slicenumpy.go

   1 // Copyright (C) The Lightning Authors. All rights reserved.
   2 //
   3 // SPDX-License-Identifier: AGPL-3.0
   4
   5 package lightning
   6
   7 import (
   8         "bufio"
   9         "bytes"
  10         "encoding/gob"
  11         "encoding/json"
  12         "errors"
  13         "flag"
  14         "fmt"
  15         "io"
  16         "io/ioutil"
  17         "math"
  18         "net/http"
  19         _ "net/http/pprof"
  20         "os"
  21         "regexp"
  22         "runtime"
  23         "runtime/debug"
  24         "sort"
  25         "strconv"
  26         "strings"
  27         "sync/atomic"
  28         "unsafe"
  29
  30         "git.arvados.org/arvados.git/sdk/go/arvados"
  31         "github.com/arvados/lightning/hgvs"
  32         "github.com/james-bowman/nlp"
  33         "github.com/kshedden/gonpy"
  34         "github.com/sirupsen/logrus"
  35         log "github.com/sirupsen/logrus"
  36         "golang.org/x/crypto/blake2b"
  37         "gonum.org/v1/gonum/mat"
  38 )
  39
  40 const annotationMaxTileSpan = 100
  41
  42 type sliceNumpy struct {
  43         filter          filter
  44         threads         int
  45         chi2Cases       []bool
  46         chi2PValue      float64
  47         pcaComponents   int
  48         minCoverage     int
  49         includeVariant1 bool
  50         debugTag        tagID
  51
  52         cgnames         []string
  53         samples         []sampleInfo
  54         trainingSet     []int // samples index => training set index, or -1 if not in training set
  55         trainingSetSize int
  56         pvalue          func(onehot []bool) float64
  57         pvalueCallCount int64
  58 }
  59
  60 func (cmd *sliceNumpy) RunCommand(prog string, args []string, stdin io.Reader, stdout, stderr io.Writer) int {
  61         err := cmd.run(prog, args, stdin, stdout, stderr)
  62         if err != nil {
  63                 fmt.Fprintf(stderr, "%s\n", err)
  64                 return 1
  65         }
  66         return 0
  67 }
  68
  69 func (cmd *sliceNumpy) run(prog string, args []string, stdin io.Reader, stdout, stderr io.Writer) error {
  70         flags := flag.NewFlagSet("", flag.ContinueOnError)
  71         flags.SetOutput(stderr)
  72         pprof := flags.String("pprof", "", "serve Go profile data at http://`[addr]:port`")
  73         runlocal := flags.Bool("local", false, "run on local host (default: run in an arvados container)")
  74         arvadosRAM := flags.Int("arvados-ram", 750000000000, "amount of memory to request for arvados container (`bytes`)")
  75         arvadosVCPUs := flags.Int("arvados-vcpus", 96, "number of VCPUs to request for arvados container")
  76         projectUUID := flags.String("project", "", "project `UUID` for output data")
  77         priority := flags.Int("priority", 500, "container request priority")
  78         preemptible := flags.Bool("preemptible", true, "request preemptible instance")
  79         inputDir := flags.String("input-dir", "./in", "input `directory`")
  80         outputDir := flags.String("output-dir", "./out", "output `directory`")
  81         ref := flags.String("ref", "", "reference name (if blank, choose last one that appears in input)")
  82         regionsFilename := flags.String("regions", "", "only output columns/annotations that intersect regions in specified bed `file`")
  83         expandRegions := flags.Int("expand-regions", 0, "expand specified regions by `N` base pairs on each side`")
  84         mergeOutput := flags.Bool("merge-output", false, "merge output into one matrix.npy and one matrix.annotations.csv")
  85         hgvsSingle := flags.Bool("single-hgvs-matrix", false, "also generate hgvs-based matrix")
  86         hgvsChunked := flags.Bool("chunked-hgvs-matrix", false, "also generate hgvs-based matrix per chromosome")
  87         onehotSingle := flags.Bool("single-onehot", false, "generate one-hot tile-based matrix")
  88         onehotChunked := flags.Bool("chunked-onehot", false, "generate one-hot tile-based matrix per input chunk")
  89         samplesFilename := flags.String("samples", "", "`samples.csv` file with training/validation and case/control groups (see 'lightning choose-samples')")
  90         caseControlOnly := flags.Bool("case-control-only", false, "drop samples that are not in case/control groups")
  91         onlyPCA := flags.Bool("pca", false, "run principal component analysis, write components to pca.npy and samples.csv")
  92         flags.IntVar(&cmd.pcaComponents, "pca-components", 4, "number of PCA components to compute / use in logistic regression")
  93         maxPCATiles := flags.Int("max-pca-tiles", 0, "maximum tiles to use as PCA input (filter, then drop every 2nd colum pair until below max)")
  94         debugTag := flags.Int("debug-tag", -1, "log debugging details about specified tag")
  95         flags.IntVar(&cmd.threads, "threads", 16, "number of memory-hungry assembly threads, and number of VCPUs to request for arvados container")
  96         flags.Float64Var(&cmd.chi2PValue, "chi2-p-value", 1, "do Χ² test (or logistic regression if -samples file has PCA components) and omit columns with p-value above this threshold")
  97         flags.BoolVar(&cmd.includeVariant1, "include-variant-1", false, "include most common variant when building one-hot matrix")
  98         cmd.filter.Flags(flags)
  99         err := flags.Parse(args)
 100         if err == flag.ErrHelp {
 101                 return nil
 102         } else if err != nil {
 103                 return err
 104         } else if flags.NArg() > 0 {
 105                 return fmt.Errorf("errant command line arguments after parsed flags: %v", flags.Args())
 106         }
 107
 108         if *pprof != "" {
 109                 go func() {
 110                         log.Println(http.ListenAndServe(*pprof, nil))
 111                 }()
 112         }
 113
 114         if cmd.chi2PValue != 1 && *samplesFilename == "" {
 115                 return fmt.Errorf("cannot use provided -chi2-p-value=%f because -samples= value is empty", cmd.chi2PValue)
 116         }
 117
 118         cmd.debugTag = tagID(*debugTag)
 119
 120         if !*runlocal {
 121                 runner := arvadosContainerRunner{
 122                         Name:        "lightning slice-numpy",
 123                         Client:      arvados.NewClientFromEnv(),
 124                         ProjectUUID: *projectUUID,
 125                         RAM:         int64(*arvadosRAM),
 126                         VCPUs:       *arvadosVCPUs,
 127                         Priority:    *priority,
 128                         KeepCache:   2,
 129                         APIAccess:   true,
 130                         Preemptible: *preemptible,
 131                 }
 132                 err = runner.TranslatePaths(inputDir, regionsFilename, samplesFilename)
 133                 if err != nil {
 134                         return err
 135                 }
 136                 runner.Args = []string{"slice-numpy", "-local=true",
 137                         "-pprof=:6060",
 138                         "-input-dir=" + *inputDir,
 139                         "-output-dir=/mnt/output",
 140                         "-threads=" + fmt.Sprintf("%d", cmd.threads),
 141                         "-regions=" + *regionsFilename,
 142                         "-expand-regions=" + fmt.Sprintf("%d", *expandRegions),
 143                         "-merge-output=" + fmt.Sprintf("%v", *mergeOutput),
 144                         "-single-hgvs-matrix=" + fmt.Sprintf("%v", *hgvsSingle),
 145                         "-chunked-hgvs-matrix=" + fmt.Sprintf("%v", *hgvsChunked),
 146                         "-single-onehot=" + fmt.Sprintf("%v", *onehotSingle),
 147                         "-chunked-onehot=" + fmt.Sprintf("%v", *onehotChunked),
 148                         "-samples=" + *samplesFilename,
 149                         "-case-control-only=" + fmt.Sprintf("%v", *caseControlOnly),
 150                         "-pca=" + fmt.Sprintf("%v", *onlyPCA),
 151                         "-pca-components=" + fmt.Sprintf("%d", cmd.pcaComponents),
 152                         "-max-pca-tiles=" + fmt.Sprintf("%d", *maxPCATiles),
 153                         "-chi2-p-value=" + fmt.Sprintf("%f", cmd.chi2PValue),
 154                         "-include-variant-1=" + fmt.Sprintf("%v", cmd.includeVariant1),
 155                         "-debug-tag=" + fmt.Sprintf("%d", cmd.debugTag),
 156                 }
 157                 runner.Args = append(runner.Args, cmd.filter.Args()...)
 158                 var output string
 159                 output, err = runner.Run()
 160                 if err != nil {
 161                         return err
 162                 }
 163                 fmt.Fprintln(stdout, output)
 164                 return nil
 165         }
 166
 167         infiles, err := allFiles(*inputDir, matchGobFile)
 168         if err != nil {
 169                 return err
 170         }
 171         if len(infiles) == 0 {
 172                 err = fmt.Errorf("no input files found in %s", *inputDir)
 173                 return err
 174         }
 175         sort.Strings(infiles)
 176
 177         var refseq map[string][]tileLibRef
 178         var reftiledata = make(map[tileLibRef][]byte, 11000000)
 179         in0, err := open(infiles[0])
 180         if err != nil {
 181                 return err
 182         }
 183
 184         matchGenome, err := regexp.Compile(cmd.filter.MatchGenome)
 185         if err != nil {
 186                 err = fmt.Errorf("-match-genome: invalid regexp: %q", cmd.filter.MatchGenome)
 187                 return err
 188         }
 189
 190         if *samplesFilename != "" {
 191                 cmd.samples, err = loadSampleInfo(*samplesFilename)
 192                 if err != nil {
 193                         return err
 194                 }
 195                 if len(cmd.samples[0].pcaComponents) > 0 {
 196                         cmd.pvalue = glmPvalueFunc(cmd.samples, cmd.pcaComponents)
 197                         // Unfortunately, statsmodel/glm lib logs
 198                         // stuff to os.Stdout when it panics on an
 199                         // unsolvable problem. We recover() from the
 200                         // panic in glm.go, but we also need to
 201                         // commandeer os.Stdout to avoid producing
 202                         // large quantities of logs.
