 			return ctx.Err()
 		}
@@ -235,13 +248,22 @@ func (tilelib *tileLibrary) LoadGob(ctx context.Context, rdr io.Reader, onLoadGe
 	return nil
 }
 
-func (tilelib *tileLibrary) TileFasta(filelabel string, rdr io.Reader) (tileSeq, error) {
+type importStats struct {
+	InputFile              string
+	InputLabel             string
+	InputLength            int
+	InputCoverage          int
+	PathLength             int
+	DroppedOutOfOrderTiles int
+}
+
+func (tilelib *tileLibrary) TileFasta(filelabel string, rdr io.Reader, matchChromosome *regexp.Regexp) (tileSeq, []importStats, error) {
 	ret := tileSeq{}
 	type jobT struct {
 		label string
 		fasta []byte
 	}
-	todo := make(chan jobT)
+	todo := make(chan jobT, 1)
 	scanner := bufio.NewScanner(rdr)
 	go func() {
 		defer close(todo)
@@ -250,8 +272,8 @@ func (tilelib *tileLibrary) TileFasta(filelabel string, rdr io.Reader) (tileSeq,
 		for scanner.Scan() {
 			buf := scanner.Bytes()
 			if len(buf) > 0 && buf[0] == '>' {
-				todo <- jobT{seqlabel, fasta}
-				seqlabel, fasta = string(buf[1:]), nil
+				todo <- jobT{seqlabel, append([]byte(nil), fasta...)}
+				seqlabel, fasta = strings.SplitN(string(buf[1:]), " ", 2)[0], fasta[:0]
 				log.Debugf("%s %s reading fasta", filelabel, seqlabel)
 			} else {
 				fasta = append(fasta, bytes.ToLower(buf)...)
@@ -260,19 +282,20 @@ func (tilelib *tileLibrary) TileFasta(filelabel string, rdr io.Reader) (tileSeq,
 		todo <- jobT{seqlabel, fasta}
 	}()
 	type foundtag struct {
-		pos    int
-		tagid  tagID
-		taglen int
+		pos   int
+		tagid tagID
 	}
 	found := make([]foundtag, 2000000)
 	path := make([]tileLibRef, 2000000)
 	totalFoundTags := 0
 	totalPathLen := 0
 	skippedSequences := 0
+	taglen := tilelib.taglib.TagLen()
+	var stats []importStats
 	for job := range todo {
 		if len(job.fasta) == 0 {
 			continue
-		} else if strings.Contains(job.label, "_") {
+		} else if !matchChromosome.MatchString(job.label) {
 			skippedSequences++
 			continue
 		}
@@ -280,14 +303,16 @@ func (tilelib *tileLibrary) TileFasta(filelabel string, rdr io.Reader) (tileSeq,
 
 		found = found[:0]
 		tilelib.taglib.FindAll(job.fasta, func(tagid tagID, pos, taglen int) {
-			found = append(found, foundtag{pos: pos, tagid: tagid, taglen: taglen})
+			found = append(found, foundtag{pos: pos, tagid: tagid})
 		})
 		totalFoundTags += len(found)
+		if len(found) == 0 {
+			log.Warnf("%s %s no tags found", filelabel, job.label)
+		}
 
 		skipped := 0
-		path = path[:0]
-		last := foundtag{tagid: -1}
 		if tilelib.skipOOO {
+			log.Infof("%s %s keeping longest increasing subsequence", filelabel, job.label)
 			keep := longestIncreasingSubsequence(len(found), func(i int) int { return int(found[i].tagid) })
 			for i, x := range keep {
 				found[i] = found[x]
@@ -295,94 +320,271 @@ func (tilelib *tileLibrary) TileFasta(filelabel string, rdr io.Reader) (tileSeq,
 			skipped = len(found) - len(keep)
 			found = found[:len(keep)]
 		}
+
+		log.Infof("%s %s getting %d librefs", filelabel, job.label, len(found))
+		throttle := &throttle{Max: runtime.NumCPU()}
+		path = path[:len(found)]
+		var lowquality int64
 		for i, f := range found {
-			log.Tracef("%s %s found[%d] == %#v", filelabel, job.label, i, f)
-			if last.taglen > 0 {
-				path = append(path, tilelib.getRef(last.tagid, job.fasta[last.pos:f.pos+f.taglen]))
-			} else {
-				f.pos = 0
-			}
-			last = f
-		}
-		if last.taglen > 0 {
-			path = append(path, tilelib.getRef(last.tagid, job.fasta[last.pos:]))
-		} else {
-			log.Warnf("%s %s no tags found", filelabel, job.label)
+			i, f := i, f
+			throttle.Acquire()
+			go func() {
+				defer throttle.Release()
+				var startpos, endpos int
+				if i == 0 {
+					startpos = 0
+				} else {
+					startpos = f.pos
+				}
+				if i == len(found)-1 {
+					endpos = len(job.fasta)
+				} else {
+					endpos = found[i+1].pos + taglen
+				}
+				path[i] = tilelib.getRef(f.tagid, job.fasta[startpos:endpos])
+				if countBases(job.fasta[startpos:endpos]) != endpos-startpos {
+					atomic.AddInt64(&lowquality, 1)
+				}
+			}()
 		}
+		throttle.Wait()
+
+		log.Infof("%s %s copying path", filelabel, job.label)
 
