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         pvalueMinFrequency float64
  48         pcaComponents      int
  49         minCoverage        int
  50         includeVariant1    bool
  51         debugTag           tagID
  52
  53         cgnames         []string
  54         samples         []sampleInfo
  55         trainingSet     []int // samples index => training set index, or -1 if not in training set
  56         trainingSetSize int
  57         pvalue          func(onehot []bool) float64
  58         pvalueCallCount int64
  59 }
  60
  61 func (cmd *sliceNumpy) RunCommand(prog string, args []string, stdin io.Reader, stdout, stderr io.Writer) int {
  62         err := cmd.run(prog, args, stdin, stdout, stderr)
  63         if err != nil {
  64                 fmt.Fprintf(stderr, "%s\n", err)
  65                 return 1
  66         }
  67         return 0
  68 }
  69
  70 func (cmd *sliceNumpy) run(prog string, args []string, stdin io.Reader, stdout, stderr io.Writer) error {
  71         flags := flag.NewFlagSet("", flag.ContinueOnError)
  72         flags.SetOutput(stderr)
  73         pprof := flags.String("pprof", "", "serve Go profile data at http://`[addr]:port`")
  74         runlocal := flags.Bool("local", false, "run on local host (default: run in an arvados container)")
  75         arvadosRAM := flags.Int("arvados-ram", 750000000000, "amount of memory to request for arvados container (`bytes`)")
  76         arvadosVCPUs := flags.Int("arvados-vcpus", 96, "number of VCPUs to request for arvados container")
  77         projectUUID := flags.String("project", "", "project `UUID` for output data")
  78         priority := flags.Int("priority", 500, "container request priority")
  79         preemptible := flags.Bool("preemptible", true, "request preemptible instance")
  80         inputDir := flags.String("input-dir", "./in", "input `directory`")
  81         outputDir := flags.String("output-dir", "./out", "output `directory`")
  82         ref := flags.String("ref", "", "reference name (if blank, choose last one that appears in input)")
  83         regionsFilename := flags.String("regions", "", "only output columns/annotations that intersect regions in specified bed `file`")
  84         expandRegions := flags.Int("expand-regions", 0, "expand specified regions by `N` base pairs on each side`")
  85         mergeOutput := flags.Bool("merge-output", false, "merge output into one matrix.npy and one matrix.annotations.csv")
  86         hgvsSingle := flags.Bool("single-hgvs-matrix", false, "also generate hgvs-based matrix")
  87         hgvsChunked := flags.Bool("chunked-hgvs-matrix", false, "also generate hgvs-based matrix per chromosome")
  88         onehotSingle := flags.Bool("single-onehot", false, "generate one-hot tile-based matrix")
  89         onehotChunked := flags.Bool("chunked-onehot", false, "generate one-hot tile-based matrix per input chunk")
  90         samplesFilename := flags.String("samples", "", "`samples.csv` file with training/validation and case/control groups (see 'lightning choose-samples')")
  91         caseControlOnly := flags.Bool("case-control-only", false, "drop samples that are not in case/control groups")
  92         onlyPCA := flags.Bool("pca", false, "run principal component analysis, write components to pca.npy and samples.csv")
  93         flags.IntVar(&cmd.pcaComponents, "pca-components", 4, "number of PCA components to compute / use in logistic regression")
  94         maxPCATiles := flags.Int("max-pca-tiles", 0, "maximum tiles to use as PCA input (filter, then drop every 2nd colum pair until below max)")
  95         debugTag := flags.Int("debug-tag", -1, "log debugging details about specified tag")
  96         flags.IntVar(&cmd.threads, "threads", 16, "number of memory-hungry assembly threads, and number of VCPUs to request for arvados container")
  97         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")
  98         flags.Float64Var(&cmd.pvalueMinFrequency, "pvalue-min-frequency", 0.01, "skip p-value calculation on tile variants below this frequency in the training set")