 203                         stdoutWas := os.Stdout
 204                         defer func() { os.Stdout = stdoutWas }()
 205                         os.Stdout, err = os.Open(os.DevNull)
 206                         if err != nil {
 207                                 return err
 208                         }
 209                 }
 210         } else if *caseControlOnly {
 211                 return fmt.Errorf("-case-control-only does not make sense without -samples")
 212         }
 213
 214         cmd.cgnames = nil
 215         var tagset [][]byte
 216         err = DecodeLibrary(in0, strings.HasSuffix(infiles[0], ".gz"), func(ent *LibraryEntry) error {
 217                 if len(ent.TagSet) > 0 {
 218                         tagset = ent.TagSet
 219                 }
 220                 for _, cseq := range ent.CompactSequences {
 221                         if cseq.Name == *ref || *ref == "" {
 222                                 refseq = cseq.TileSequences
 223                         }
 224                 }
 225                 for _, cg := range ent.CompactGenomes {
 226                         if matchGenome.MatchString(cg.Name) {
 227                                 cmd.cgnames = append(cmd.cgnames, cg.Name)
 228                         }
 229                 }
 230                 for _, tv := range ent.TileVariants {
 231                         if tv.Ref {
 232                                 reftiledata[tileLibRef{tv.Tag, tv.Variant}] = tv.Sequence
 233                         }
 234                 }
 235                 return nil
 236         })
 237         if err != nil {
 238                 return err
 239         }
 240         in0.Close()
 241         if refseq == nil {
 242                 err = fmt.Errorf("%s: reference sequence not found", infiles[0])
 243                 return err
 244         }
 245         if len(tagset) == 0 {
 246                 err = fmt.Errorf("tagset not found")
 247                 return err
 248         }
 249
 250         taglib := &tagLibrary{}
 251         err = taglib.setTags(tagset)
 252         if err != nil {
 253                 return err
 254         }
 255         taglen := taglib.TagLen()
 256         sort.Strings(cmd.cgnames)
 257
 258         if len(cmd.cgnames) == 0 {
 259                 return fmt.Errorf("fatal: 0 matching samples in library, nothing to do")
 260         }
 261         cmd.trainingSet = make([]int, len(cmd.cgnames))
 262         if *samplesFilename == "" {
 263                 cmd.trainingSetSize = len(cmd.cgnames)
 264                 for i, name := range cmd.cgnames {
 265                         cmd.samples = append(cmd.samples, sampleInfo{
 266                                 id:         trimFilenameForLabel(name),
 267                                 isTraining: true,
 268                         })
 269                         cmd.trainingSet[i] = i
 270                 }
 271         } else if len(cmd.cgnames) != len(cmd.samples) {
 272                 return fmt.Errorf("mismatched sample list: %d samples in library, %d in %s", len(cmd.cgnames), len(cmd.samples), *samplesFilename)
 273         } else {
 274                 for i, name := range cmd.cgnames {
 275                         if s := trimFilenameForLabel(name); s != cmd.samples[i].id {
 276                                 return fmt.Errorf("mismatched sample list: sample %d is %q in library, %q in %s", i, s, cmd.samples[i].id, *samplesFilename)
 277                         }
 278                 }
 279                 if *caseControlOnly {
 280                         for i := 0; i < len(cmd.samples); i++ {
 281                                 if !cmd.samples[i].isTraining && !cmd.samples[i].isValidation {
 282                                         if i+1 < len(cmd.samples) {
 283                                                 copy(cmd.samples[i:], cmd.samples[i+1:])
 284                                                 copy(cmd.cgnames[i:], cmd.cgnames[i+1:])
 285                                         }
 286                                         cmd.samples = cmd.samples[:len(cmd.samples)-1]
 287                                         cmd.cgnames = cmd.cgnames[:len(cmd.cgnames)-1]
 288                                         i--
 289                                 }
 290                         }
 291                 }
 292                 cmd.chi2Cases = nil
 293                 cmd.trainingSetSize = 0
 294                 for i := range cmd.cgnames {
 295                         if cmd.samples[i].isTraining {
 296                                 cmd.trainingSet[i] = cmd.trainingSetSize
 297                                 cmd.trainingSetSize++
 298                                 cmd.chi2Cases = append(cmd.chi2Cases, cmd.samples[i].isCase)
 299                         } else {
 300                                 cmd.trainingSet[i] = -1
 301                         }
 302                 }
 303                 if cmd.pvalue == nil {
 304                         cmd.pvalue = func(onehot []bool) float64 {
 305                                 return pvalue(onehot, cmd.chi2Cases)
 306                         }
 307                 }
 308         }
 309         if cmd.filter.MinCoverage == 1 {
 310                 // In the generic formula below, floating point
 311                 // arithmetic can effectively push the coverage
 312                 // threshold above 1.0, which is impossible/useless.
 313                 // 1.0 needs to mean exactly 100% coverage.
 314                 cmd.minCoverage = len(cmd.cgnames)
 315         } else {
 316                 cmd.minCoverage = int(math.Ceil(cmd.filter.MinCoverage * float64(len(cmd.cgnames))))
 317         }
 318
 319         // cgnamemap[name]==true for samples that we are including in
 320         // output
 321         cgnamemap := map[string]bool{}
 322         for _, name := range cmd.cgnames {
 323                 cgnamemap[name] = true
 324         }
 325
 326         {
 327                 samplesOutFilename := *outputDir + "/samples.csv"
 328                 log.Infof("writing sample metadata to %s", samplesOutFilename)
 329                 var f *os.File
 330                 f, err = os.Create(samplesOutFilename)
 331                 if err != nil {
 332                         return err
 333                 }
 334                 defer f.Close()
 335                 for i, si := range cmd.samples {
 336                         var cc, tv string
 337                         if si.isCase {
 338                                 cc = "1"
 339                         } else if si.isControl {
 340                                 cc = "0"
 341                         }
 342                         if si.isTraining {
 343                                 tv = "1"
 344                         } else {
 345                                 tv = "0"
 346                         }
 347                         _, err = fmt.Fprintf(f, "%d,%s,%s,%s\n", i, si.id, cc, tv)
 348                         if err != nil {
 349                                 err = fmt.Errorf("write %s: %w", samplesOutFilename, err)
 350                                 return err
 351                         }
 352                 }
 353                 err = f.Close()
 354                 if err != nil {
 355                         err = fmt.Errorf("close %s: %w", samplesOutFilename, err)
 356                         return err
 357                 }
 358                 log.Print("done")
 359         }
 360
 361         log.Info("indexing reference tiles")
 362         type reftileinfo struct {
 363                 variant  tileVariantID
 364                 seqname  string // chr1
 365                 pos      int    // distance from start of chromosome to starttag
 366                 tiledata []byte // acgtggcaa...