 		pathcopy := make([]tileLibRef, len(path))
 		copy(pathcopy, path)
 		ret[job.label] = pathcopy
-		log.Debugf("%s %s tiled with path len %d, skipped %d", filelabel, job.label, len(path), skipped)
+
+		basesIn := countBases(job.fasta)
+		log.Infof("%s %s fasta in %d coverage in %d path len %d low-quality %d skipped-out-of-order %d", filelabel, job.label, len(job.fasta), basesIn, len(path), lowquality, skipped)
+		stats = append(stats, importStats{
+			InputFile:              filelabel,
+			InputLabel:             job.label,
+			InputLength:            len(job.fasta),
+			InputCoverage:          basesIn,
+			PathLength:             len(path),
+			DroppedOutOfOrderTiles: skipped,
+		})
+
 		totalPathLen += len(path)
 	}
-	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)
-	return ret, scanner.Err()
+	log.Printf("%s tiled with total path len %d in %d sequences (skipped %d sequences that did not match chromosome regexp, skipped %d out-of-order tags)", filelabel, totalPathLen, len(ret), skippedSequences, totalFoundTags-totalPathLen)
+	return ret, stats, scanner.Err()
 }
 
-func (tilelib *tileLibrary) Len() int {
-	tilelib.mtx.Lock()
-	defer tilelib.mtx.Unlock()
-	return tilelib.variants
+func (tilelib *tileLibrary) Len() int64 {
+	return atomic.LoadInt64(&tilelib.variants)
 }
 
 // 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 {
-	if !tilelib.includeNoCalls {
+	dropSeq := false
+	if !tilelib.retainNoCalls {
 		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}
+				dropSeq = true
+				break
 			}
 		}
 	}
-	tilelib.mtx.Lock()
-	// if tilelib.seq == nil {
-	// 	tilelib.seq = map[[blake2b.Size]byte][]byte{}
-	// }
-	if tilelib.variant == nil && tilelib.taglib != nil {
-		tilelib.variant = make([][][blake2b.Size256]byte, tilelib.taglib.Len())
+	seqhash := blake2b.Sum256(seq)
+	var vlock sync.Locker
+
+	tilelib.mtx.RLock()
+	if len(tilelib.vlock) > int(tag) {
+		vlock = tilelib.vlock[tag]
 	}
-	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]
+	tilelib.mtx.RUnlock()
+
+	if vlock != nil {
+		vlock.Lock()
+		for i, varhash := range tilelib.variant[tag] {
+			if varhash == seqhash {
+				vlock.Unlock()
+				return tileLibRef{Tag: tag, Variant: tileVariantID(i + 1)}
+			}
+		}
+		vlock.Unlock()
+	} else {
+		tilelib.mtx.Lock()
+		if tilelib.variant == nil && tilelib.taglib != nil {
+			tilelib.variant = make([][][blake2b.Size256]byte, tilelib.taglib.Len())
+			tilelib.vlock = make([]sync.Locker, tilelib.taglib.Len())
+			for i := range tilelib.vlock {
+				tilelib.vlock[i] = new(sync.Mutex)
+			}
 		}
+		if int(tag) >= len(tilelib.variant) {
+			oldlen := len(tilelib.vlock)
+			for i := 0; i < oldlen; i++ {
+				tilelib.vlock[i].Lock()
+			}
+			// 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 and
+			// tilelib.vlock slices 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]
+				newvlock := make([]sync.Locker, int(tag)+1, (int(tag)+1)*2)
+				copy(newvlock, tilelib.vlock)
+				tilelib.vlock = newvlock[:int(tag)+1]
+			} else {
+				// Use previously allocated capacity,
+				// avoiding copy.
+				tilelib.variant = tilelib.variant[:int(tag)+1]
+				tilelib.vlock = tilelib.vlock[:int(tag)+1]
+			}
+			for i := oldlen; i < len(tilelib.vlock); i++ {
+				tilelib.vlock[i] = new(sync.Mutex)
+			}
+			for i := 0; i < oldlen; i++ {
+				tilelib.vlock[i].Unlock()
+			}
+		}
+		vlock = tilelib.vlock[tag]
+		tilelib.mtx.Unlock()
 	}
-	seqhash := blake2b.Sum256(seq)
+
+	vlock.Lock()
 	for i, varhash := range tilelib.variant[tag] {
 		if varhash == seqhash {
-			tilelib.mtx.Unlock()
+			vlock.Unlock()
 			return tileLibRef{Tag: tag, Variant: tileVariantID(i + 1)}
 		}
 	}
-	tilelib.variants++
+	atomic.AddInt64(&tilelib.variants, 1)
 	tilelib.variant[tag] = append(tilelib.variant[tag], seqhash)
-	// tilelib.seq[seqhash] = append([]byte(nil), seq...)
 	variant := tileVariantID(len(tilelib.variant[tag]))
-	tilelib.mtx.Unlock()
+	vlock.Unlock()
+
+	if tilelib.retainTileSequences && !dropSeq {
+		tilelib.mtx.Lock()
+		if tilelib.seq == nil {
+			tilelib.seq = map[[blake2b.Size256]byte][]byte{}
+		}
+		tilelib.seq[seqhash] = append([]byte(nil), seq...)
+		tilelib.mtx.Unlock()
+	}
 