 367                 excluded bool   // true if excluded by regions file
 368                 nexttag  tagID  // tagID of following tile (-1 for last tag of chromosome)
 369         }
 370         isdup := map[tagID]bool{}
 371         reftile := map[tagID]*reftileinfo{}
 372         for seqname, cseq := range refseq {
 373                 pos := 0
 374                 lastreftag := tagID(-1)
 375                 for _, libref := range cseq {
 376                         if cmd.filter.MaxTag >= 0 && libref.Tag > tagID(cmd.filter.MaxTag) {
 377                                 continue
 378                         }
 379                         tiledata := reftiledata[libref]
 380                         if len(tiledata) == 0 {
 381                                 err = fmt.Errorf("missing tiledata for tag %d variant %d in %s in ref", libref.Tag, libref.Variant, seqname)
 382                                 return err
 383                         }
 384                         foundthistag := false
 385                         taglib.FindAll(tiledata[:len(tiledata)-1], func(tagid tagID, offset, _ int) {
 386                                 if !foundthistag && tagid == libref.Tag {
 387                                         foundthistag = true
 388                                         return
 389                                 }
 390                                 if dupref, ok := reftile[tagid]; ok {
 391                                         log.Printf("dropping reference tile %+v from %s @ %d, tag not unique, also found inside %+v from %s @ %d", tileLibRef{Tag: tagid, Variant: dupref.variant}, dupref.seqname, dupref.pos, libref, seqname, pos+offset+1)
 392                                         delete(reftile, tagid)
 393                                 } else {
 394                                         log.Printf("found tag %d at offset %d inside tile variant %+v on %s @ %d", tagid, offset, libref, seqname, pos+offset+1)
 395                                 }
 396                                 isdup[tagid] = true
 397                         })
 398                         if isdup[libref.Tag] {
 399                                 log.Printf("dropping reference tile %+v from %s @ %d, tag not unique", libref, seqname, pos)
 400                         } else if reftile[libref.Tag] != nil {
 401                                 log.Printf("dropping reference tile %+v from %s @ %d, tag not unique", tileLibRef{Tag: libref.Tag, Variant: reftile[libref.Tag].variant}, reftile[libref.Tag].seqname, reftile[libref.Tag].pos)
 402                                 delete(reftile, libref.Tag)
 403                                 log.Printf("dropping reference tile %+v from %s @ %d, tag not unique", libref, seqname, pos)
 404                                 isdup[libref.Tag] = true
 405                         } else {
 406                                 reftile[libref.Tag] = &reftileinfo{
 407                                         seqname:  seqname,
 408                                         variant:  libref.Variant,
 409                                         tiledata: tiledata,
 410                                         pos:      pos,
 411                                         nexttag:  -1,
 412                                 }
 413                                 if lastreftag >= 0 {
 414                                         reftile[lastreftag].nexttag = libref.Tag
 415                                 }
 416                                 lastreftag = libref.Tag
 417                         }
 418                         pos += len(tiledata) - taglen
 419                 }
 420                 log.Printf("... %s done, len %d", seqname, pos+taglen)
 421         }
 422
 423         var mask *mask
 424         if *regionsFilename != "" {
 425                 log.Printf("loading regions from %s", *regionsFilename)
 426                 mask, err = makeMask(*regionsFilename, *expandRegions)
 427                 if err != nil {
 428                         return err
 429                 }
 430                 log.Printf("before applying mask, len(reftile) == %d", len(reftile))
 431                 log.Printf("deleting reftile entries for regions outside %d intervals", mask.Len())
 432                 for _, rt := range reftile {
 433                         if !mask.Check(strings.TrimPrefix(rt.seqname, "chr"), rt.pos, rt.pos+len(rt.tiledata)) {
 434                                 rt.excluded = true
 435                         }
 436                 }
 437                 log.Printf("after applying mask, len(reftile) == %d", len(reftile))
 438         }
 439
 440         type hgvsColSet map[hgvs.Variant][2][]int8
 441         encodeHGVS := throttle{Max: len(refseq)}
 442         encodeHGVSTodo := map[string]chan hgvsColSet{}
 443         tmpHGVSCols := map[string]*os.File{}
 444         if *hgvsChunked {
 445                 for seqname := range refseq {
 446                         var f *os.File
 447                         f, err = os.Create(*outputDir + "/tmp." + seqname + ".gob")
 448                         if err != nil {
 449                                 return err
 450                         }
 451                         defer os.Remove(f.Name())
 452                         bufw := bufio.NewWriterSize(f, 1<<24)
 453                         enc := gob.NewEncoder(bufw)
 454                         tmpHGVSCols[seqname] = f
 455                         todo := make(chan hgvsColSet, 128)
 456                         encodeHGVSTodo[seqname] = todo
 457                         encodeHGVS.Go(func() error {
 458                                 for colset := range todo {
 459                                         err := enc.Encode(colset)
 460                                         if err != nil {
 461                                                 encodeHGVS.Report(err)
 462                                                 for range todo {
 463                                                 }
 464                                                 return err
 465                                         }
 466                                 }
 467                                 return bufw.Flush()
 468                         })
 469                 }
 470         }
 471
 472         var toMerge [][]int16
 473         if *mergeOutput || *hgvsSingle {
 474                 toMerge = make([][]int16, len(infiles))
 475         }
 476         var onehotIndirect [][2][]uint32 // [chunkIndex][axis][index]
 477         var onehotChunkSize []uint32
 478         var onehotXrefs [][]onehotXref
 479         if *onehotSingle || *onlyPCA {
 480                 onehotIndirect = make([][2][]uint32, len(infiles))
 481                 onehotChunkSize = make([]uint32, len(infiles))
 482                 onehotXrefs = make([][]onehotXref, len(infiles))
 483         }
 484         chunkStartTag := make([]tagID, len(infiles))
 485
 486         throttleMem := throttle{Max: cmd.threads} // TODO: estimate using mem and data size
 487         throttleNumpyMem := throttle{Max: cmd.threads/2 + 1}
 488         log.Info("generating annotations and numpy matrix for each slice")
 489         var errSkip = errors.New("skip infile")
 490         var done int64
 491         for infileIdx, infile := range infiles {
 492                 infileIdx, infile := infileIdx, infile
 493                 throttleMem.Go(func() error {
 494                         seq := make(map[tagID][]TileVariant, 50000)
 495                         cgs := make(map[string]CompactGenome, len(cmd.cgnames))
 496                         f, err := open(infile)
 497                         if err != nil {
 498                                 return err
 499                         }
 500                         defer f.Close()
 501                         log.Infof("%04d: reading %s", infileIdx, infile)
 502                         err = DecodeLibrary(f, strings.HasSuffix(infile, ".gz"), func(ent *LibraryEntry) error {
 503                                 for _, tv := range ent.TileVariants {
 504                                         if tv.Ref {
 505                                                 continue
 506                                         }
 507                                         // Skip tile with no
 508                                         // corresponding ref tile, if
 509                                         // mask is in play (we can't
 510                                         // determine coordinates for
 511                                         // these)
 512                                         if mask != nil && reftile[tv.Tag] == nil {
 513                                                 continue
 514                                         }
 515                                         // Skip tile whose
 516                                         // corresponding ref tile is
 517                                         // outside target regions --
 518                                         // unless it's a potential
 519                                         // spanning tile.
 520                                         if mask != nil && reftile[tv.Tag].excluded &&
 521                                                 (int(tv.Tag+1) >= len(tagset) ||
 522                                                         (bytes.HasSuffix(tv.Sequence, tagset[tv.Tag+1]) && reftile[tv.Tag+1] != nil && !reftile[tv.Tag+1].excluded)) {
 523                                                 continue
 524                                         }
 525                                         if tv.Tag == cmd.debugTag {
 526                                                 log.Printf("infile %d %s tag %d variant %d hash %x", infileIdx, infile, tv.Tag, tv.Variant, tv.Blake2b[:3])
 527                                         }
 528                                         variants := seq[tv.Tag]
 529                                         if len(variants) == 0 {
 530                                                 variants = make([]TileVariant, 100)
 531                                         }
 532                                         for len(variants) <= int(tv.Variant) {
 533                                                 variants = append(variants, TileVariant{})
 534                                         }
 535                                         variants[int(tv.Variant)] = tv
 536                                         seq[tv.Tag] = variants
 537                                 }
 538                                 for _, cg := range ent.CompactGenomes {
 539                                         if cmd.filter.MaxTag >= 0 && cg.StartTag > tagID(cmd.filter.MaxTag) {
 540                                                 return errSkip
 541                                         }
 542                                         if !cgnamemap[cg.Name] {
 543                                                 continue
 544                                         }
 545                                         // pad to full slice size
 546                                         // to avoid out-of-bounds
 547                                         // checks later
 548                                         if sliceSize := 2 * int(cg.EndTag-cg.StartTag); len(cg.Variants) < sliceSize {
 549                                                 cg.Variants = append(cg.Variants, make([]tileVariantID, sliceSize-len(cg.Variants))...)
 550                                         }
 551                                         cgs[cg.Name] = cg
 552                                 }
 553                                 return nil
 554                         })
 555                         if err == errSkip {
 556                                 return nil
 557                         } else if err != nil {
 558                                 return fmt.Errorf("%04d: DecodeLibrary(%s): err", infileIdx, infile)
 559                         }
 560                         tagstart := cgs[cmd.cgnames[0]].StartTag
 561                         tagend := cgs[cmd.cgnames[0]].EndTag
 562                         chunkStartTag[infileIdx] = tagstart
 563
 564                         // TODO: filters
 565
 566                         log.Infof("%04d: renumber/dedup variants for tags %d-%d", infileIdx, tagstart, tagend)
 567                         variantRemap := make([][]tileVariantID, tagend-tagstart)
 568                         throttleCPU := throttle{Max: runtime.GOMAXPROCS(0)}
 569                         for tag, variants := range seq {
 570                                 tag, variants := tag, variants
 571                                 throttleCPU.Go(func() error {
 572                                         alleleCoverage := 0
 573                                         count := make(map[[blake2b.Size256]byte]int, len(variants))
 574
 575                                         rt := reftile[tag]
 576                                         if rt != nil {
 577                                                 count[blake2b.Sum256(rt.tiledata)] = 0
 578                                         }
 579
 580                                         for cgname, cg := range cgs {
 581                                                 idx := int(tag-tagstart) * 2
 582                                                 for allele := 0; allele < 2; allele++ {
 583                                                         v := cg.Variants[idx+allele]
 584                                                         if v > 0 && len(variants[v].Sequence) > 0 {
 585                                                                 count[variants[v].Blake2b]++
 586                                                                 alleleCoverage++
 587                                                         }
 588                                                         if v > 0 && tag == cmd.debugTag {
 589                                                                 log.Printf("tag %d cg %s allele %d tv %d hash %x count is now %d", tag, cgname, allele, v, variants[v].Blake2b[:3], count[variants[v].Blake2b])
 590                                                         }
 591                                                 }
 592                                         }
 593                                         if alleleCoverage < cmd.minCoverage*2 {
 594                                                 idx := int(tag-tagstart) * 2
 595                                                 for _, cg := range cgs {
 596                                                         cg.Variants[idx] = 0
 597                                                         cg.Variants[idx+1] = 0
 598                                                 }
 599                                                 if tag == cmd.debugTag {
 600                                                         log.Printf("tag %d alleleCoverage %d < min %d, sample data wiped", tag, alleleCoverage, cmd.minCoverage*2)
 601                                                 }
 602                                                 return nil
 603                                         }
 604
 605                                         // hash[i] will be the hash of
 606                                         // the variant(s) that should
 607                                         // be at rank i (0-based).
 608                                         hash := make([][blake2b.Size256]byte, 0, len(count))
 609                                         for b := range count {
 610                                                 hash = append(hash, b)
 611                                         }
 612                                         sort.Slice(hash, func(i, j int) bool {
 613                                                 bi, bj := &hash[i], &hash[j]
 614                                                 if ci, cj := count[*bi], count[*bj]; ci != cj {
 615                                                         return ci > cj
 616                                                 } else {
 617                                                         return bytes.Compare((*bi)[:], (*bj)[:]) < 0
 618                                                 }
 619                                         })
 620                                         // rank[b] will be the 1-based
 621                                         // new variant number for
 622                                         // variants whose hash is b.
 623                                         rank := make(map[[blake2b.Size256]byte]tileVariantID, len(hash))
 624                                         for i, h := range hash {
 625                                                 rank[h] = tileVariantID(i + 1)
 626                                         }
 627                                         if tag == cmd.debugTag {
 628                                                 for h, r := range rank {
 629                                                         log.Printf("tag %d rank(%x) = %v", tag, h[:3], r)
 630                                                 }
 631                                         }
 632                                         // remap[v] will be the new
 633                                         // variant number for original
 634                                         // variant number v.
 635                                         remap := make([]tileVariantID, len(variants))
 636                                         for i, tv := range variants {
 637                                                 remap[i] = rank[tv.Blake2b]
 638                                         }
 639                                         if tag == cmd.debugTag {
 640                                                 for in, out := range remap {
 641                                                         if out > 0 {
 642                                                                 log.Printf("tag %d remap %d => %d", tag, in, out)
 643                                                         }
 644                                                 }
 645                                         }
 646                                         variantRemap[tag-tagstart] = remap
 647                                         if rt != nil {
 648                                                 refrank := rank[blake2b.Sum256(rt.tiledata)]
 649                                                 if tag == cmd.debugTag {
 650                                                         log.Printf("tag %d reftile variant %d => %d", tag, rt.variant, refrank)
 651                                                 }
 652                                                 rt.variant = refrank
 653                                         }
 654                                         return nil
 655                                 })
 656                         }
 657                         throttleCPU.Wait()
 658
 659                         var onehotChunk [][]int8
 660                         var onehotXref []onehotXref
 661
 662                         var annotationsFilename string
 663                         if *onlyPCA {
 664                                 annotationsFilename = "/dev/null"
 665                         } else {
 666                                 annotationsFilename = fmt.Sprintf("%s/matrix.%04d.annotations.csv", *outputDir, infileIdx)
 667                                 log.Infof("%04d: writing %s", infileIdx, annotationsFilename)
 668                         }
 669                         annof, err := os.Create(annotationsFilename)
 670                         if err != nil {
 671                                 return err
 672                         }
 673                         annow := bufio.NewWriterSize(annof, 1<<20)
 674                         outcol := 0
 675                         for tag := tagstart; tag < tagend; tag++ {
 676                                 rt := reftile[tag]
 677                                 if rt == nil && mask != nil {
 678                                         // With no ref tile, we don't
 679                                         // have coordinates to say
 680                                         // this is in the desired
 681                                         // regions -- so it's not.
 682                                         // TODO: handle ref spanning
 683                                         // tile case.
 684                                         continue
 685                                 }
 686                                 if rt != nil && rt.excluded {
 687                                         // TODO: don't skip yet --
 688                                         // first check for spanning
 689                                         // tile variants that
 690                                         // intersect non-excluded ref
 691                                         // tiles.
 692                                         continue
 693                                 }
 694                                 if cmd.filter.MaxTag >= 0 && tag > tagID(cmd.filter.MaxTag) {
 695                                         break
 696                                 }
 697                                 remap := variantRemap[tag-tagstart]
 698                                 if remap == nil {
 699                                         // was not assigned above,
 700                                         // because minCoverage
 701                                         outcol++
 702                                         continue
 703                                 }
 704                                 maxv := tileVariantID(0)
 705                                 for _, v := range remap {
 706                                         if maxv < v {
 707                                                 maxv = v
 708                                         }
 709                                 }
 710                                 if *onehotChunked || *onehotSingle || *onlyPCA {
 711                                         onehot, xrefs := cmd.tv2homhet(cgs, maxv, remap, tag, tagstart, seq)
 712                                         if tag == cmd.debugTag {
 713                                                 log.WithFields(logrus.Fields{
 714                                                         "onehot": onehot,
 715                                                         "xrefs":  xrefs,
 716                                                 }).Info("tv2homhet()")
 717                                         }
 718                                         onehotChunk = append(onehotChunk, onehot...)
 719                                         onehotXref = append(onehotXref, xrefs...)
 720                                 }
 721                                 if *onlyPCA {
 722                                         outcol++
 723                                         continue
 724                                 }
 725                                 if rt == nil {
 726                                         // Reference does not use any
 727                                         // variant of this tile
 728                                         //
 729                                         // TODO: diff against the
 730                                         // relevant portion of the
 731                                         // ref's spanning tile
 732                                         outcol++
 733                                         continue
 734                                 }
 735                                 fmt.Fprintf(annow, "%d,%d,%d,=,%s,%d,,,\n", tag, outcol, rt.variant, rt.seqname, rt.pos)
 736                                 variants := seq[tag]
 737                                 reftilestr := strings.ToUpper(string(rt.tiledata))
 738
 739                                 done := make([]bool, maxv+1)
 740                                 variantDiffs := make([][]hgvs.Variant, maxv+1)
 741                                 for v, tv := range variants {
 742                                         v := remap[v]
 743                                         if v == 0 || v == rt.variant || done[v] {
 744                                                 continue
 745                                         } else {
 746                                                 done[v] = true
 747                                         }
 748                                         if len(tv.Sequence) < taglen {
 749                                                 continue
 750                                         }
 751                                         // if reftilestr doesn't end
 752                                         // in the same tag as tv,
 753                                         // extend reftilestr with
 754                                         // following ref tiles until
 755                                         // it does (up to an arbitrary
 756                                         // sanity-check limit)
 757                                         reftilestr := reftilestr
 758                                         endtagstr := strings.ToUpper(string(tv.Sequence[len(tv.Sequence)-taglen:]))
 759                                         for i, rt := 0, rt; i < annotationMaxTileSpan && !strings.HasSuffix(reftilestr, endtagstr) && rt.nexttag >= 0; i++ {
 760                                                 rt = reftile[rt.nexttag]
 761                                                 if rt == nil {
 762                                                         break
 763                                                 }
 764                                                 reftilestr += strings.ToUpper(string(rt.tiledata[taglen:]))
 765                                         }
 766                                         if mask != nil && !mask.Check(strings.TrimPrefix(rt.seqname, "chr"), rt.pos, rt.pos+len(reftilestr)) {
 767                                                 continue
 768                                         }
 769                                         if !strings.HasSuffix(reftilestr, endtagstr) {
 770                                                 fmt.Fprintf(annow, "%d,%d,%d,,%s,%d,,,\n", tag, outcol, v, rt.seqname, rt.pos)
 771                                                 continue
 772                                         }
 773                                         if lendiff := len(reftilestr) - len(tv.Sequence); lendiff < -1000 || lendiff > 1000 {
 774                                                 fmt.Fprintf(annow, "%d,%d,%d,,%s,%d,,,\n", tag, outcol, v, rt.seqname, rt.pos)
 775                                                 continue
 776                                         }
 777                                         diffs, _ := hgvs.Diff(reftilestr, strings.ToUpper(string(tv.Sequence)), 0)
 778                                         for i := range diffs {
 779                                                 diffs[i].Position += rt.pos
 780                                         }
 781                                         for _, diff := range diffs {
 782                                                 fmt.Fprintf(annow, "%d,%d,%d,%s:g.%s,%s,%d,%s,%s,%s\n", tag, outcol, v, rt.seqname, diff.String(), rt.seqname, diff.Position, diff.Ref, diff.New, diff.Left)
 783                                         }
 784                                         if *hgvsChunked {
 785                                                 variantDiffs[v] = diffs
 786                                         }
 787                                 }
 788                                 if *hgvsChunked {
 789                                         // We can now determine, for each HGVS
 790                                         // variant (diff) in this reftile
 791                                         // region, whether a given genome
 792                                         // phase/allele (1) has the variant, (0) has
 793                                         // =ref or a different variant in that
 794                                         // position, or (-1) is lacking
 795                                         // coverage / couldn't be diffed.