 	if tilelib.encoder != nil {
+		saveSeq := seq
+		if dropSeq {
+			// Save the hash, but not the sequence
+			saveSeq = nil
+		}
 		tilelib.encoder.Encode(LibraryEntry{
 			TileVariants: []TileVariant{{
 				Tag:      tag,
 				Variant:  variant,
 				Blake2b:  seqhash,
-				Sequence: seq,
+				Sequence: saveSeq,
 			}},
 		})
 	}
 	return tileLibRef{Tag: tag, Variant: variant}
 }
+
+func (tilelib *tileLibrary) TileVariantSequence(libref tileLibRef) []byte {
+	if libref.Variant == 0 || len(tilelib.variant) <= int(libref.Tag) || len(tilelib.variant[libref.Tag]) < int(libref.Variant) {
+		return nil
+	}
+	return tilelib.seq[tilelib.variant[libref.Tag][libref.Variant-1]]
+}
+
+// Tidy deletes unreferenced tile variants and renumbers variants so
+// more common variants have smaller IDs.
+func (tilelib *tileLibrary) Tidy() {
+	log.Print("Tidy: compute inref")
+	inref := map[tileLibRef]bool{}
+	for _, refseq := range tilelib.refseqs {
+		for _, librefs := range refseq {
+			for _, libref := range librefs {
+				inref[libref] = true
+			}
+		}
+	}
+	log.Print("Tidy: compute remap")
+	remap := make([][]tileVariantID, len(tilelib.variant))
+	throttle := throttle{Max: runtime.NumCPU() + 1}
+	for tag, oldvariants := range tilelib.variant {
+		tag, oldvariants := tagID(tag), oldvariants
+		if tag%1000000 == 0 {
+			log.Printf("Tidy: tag %d", tag)
+		}
+		throttle.Acquire()
+		go func() {
+			defer throttle.Release()
+			uses := make([]int, len(oldvariants))
+			for _, cg := range tilelib.compactGenomes {
+				for phase := 0; phase < 2; phase++ {
+					cgi := int(tag)*2 + phase
+					if cgi < len(cg) && cg[cgi] > 0 {
+						uses[cg[cgi]-1]++
+					}
+				}
+			}
+
+			// Compute desired order of variants:
+			// neworder[x] == index in oldvariants that
+			// should move to position x.
+			neworder := make([]int, len(oldvariants))
+			for i := range neworder {
+				neworder[i] = i
+			}
+			sort.Slice(neworder, func(i, j int) bool {
+				if cmp := uses[neworder[i]] - uses[neworder[j]]; cmp != 0 {
+					return cmp > 0
+				} else {
+					return bytes.Compare(oldvariants[neworder[i]][:], oldvariants[neworder[j]][:]) < 0
+				}
+			})
+
+			// Replace tilelib.variant[tag] with a new
+			// re-ordered slice of hashes, and make a
+			// mapping from old to new variant IDs.
+			remaptag := make([]tileVariantID, len(oldvariants)+1)
+			newvariants := make([][blake2b.Size256]byte, 0, len(neworder))
+			for _, oldi := range neworder {
+				if uses[oldi] > 0 || inref[tileLibRef{Tag: tag, Variant: tileVariantID(oldi + 1)}] {
+					newvariants = append(newvariants, oldvariants[oldi])
+					remaptag[oldi+1] = tileVariantID(len(newvariants))
+				}
+			}
+			tilelib.variant[tag] = newvariants
+			remap[tag] = remaptag
+		}()
+	}
+	throttle.Wait()
+
+	// Apply remap to genomes and reference sequences, so they
+	// refer to the same tile variants using the changed IDs.
+	log.Print("Tidy: apply remap")
+	for _, cg := range tilelib.compactGenomes {
+		for idx, variant := range cg {
+			cg[idx] = remap[tagID(idx/2)][variant]
+		}
+	}
+	for _, refcs := range tilelib.refseqs {
+		for _, refseq := range refcs {
+			for i, tv := range refseq {
+				refseq[i].Variant = remap[tv.Tag][tv.Variant]
+			}
+		}
+	}
+	log.Print("Tidy: done")
+}
+
+func countBases(seq []byte) int {
+	n := 0
+	for _, c := range seq {
+		if isbase[c] {
+			n++
+		}
+	}
+	return n
+}