 796                                         hgvsCol := hgvsColSet{}
 797                                         for _, diffs := range variantDiffs {
 798                                                 for _, diff := range diffs {
 799                                                         if _, ok := hgvsCol[diff]; ok {
 800                                                                 continue
 801                                                         }
 802                                                         hgvsCol[diff] = [2][]int8{
 803                                                                 make([]int8, len(cmd.cgnames)),
 804                                                                 make([]int8, len(cmd.cgnames)),
 805                                                         }
 806                                                 }
 807                                         }
 808                                         for row, name := range cmd.cgnames {
 809                                                 variants := cgs[name].Variants[(tag-tagstart)*2:]
 810                                                 for ph := 0; ph < 2; ph++ {
 811                                                         v := variants[ph]
 812                                                         if int(v) >= len(remap) {
 813                                                                 v = 0
 814                                                         } else {
 815                                                                 v = remap[v]
 816                                                         }
 817                                                         if v == rt.variant {
 818                                                                 // hgvsCol[*][ph][row] is already 0
 819                                                         } else if len(variantDiffs[v]) == 0 {
 820                                                                 // lacking coverage / couldn't be diffed
 821                                                                 for _, col := range hgvsCol {
 822                                                                         col[ph][row] = -1
 823                                                                 }
 824                                                         } else {
 825                                                                 for _, diff := range variantDiffs[v] {
 826                                                                         hgvsCol[diff][ph][row] = 1
 827                                                                 }
 828                                                         }
 829                                                 }
 830                                         }
 831                                         for diff, colpair := range hgvsCol {
 832                                                 allele2homhet(colpair)
 833                                                 if !cmd.filterHGVScolpair(colpair) {
 834                                                         delete(hgvsCol, diff)
 835                                                 }
 836                                         }
 837                                         if len(hgvsCol) > 0 {
 838                                                 encodeHGVSTodo[rt.seqname] <- hgvsCol
 839                                         }
 840                                 }
 841                                 outcol++
 842                         }
 843                         err = annow.Flush()
 844                         if err != nil {
 845                                 return err
 846                         }
 847                         err = annof.Close()
 848                         if err != nil {
 849                                 return err
 850                         }
 851
 852                         if *onehotChunked {
 853                                 // transpose onehotChunk[col][row] to numpy[row*ncols+col]
 854                                 rows := len(cmd.cgnames)
 855                                 cols := len(onehotChunk)
 856                                 log.Infof("%04d: preparing onehot numpy (rows=%d, cols=%d, mem=%d)", infileIdx, rows, cols, rows*cols)
 857                                 throttleNumpyMem.Acquire()
 858                                 out := onehotcols2int8(onehotChunk)
 859                                 fnm := fmt.Sprintf("%s/onehot.%04d.npy", *outputDir, infileIdx)
 860                                 err = writeNumpyInt8(fnm, out, rows, cols)
 861                                 if err != nil {
 862                                         return err
 863                                 }
 864                                 fnm = fmt.Sprintf("%s/onehot-columns.%04d.npy", *outputDir, infileIdx)
 865                                 err = writeNumpyInt32(fnm, onehotXref2int32(onehotXref), 4, len(onehotXref))
 866                                 if err != nil {
 867                                         return err
 868                                 }
 869                                 debug.FreeOSMemory()
 870                                 throttleNumpyMem.Release()
 871                         }
 872                         if *onehotSingle || *onlyPCA {
 873                                 onehotIndirect[infileIdx] = onehotChunk2Indirect(onehotChunk)
 874                                 onehotChunkSize[infileIdx] = uint32(len(onehotChunk))
 875                                 onehotXrefs[infileIdx] = onehotXref
 876                                 n := len(onehotIndirect[infileIdx][0])
 877                                 log.Infof("%04d: keeping onehot coordinates in memory (n=%d, mem=%d)", infileIdx, n, n*8*2)
 878                         }
 879                         if !(*onehotSingle || *onehotChunked || *onlyPCA) || *mergeOutput || *hgvsSingle {
 880                                 log.Infof("%04d: preparing numpy (rows=%d, cols=%d)", infileIdx, len(cmd.cgnames), 2*outcol)
 881                                 throttleNumpyMem.Acquire()
 882                                 rows := len(cmd.cgnames)
 883                                 cols := 2 * outcol
 884                                 out := make([]int16, rows*cols)
 885                                 for row, name := range cmd.cgnames {
 886                                         outidx := row * cols
 887                                         for col, v := range cgs[name].Variants {
 888                                                 tag := tagstart + tagID(col/2)
 889                                                 if cmd.filter.MaxTag >= 0 && tag > tagID(cmd.filter.MaxTag) {
 890                                                         break
 891                                                 }
 892                                                 if rt := reftile[tag]; rt == nil || rt.excluded {
 893                                                         continue
 894                                                 }
 895                                                 if v == 0 {
 896                                                         out[outidx] = 0 // tag not found / spanning tile
 897                                                 } else if variants, ok := seq[tag]; ok && int(v) < len(variants) && len(variants[v].Sequence) > 0 {
 898                                                         out[outidx] = int16(variantRemap[tag-tagstart][v])
 899                                                 } else {
 900                                                         out[outidx] = -1 // low quality tile variant
 901                                                 }
 902                                                 if tag == cmd.debugTag {
 903                                                         log.Printf("tag %d row %d col %d outidx %d v %d out %d", tag, row, col, outidx, v, out[outidx])
 904                                                 }
 905                                                 outidx++
 906                                         }
 907                                 }
 908                                 seq = nil
 909                                 cgs = nil
 910                                 debug.FreeOSMemory()
 911                                 throttleNumpyMem.Release()
 912                                 if *mergeOutput || *hgvsSingle {
 913                                         log.Infof("%04d: matrix fragment %d rows x %d cols", infileIdx, rows, cols)
 914                                         toMerge[infileIdx] = out
 915                                 }
 916                                 if !*mergeOutput && !*onehotChunked && !*onehotSingle {
 917                                         fnm := fmt.Sprintf("%s/matrix.%04d.npy", *outputDir, infileIdx)
 918                                         err = writeNumpyInt16(fnm, out, rows, cols)
 919                                         if err != nil {
 920                                                 return err
 921                                         }
 922                                 }
 923                         }
 924                         debug.FreeOSMemory()
 925                         log.Infof("%s: done (%d/%d)", infile, int(atomic.AddInt64(&done, 1)), len(infiles))
 926                         return nil
 927                 })
 928         }
 929         if err = throttleMem.Wait(); err != nil {
 930                 return err
 931         }
 932
 933         if *hgvsChunked {
 934                 log.Info("flushing hgvsCols temp files")
 935                 for seqname := range refseq {
 936                         close(encodeHGVSTodo[seqname])
 937                 }
 938                 err = encodeHGVS.Wait()
 939                 if err != nil {
 940                         return err
 941                 }
 942                 for seqname := range refseq {
 943                         log.Infof("%s: reading hgvsCols from temp file", seqname)
 944                         f := tmpHGVSCols[seqname]
 945                         _, err = f.Seek(0, io.SeekStart)
 946                         if err != nil {
 947                                 return err
 948                         }
 949                         var hgvsCols hgvsColSet
 950                         dec := gob.NewDecoder(bufio.NewReaderSize(f, 1<<24))
 951                         for err == nil {
 952                                 err = dec.Decode(&hgvsCols)
 953                         }
 954                         if err != io.EOF {
 955                                 return err
 956                         }
 957                         log.Infof("%s: sorting %d hgvs variants", seqname, len(hgvsCols))
 958                         variants := make([]hgvs.Variant, 0, len(hgvsCols))
 959                         for v := range hgvsCols {
 960                                 variants = append(variants, v)
 961                         }
 962                         sort.Slice(variants, func(i, j int) bool {
 963                                 vi, vj := &variants[i], &variants[j]
 964                                 if vi.Position != vj.Position {
 965                                         return vi.Position < vj.Position
 966                                 } else if vi.Ref != vj.Ref {
 967                                         return vi.Ref < vj.Ref
 968                                 } else {
 969                                         return vi.New < vj.New
 970                                 }
 971                         })
 972                         rows := len(cmd.cgnames)
 973                         cols := len(variants) * 2
 974                         log.Infof("%s: building hgvs matrix (rows=%d, cols=%d, mem=%d)", seqname, rows, cols, rows*cols)
 975                         out := make([]int8, rows*cols)
 976                         for varIdx, variant := range variants {
 977                                 hgvsCols := hgvsCols[variant]
 978                                 for row := range cmd.cgnames {
 979                                         for ph := 0; ph < 2; ph++ {
 980                                                 out[row*cols+varIdx+ph] = hgvsCols[ph][row]
 981                                         }
 982                                 }
 983                         }
 984                         err = writeNumpyInt8(fmt.Sprintf("%s/hgvs.%s.npy", *outputDir, seqname), out, rows, cols)
 985                         if err != nil {
 986                                 return err
 987                         }
 988                         out = nil
 989
 990                         fnm := fmt.Sprintf("%s/hgvs.%s.annotations.csv", *outputDir, seqname)
 991                         log.Infof("%s: writing hgvs column labels to %s", seqname, fnm)
 992                         var hgvsLabels bytes.Buffer
 993                         for varIdx, variant := range variants {
 994                                 fmt.Fprintf(&hgvsLabels, "%d,%s:g.%s\n", varIdx, seqname, variant.String())
 995                         }
 996                         err = ioutil.WriteFile(fnm, hgvsLabels.Bytes(), 0666)
 997                         if err != nil {
 998                                 return err
 999                         }
1000                 }
1001         }
1002
1003         if *mergeOutput || *hgvsSingle {
1004                 var annow *bufio.Writer
1005                 var annof *os.File
1006                 if *mergeOutput {
1007                         annoFilename := fmt.Sprintf("%s/matrix.annotations.csv", *outputDir)
1008                         annof, err = os.Create(annoFilename)
1009                         if err != nil {
1010                                 return err
1011                         }
1012                         annow = bufio.NewWriterSize(annof, 1<<20)
1013                 }
1014
1015                 rows := len(cmd.cgnames)
1016                 cols := 0
1017                 for _, chunk := range toMerge {
1018                         cols += len(chunk) / rows
1019                 }
1020                 log.Infof("merging output matrix (rows=%d, cols=%d, mem=%d) and annotations", rows, cols, rows*cols*2)
1021                 var out []int16
1022                 if *mergeOutput {
1023                         out = make([]int16, rows*cols)
1024                 }
1025                 hgvsCols := map[string][2][]int16{} // hgvs -> [[g0,g1,g2,...], [g0,g1,g2,...]] (slice of genomes for each phase)
1026                 startcol := 0
1027                 for outIdx, chunk := range toMerge {
1028                         chunkcols := len(chunk) / rows
1029                         if *mergeOutput {
1030                                 for row := 0; row < rows; row++ {
1031                                         copy(out[row*cols+startcol:], chunk[row*chunkcols:(row+1)*chunkcols])
1032                                 }
1033                         }
1034                         toMerge[outIdx] = nil
1035
1036                         annotationsFilename := fmt.Sprintf("%s/matrix.%04d.annotations.csv", *outputDir, outIdx)
1037                         log.Infof("reading %s", annotationsFilename)
1038                         buf, err := os.ReadFile(annotationsFilename)
1039                         if err != nil {
1040                                 return err
1041                         }
1042                         if *mergeOutput {
1043                                 err = os.Remove(annotationsFilename)
1044                                 if err != nil {
1045                                         return err
1046                                 }
1047                         }
1048                         for _, line := range bytes.Split(buf, []byte{'\n'}) {
1049                                 if len(line) == 0 {
1050                                         continue
1051                                 }
1052                                 fields := bytes.SplitN(line, []byte{','}, 9)
1053                                 tag, _ := strconv.Atoi(string(fields[0]))
1054                                 incol, _ := strconv.Atoi(string(fields[1]))
1055                                 tileVariant, _ := strconv.Atoi(string(fields[2]))
1056                                 hgvsID := string(fields[3])
1057                                 seqname := string(fields[4])
1058                                 pos, _ := strconv.Atoi(string(fields[5]))
1059                                 refseq := fields[6]
1060                                 if hgvsID == "" {
1061                                         // Null entry for un-diffable
1062                                         // tile variant
1063                                         continue
1064                                 }
1065                                 if hgvsID == "=" {
1066                                         // Null entry for ref tile
1067                                         continue
1068                                 }
1069                                 if mask != nil && !mask.Check(strings.TrimPrefix(seqname, "chr"), pos, pos+len(refseq)) {
1070                                         // The tile intersects one of
1071                                         // the selected regions, but
1072                                         // this particular HGVS
1073                                         // variant does not.
1074                                         continue
1075                                 }
1076                                 hgvsColPair := hgvsCols[hgvsID]
1077                                 if hgvsColPair[0] == nil {
1078                                         // values in new columns start
1079                                         // out as -1 ("no data yet")
1080                                         // or 0 ("=ref") here, may
1081                                         // change to 1 ("hgvs variant
1082                                         // present") below, either on
1083                                         // this line or a future line.
1084                                         hgvsColPair = [2][]int16{make([]int16, len(cmd.cgnames)), make([]int16, len(cmd.cgnames))}
1085                                         rt, ok := reftile[tagID(tag)]
1086                                         if !ok {
1087                                                 err = fmt.Errorf("bug: seeing annotations for tag %d, but it has no reftile entry", tag)
1088                                                 return err
1089                                         }
1090                                         for ph := 0; ph < 2; ph++ {
1091                                                 for row := 0; row < rows; row++ {
1092                                                         v := chunk[row*chunkcols+incol*2+ph]
1093                                                         if tileVariantID(v) == rt.variant {
1094                                                                 hgvsColPair[ph][row] = 0
1095                                                         } else {
1096                                                                 hgvsColPair[ph][row] = -1
1097                                                         }
1098                                                 }
1099                                         }
1100                                         hgvsCols[hgvsID] = hgvsColPair
1101                                         if annow != nil {
1102                                                 hgvsref := hgvs.Variant{
1103                                                         Position: pos,
1104                                                         Ref:      string(refseq),
1105                                                         New:      string(refseq),
1106                                                 }
1107                                                 fmt.Fprintf(annow, "%d,%d,%d,%s:g.%s,%s,%d,%s,%s,%s\n", tag, incol+startcol/2, rt.variant, seqname, hgvsref.String(), seqname, pos, refseq, refseq, fields[8])
1108                                         }
1109                                 }
1110                                 if annow != nil {
1111                                         fmt.Fprintf(annow, "%d,%d,%d,%s,%s,%d,%s,%s,%s\n", tag, incol+startcol/2, tileVariant, hgvsID, seqname, pos, refseq, fields[7], fields[8])
1112                                 }
1113                                 for ph := 0; ph < 2; ph++ {
1114                                         for row := 0; row < rows; row++ {
1115                                                 v := chunk[row*chunkcols+incol*2+ph]
1116                                                 if int(v) == tileVariant {
1117                                                         hgvsColPair[ph][row] = 1
1118                                                 }
1119                                         }
1120                                 }
1121                         }
1122
1123                         startcol += chunkcols
1124                 }
1125                 if *mergeOutput {
1126                         err = annow.Flush()
1127                         if err != nil {
1128                                 return err
1129                         }
1130                         err = annof.Close()
1131                         if err != nil {
1132                                 return err
1133                         }
1134                         err = writeNumpyInt16(fmt.Sprintf("%s/matrix.npy", *outputDir), out, rows, cols)
1135                         if err != nil {
1136                                 return err
1137                         }
1138                 }
1139                 out = nil
1140
1141                 if *hgvsSingle {
1142                         cols = len(hgvsCols) * 2
1143                         log.Printf("building hgvs-based matrix: %d rows x %d cols", rows, cols)
1144                         out = make([]int16, rows*cols)
1145                         hgvsIDs := make([]string, 0, cols/2)
1146                         for hgvsID := range hgvsCols {
1147                                 hgvsIDs = append(hgvsIDs, hgvsID)
1148                         }
1149                         sort.Strings(hgvsIDs)
1150                         var hgvsLabels bytes.Buffer
1151                         for idx, hgvsID := range hgvsIDs {
1152                                 fmt.Fprintf(&hgvsLabels, "%d,%s\n", idx, hgvsID)
1153                                 for ph := 0; ph < 2; ph++ {
1154                                         hgvscol := hgvsCols[hgvsID][ph]
1155                                         for row, val := range hgvscol {
1156                                                 out[row*cols+idx*2+ph] = val
1157                                         }
1158                                 }
1159                         }
1160                         err = writeNumpyInt16(fmt.Sprintf("%s/hgvs.npy", *outputDir), out, rows, cols)
1161                         if err != nil {
1162                                 return err
1163                         }
1164
1165                         fnm := fmt.Sprintf("%s/hgvs.annotations.csv", *outputDir)
1166                         log.Printf("writing hgvs labels: %s", fnm)
1167                         err = ioutil.WriteFile(fnm, hgvsLabels.Bytes(), 0777)
1168                         if err != nil {
1169                                 return err
1170                         }
1171                 }
1172         }
1173         if *onehotSingle || *onlyPCA {
1174                 nzCount := 0
1175                 for _, part := range onehotIndirect {
1176                         nzCount += len(part[0])
1177                 }
1178                 onehot := make([]uint32, nzCount*2) // [r,r,r,...,c,c,c,...]
1179                 var xrefs []onehotXref
1180                 chunkOffset := uint32(0)
1181                 outcol := 0
1182                 for i, part := range onehotIndirect {
1183                         for i := range part[1] {
1184                                 part[1][i] += chunkOffset
1185                         }
1186                         copy(onehot[outcol:], part[0])
1187                         copy(onehot[outcol+nzCount:], part[1])
1188                         xrefs = append(xrefs, onehotXrefs[i]...)
1189
1190                         outcol += len(part[0])
1191                         chunkOffset += onehotChunkSize[i]
1192
1193                         part[0] = nil
1194                         part[1] = nil
1195                         onehotXrefs[i] = nil
1196                         debug.FreeOSMemory()
1197                 }
1198                 if *onehotSingle {
1199                         fnm := fmt.Sprintf("%s/onehot.npy", *outputDir)
1200                         err = writeNumpyUint32(fnm, onehot, 2, nzCount)
1201                         if err != nil {
1202                                 return err
1203                         }
1204                         fnm = fmt.Sprintf("%s/onehot-columns.npy", *outputDir)
1205                         err = writeNumpyInt32(fnm, onehotXref2int32(xrefs), 5, len(xrefs))
1206                         if err != nil {
1207                                 return err
1208                         }
1209                         fnm = fmt.Sprintf("%s/stats.json", *outputDir)
1210                         j, err := json.Marshal(map[string]interface{}{
1211                                 "pvalueCallCount": cmd.pvalueCallCount,
1212                         })
1213                         if err != nil {
1214                                 return err
1215                         }
1216                         err = os.WriteFile(fnm, j, 0777)
1217                         if err != nil {
1218                                 return err
1219                         }
1220                 }
1221                 if *onlyPCA {
1222                         cols := 0
1223                         for _, c := range onehot[nzCount:] {
1224                                 if int(c) >= cols {
1225                                         cols = int(c) + 1
1226                                 }
1227                         }
1228                         if cols == 0 {
1229                                 return fmt.Errorf("cannot do PCA: one-hot matrix is empty")
1230                         }
1231                         log.Printf("have %d one-hot cols", cols)
1232                         stride := 1
1233                         for *maxPCATiles > 0 && cols > *maxPCATiles*2 {
1234                                 cols = (cols + 1) / 2
1235                                 stride = stride * 2
1236                         }
1237                         if cols%2 == 1 {
1238                                 // we work with pairs of columns
1239                                 cols++
1240                         }
1241                         log.Printf("creating full matrix (%d rows) and training matrix (%d rows) with %d cols, stride %d", len(cmd.cgnames), cmd.trainingSetSize, cols, stride)
1242                         mtxFull := mat.NewDense(len(cmd.cgnames), cols, nil)
1243                         mtxTrain := mat.NewDense(cmd.trainingSetSize, cols, nil)
1244                         for i, c := range onehot[nzCount:] {
1245                                 if int(c/2)%stride == 0 {
1246                                         outcol := int(c/2)/stride*2 + int(c)%2
1247                                         mtxFull.Set(int(onehot[i]), outcol, 1)
1248                                         if trainRow := cmd.trainingSet[int(onehot[i])]; trainRow >= 0 {
1249                                                 mtxTrain.Set(trainRow, outcol, 1)
1250                                         }
1251                                 }
1252                         }
1253                         log.Print("fitting")
1254                         transformer := nlp.NewPCA(cmd.pcaComponents)
1255                         transformer.Fit(mtxTrain.T())
1256                         log.Printf("transforming")
1257                         pca, err := transformer.Transform(mtxFull.T())
1258                         if err != nil {
1259                                 return err
1260                         }
1261                         pca = pca.T()
1262                         outrows, outcols := pca.Dims()
1263                         log.Printf("copying result to numpy output array: %d rows, %d cols", outrows, outcols)
1264                         out := make([]float64, outrows*outcols)
1265                         for i := 0; i < outrows; i++ {
1266                                 for j := 0; j < outcols; j++ {
1267                                         out[i*outcols+j] = pca.At(i, j)
1268                                 }
1269                         }
1270                         fnm := fmt.Sprintf("%s/pca.npy", *outputDir)
1271                         log.Printf("writing numpy: %s", fnm)
1272                         output, err := os.OpenFile(fnm, os.O_CREATE|os.O_TRUNC|os.O_WRONLY, 0777)
1273                         if err != nil {
1274                                 return err
1275                         }
1276                         npw, err := gonpy.NewWriter(nopCloser{output})
1277                         if err != nil {
1278                                 return fmt.Errorf("gonpy.NewWriter: %w", err)
1279                         }
1280                         npw.Shape = []int{outrows, outcols}
1281                         err = npw.WriteFloat64(out)
1282                         if err != nil {
1283                                 return fmt.Errorf("WriteFloat64: %w", err)
1284                         }
1285                         err = output.Close()
1286                         if err != nil {
1287                                 return err
1288                         }
1289                         log.Print("done")
1290
1291                         samplesOutFilename := *outputDir + "/samples.csv"
1292                         log.Infof("writing sample metadata to %s", samplesOutFilename)
1293                         var f *os.File
1294                         f, err = os.Create(samplesOutFilename)
1295                         if err != nil {
1296                                 return err
1297                         }
1298                         defer f.Close()
1299                         pcaLabels := ""
1300                         for i := 0; i < outcols; i++ {
1301                                 pcaLabels += fmt.Sprintf(",PCA%d", i)
1302                         }
1303                         _, err = fmt.Fprintf(f, "Index,SampleID,CaseControl,TrainingValidation%s\n", pcaLabels)
1304                         if err != nil {
1305                                 return err
1306                         }
1307                         for i, si := range cmd.samples {
1308                                 var cc, tv string
1309                                 if si.isCase {
1310                                         cc = "1"
1311                                 } else if si.isControl {
1312                                         cc = "0"
1313                                 }
1314                                 if si.isTraining {
1315                                         tv = "1"
1316                                 } else if si.isValidation {
1317                                         tv = "0"
1318                                 }
1319                                 var pcavals string
1320                                 for c := 0; c < outcols; c++ {
1321                                         pcavals += fmt.Sprintf(",%f", pca.At(i, c))
1322                                 }
1323                                 _, err = fmt.Fprintf(f, "%d,%s,%s,%s%s\n", i, si.id, cc, tv, pcavals)
1324                                 if err != nil {
1325                                         err = fmt.Errorf("write %s: %w", samplesOutFilename, err)
1326                                         return err
1327                                 }
1328                         }
1329                         err = f.Close()
1330                         if err != nil {
1331                                 err = fmt.Errorf("close %s: %w", samplesOutFilename, err)
1332                                 return err
1333                         }
1334                         log.Print("done")
1335                 }
1336         }
1337         if !*mergeOutput && !*onehotChunked && !*onehotSingle && !*onlyPCA {
1338                 tagoffsetFilename := *outputDir + "/chunk-tag-offset.csv"
1339                 log.Infof("writing tag offsets to %s", tagoffsetFilename)
1340                 var f *os.File
1341                 f, err = os.Create(tagoffsetFilename)
1342                 if err != nil {
1343                         return err
1344                 }
1345                 defer f.Close()
1346                 for idx, offset := range chunkStartTag {
1347                         _, err = fmt.Fprintf(f, "%q,%d\n", fmt.Sprintf("matrix.%04d.npy", idx), offset)
1348                         if err != nil {
1349                                 err = fmt.Errorf("write %s: %w", tagoffsetFilename, err)
1350                                 return err
1351                         }
1352                 }
1353                 err = f.Close()
1354                 if err != nil {
1355                         err = fmt.Errorf("close %s: %w", tagoffsetFilename, err)
1356                         return err
1357                 }
1358         }
1359
1360         return nil
1361 }
1362
1363 type sampleInfo struct {
1364         id            string
1365         isCase        bool
1366         isControl     bool
1367         isTraining    bool
1368         isValidation  bool
1369         pcaComponents []float64
1370 }
1371
1372 // Read samples.csv file with case/control and training/validation
1373 // flags.
1374 func loadSampleInfo(samplesFilename string) ([]sampleInfo, error) {
1375         var si []sampleInfo
1376         f, err := open(samplesFilename)
1377         if err != nil {
1378                 return nil, err
1379         }
1380         buf, err := io.ReadAll(f)
1381         f.Close()
1382         if err != nil {
1383                 return nil, err
1384         }
1385         lineNum := 0
1386         for _, csv := range bytes.Split(buf, []byte{'\n'}) {
1387                 lineNum++
1388                 if len(csv) == 0 {
1389                         continue
1390                 }
1391                 split := strings.Split(string(csv), ",")
1392                 if len(split) < 4 {
1393                         return nil, fmt.Errorf("%d fields < 4 in %s line %d: %q", len(split), samplesFilename, lineNum, csv)
1394                 }
1395                 if split[0] == "Index" && split[1] == "SampleID" && split[2] == "CaseControl" && split[3] == "TrainingValidation" {
1396                         continue
1397                 }
1398                 idx, err := strconv.Atoi(split[0])
1399                 if err != nil {
1400                         if lineNum == 1 {
1401                                 return nil, fmt.Errorf("header does not look right: %q", csv)
1402                         }
1403                         return nil, fmt.Errorf("%s line %d: index: %s", samplesFilename, lineNum, err)
1404                 }
1405                 if idx != len(si) {
1406                         return nil, fmt.Errorf("%s line %d: index %d out of order", samplesFilename, lineNum, idx)
1407                 }
1408                 var pcaComponents []float64
1409                 if len(split) > 4 {
1410                         for _, s := range split[4:] {
1411                                 f, err := strconv.ParseFloat(s, 64)
1412                                 if err != nil {
1413                                         return nil, fmt.Errorf("%s line %d: cannot parse float %q: %s", samplesFilename, lineNum, s, err)
1414                                 }
1415                                 pcaComponents = append(pcaComponents, f)
1416                         }
1417                 }
1418                 si = append(si, sampleInfo{
1419                         id:            split[1],
1420                         isCase:        split[2] == "1",
1421                         isControl:     split[2] == "0",
1422                         isTraining:    split[3] == "1",
1423                         isValidation:  split[3] == "0",
1424                         pcaComponents: pcaComponents,
1425                 })
1426         }
1427         return si, nil
1428 }
1429
1430 func (cmd *sliceNumpy) filterHGVScolpair(colpair [2][]int8) bool {
1431         if cmd.chi2PValue >= 1 {
1432                 return true
1433         }
1434         col0 := make([]bool, 0, len(cmd.chi2Cases))
1435         col1 := make([]bool, 0, len(cmd.chi2Cases))
1436         cases := make([]bool, 0, len(cmd.chi2Cases))
1437         for i, c := range cmd.chi2Cases {
1438                 if colpair[0][i] < 0 {
1439                         continue
1440                 }
1441                 col0 = append(col0, colpair[0][i] != 0)
1442                 col1 = append(col1, colpair[1][i] != 0)
1443                 cases = append(cases, c)
1444         }
1445         return len(cases) >= cmd.minCoverage &&
1446                 (pvalue(col0, cases) <= cmd.chi2PValue || pvalue(col1, cases) <= cmd.chi2PValue)
1447 }
1448
1449 func writeNumpyUint32(fnm string, out []uint32, rows, cols int) error {
1450         output, err := os.Create(fnm)
1451         if err != nil {
1452                 return err
1453         }
1454         defer output.Close()
1455         bufw := bufio.NewWriterSize(output, 1<<26)
1456         npw, err := gonpy.NewWriter(nopCloser{bufw})
1457         if err != nil {
1458                 return err
1459         }
1460         log.WithFields(log.Fields{
1461                 "filename": fnm,
1462                 "rows":     rows,
1463                 "cols":     cols,
1464                 "bytes":    rows * cols * 4,
1465         }).Infof("writing numpy: %s", fnm)
1466         npw.Shape = []int{rows, cols}
1467         npw.WriteUint32(out)
1468         err = bufw.Flush()
1469         if err != nil {
1470                 return err
1471         }
1472         return output.Close()
1473 }
1474
1475 func writeNumpyInt32(fnm string, out []int32, rows, cols int) error {
1476         output, err := os.Create(fnm)
1477         if err != nil {
1478                 return err
1479         }
1480         defer output.Close()
1481         bufw := bufio.NewWriterSize(output, 1<<26)
1482         npw, err := gonpy.NewWriter(nopCloser{bufw})
1483         if err != nil {
1484                 return err
1485         }
1486         log.WithFields(log.Fields{
1487                 "filename": fnm,
1488                 "rows":     rows,
1489                 "cols":     cols,
1490                 "bytes":    rows * cols * 4,
1491         }).Infof("writing numpy: %s", fnm)
1492         npw.Shape = []int{rows, cols}
1493         npw.WriteInt32(out)
1494         err = bufw.Flush()
1495         if err != nil {
1496                 return err
1497         }
1498         return output.Close()
1499 }
1500
1501 func writeNumpyInt16(fnm string, out []int16, rows, cols int) error {
1502         output, err := os.Create(fnm)
1503         if err != nil {
1504                 return err
1505         }
1506         defer output.Close()
1507         bufw := bufio.NewWriterSize(output, 1<<26)
1508         npw, err := gonpy.NewWriter(nopCloser{bufw})
1509         if err != nil {
1510                 return err
1511         }
1512         log.WithFields(log.Fields{
1513                 "filename": fnm,
1514                 "rows":     rows,
1515                 "cols":     cols,
1516                 "bytes":    rows * cols * 2,
1517         }).Infof("writing numpy: %s", fnm)
1518         npw.Shape = []int{rows, cols}
1519         npw.WriteInt16(out)
1520         err = bufw.Flush()
1521         if err != nil {
1522                 return err
1523         }
1524         return output.Close()
1525 }
1526
1527 func writeNumpyInt8(fnm string, out []int8, rows, cols int) error {
1528         output, err := os.Create(fnm)
1529         if err != nil {
1530                 return err
1531         }
1532         defer output.Close()
1533         bufw := bufio.NewWriterSize(output, 1<<26)
1534         npw, err := gonpy.NewWriter(nopCloser{bufw})
1535         if err != nil {
1536                 return err
1537         }
1538         log.WithFields(log.Fields{
1539                 "filename": fnm,
1540                 "rows":     rows,
1541                 "cols":     cols,
1542                 "bytes":    rows * cols,
1543         }).Infof("writing numpy: %s", fnm)
1544         npw.Shape = []int{rows, cols}
1545         npw.WriteInt8(out)
1546         err = bufw.Flush()
1547         if err != nil {
1548                 return err
1549         }
1550         return output.Close()
1551 }
1552
1553 func allele2homhet(colpair [2][]int8) {
1554         a, b := colpair[0], colpair[1]
1555         for i, av := range a {
1556                 bv := b[i]
1557                 if av < 0 || bv < 0 {
1558                         // no-call
1559                         a[i], b[i] = -1, -1
1560                 } else if av > 0 && bv > 0 {
1561                         // hom
1562                         a[i], b[i] = 1, 0
1563                 } else if av > 0 || bv > 0 {
1564                         // het
1565                         a[i], b[i] = 0, 1
1566                 } else {
1567                         // ref (or a different variant in same position)
1568                         // (this is a no-op) a[i], b[i] = 0, 0
1569                 }
1570         }
1571 }
1572
1573 type onehotXref struct {
1574         tag     tagID
1575         variant tileVariantID
1576         hom     bool
1577         pvalue  float64
1578 }
1579
1580 const onehotXrefSize = unsafe.Sizeof(onehotXref{})
1581
1582 // Build onehot matrix (m[tileVariantIndex][genome] == 0 or 1) for all
1583 // variants of a single tile/tag#.
1584 //
1585 // Return nil if no tile variant passes Χ² filter.
1586 func (cmd *sliceNumpy) tv2homhet(cgs map[string]CompactGenome, maxv tileVariantID, remap []tileVariantID, tag, chunkstarttag tagID, seq map[tagID][]TileVariant) ([][]int8, []onehotXref) {
1587         if tag == cmd.debugTag {
1588                 tv := make([]tileVariantID, len(cmd.cgnames)*2)
1589                 for i, name := range cmd.cgnames {
1590                         copy(tv[i*2:(i+1)*2], cgs[name].Variants[(tag-chunkstarttag)*2:])
1591                 }
1592                 log.WithFields(logrus.Fields{
1593                         "cgs[i].Variants[tag*2+j]": tv,
1594                         "maxv":                     maxv,
1595                         "remap":                    remap,
1596                         "tag":                      tag,
1597                         "chunkstarttag":            chunkstarttag,
1598                 }).Info("tv2homhet()")
1599         }
1600         if maxv < 1 || (maxv < 2 && !cmd.includeVariant1) {
1601                 // everyone has the most common variant (of the variants we don't drop)
1602                 return nil, nil
1603         }
1604         tagoffset := tag - chunkstarttag
1605         coverage := 0
1606         for _, cg := range cgs {
1607                 alleles := 0
1608                 for _, v := range cg.Variants[tagoffset*2 : tagoffset*2+2] {
1609                         if v > 0 && int(v) < len(seq[tag]) && len(seq[tag][v].Sequence) > 0 {
1610                                 alleles++
1611                         }
1612                 }
1613                 if alleles == 2 {
1614                         coverage++
1615                 }
1616         }
1617         if coverage < cmd.minCoverage {
1618                 return nil, nil
1619         }
1620         // "observed" array for p-value calculation (training set
1621         // only)
1622         obs := make([][]bool, (maxv+1)*2) // 2 slices (hom + het) for each variant#
1623         // one-hot output (all samples)
1624         outcols := make([][]int8, (maxv+1)*2)
1625         for i := range obs {
1626                 obs[i] = make([]bool, cmd.trainingSetSize)
1627                 outcols[i] = make([]int8, len(cmd.cgnames))
1628         }
1629         for cgid, name := range cmd.cgnames {
1630                 tsid := cmd.trainingSet[cgid]
1631                 cgvars := cgs[name].Variants[tagoffset*2:]
1632                 tv0, tv1 := remap[cgvars[0]], remap[cgvars[1]]
1633                 for v := tileVariantID(1); v <= maxv; v++ {
1634                         if tv0 == v && tv1 == v {
1635                                 if tsid >= 0 {
1636                                         obs[v*2][tsid] = true
1637                                 }
1638                                 outcols[v*2][cgid] = 1
1639                         } else if tv0 == v || tv1 == v {
1640                                 if tsid >= 0 {
1641                                         obs[v*2+1][tsid] = true
1642                                 }
1643                                 outcols[v*2+1][cgid] = 1
1644                         }
1645                 }
1646         }
1647         var onehot [][]int8
1648         var xref []onehotXref
1649         for col := 2; col < len(obs); col++ {
1650                 // col 0,1 correspond to tile variant 0, i.e.,
1651                 // no-call; col 2,3 correspond to the most common
1652                 // variant; so we (normally) start at col 4.
1653                 if col < 4 && !cmd.includeVariant1 {
1654                         continue
1655                 }
1656                 atomic.AddInt64(&cmd.pvalueCallCount, 1)
1657                 p := cmd.pvalue(obs[col])
1658                 if cmd.chi2PValue < 1 && !(p < cmd.chi2PValue) {
1659                         continue
1660                 }
1661                 onehot = append(onehot, outcols[col])
1662                 xref = append(xref, onehotXref{
1663                         tag:     tag,
1664                         variant: tileVariantID(col >> 1),
1665                         hom:     col&1 == 0,
1666                         pvalue:  p,
1667                 })
1668         }
1669         return onehot, xref
1670 }
1671
1672 // convert a []onehotXref with length N to a numpy-style []int32
1673 // matrix with N columns, one row per field of onehotXref struct.
1674 //
1675 // Hom/het row contains hom=0, het=1.
1676 //
1677 // P-value row contains 1000000x actual p-value.
1678 func onehotXref2int32(xrefs []onehotXref) []int32 {
1679         xcols := len(xrefs)
1680         xdata := make([]int32, 5*xcols)
1681         for i, xref := range xrefs {
1682                 xdata[i] = int32(xref.tag)
1683                 xdata[xcols+i] = int32(xref.variant)
1684                 if xref.hom {
1685                         xdata[xcols*2+i] = 1
1686                 }
1687                 xdata[xcols*3+i] = int32(xref.pvalue * 1000000)
1688                 xdata[xcols*4+i] = int32(-math.Log10(xref.pvalue) * 1000000)
1689         }
1690         return xdata
1691 }
1692
1693 // transpose onehot data from in[col][row] to numpy-style
1694 // out[row*cols+col].
1695 func onehotcols2int8(in [][]int8) []int8 {
1696         if len(in) == 0 {
1697                 return nil
1698         }
1699         cols := len(in)
1700         rows := len(in[0])
1701         out := make([]int8, rows*cols)
1702         for row := 0; row < rows; row++ {
1703                 outrow := out[row*cols:]
1704                 for col, incol := range in {
1705                         outrow[col] = incol[row]
1706                 }
1707         }
1708         return out
1709 }
1710
1711 // Return [2][]uint32{rowIndices, colIndices} indicating which
1712 // elements of matrixT[c][r] have non-zero values.
1713 func onehotChunk2Indirect(matrixT [][]int8) [2][]uint32 {
1714         var nz [2][]uint32
1715         for c, col := range matrixT {
1716                 for r, val := range col {
1717                         if val != 0 {
1718                                 nz[0] = append(nz[0], uint32(r))
1719                                 nz[1] = append(nz[1], uint32(c))
1720                         }
1721                 }
1722         }
1723         return nz
1724 }