  99         flags.BoolVar(&cmd.includeVariant1, "include-variant-1", false, "include most common variant when building one-hot matrix")
 100         cmd.filter.Flags(flags)
 101         err := flags.Parse(args)
 102         if err == flag.ErrHelp {
 103                 return nil
 104         } else if err != nil {
 105                 return err
 106         } else if flags.NArg() > 0 {
 107                 return fmt.Errorf("errant command line arguments after parsed flags: %v", flags.Args())
 108         }
 109
 110         if *pprof != "" {
 111                 go func() {
 112                         log.Println(http.ListenAndServe(*pprof, nil))
 113                 }()
 114         }
 115
 116         if cmd.chi2PValue != 1 && *samplesFilename == "" {
 117                 return fmt.Errorf("cannot use provided -chi2-p-value=%f because -samples= value is empty", cmd.chi2PValue)
 118         }
 119
 120         cmd.debugTag = tagID(*debugTag)
 121
 122         if !*runlocal {
 123                 runner := arvadosContainerRunner{
 124                         Name:        "lightning slice-numpy",
 125                         Client:      arvados.NewClientFromEnv(),
 126                         ProjectUUID: *projectUUID,
 127                         RAM:         int64(*arvadosRAM),
 128                         VCPUs:       *arvadosVCPUs,
 129                         Priority:    *priority,
 130                         KeepCache:   2,
 131                         APIAccess:   true,
 132                         Preemptible: *preemptible,
 133                 }
 134                 err = runner.TranslatePaths(inputDir, regionsFilename, samplesFilename)
 135                 if err != nil {
 136                         return err
 137                 }
 138                 runner.Args = []string{"slice-numpy", "-local=true",
 139                         "-pprof=:6060",
 140                         "-input-dir=" + *inputDir,
 141                         "-output-dir=/mnt/output",
 142                         "-threads=" + fmt.Sprintf("%d", cmd.threads),
 143                         "-regions=" + *regionsFilename,
 144                         "-expand-regions=" + fmt.Sprintf("%d", *expandRegions),
 145                         "-merge-output=" + fmt.Sprintf("%v", *mergeOutput),
 146                         "-single-hgvs-matrix=" + fmt.Sprintf("%v", *hgvsSingle),
 147                         "-chunked-hgvs-matrix=" + fmt.Sprintf("%v", *hgvsChunked),
 148                         "-single-onehot=" + fmt.Sprintf("%v", *onehotSingle),
 149                         "-chunked-onehot=" + fmt.Sprintf("%v", *onehotChunked),
 150                         "-samples=" + *samplesFilename,
 151                         "-case-control-only=" + fmt.Sprintf("%v", *caseControlOnly),
 152                         "-pca=" + fmt.Sprintf("%v", *onlyPCA),
 153                         "-pca-components=" + fmt.Sprintf("%d", cmd.pcaComponents),
 154                         "-max-pca-tiles=" + fmt.Sprintf("%d", *maxPCATiles),
 155                         "-chi2-p-value=" + fmt.Sprintf("%f", cmd.chi2PValue),
 156                         "-pvalue-min-frequency=" + fmt.Sprintf("%f", cmd.pvalueMinFrequency),
 157                         "-include-variant-1=" + fmt.Sprintf("%v", cmd.includeVariant1),
 158                         "-debug-tag=" + fmt.Sprintf("%d", cmd.debugTag),
 159                 }
 160                 runner.Args = append(runner.Args, cmd.filter.Args()...)
 161                 var output string
 162                 output, err = runner.Run()
 163                 if err != nil {
 164                         return err
 165                 }
 166                 fmt.Fprintln(stdout, output)
 167                 return nil
 168         }
 169
 170         infiles, err := allFiles(*inputDir, matchGobFile)
 171         if err != nil {
 172                 return err
 173         }
 174         if len(infiles) == 0 {
 175                 err = fmt.Errorf("no input files found in %s", *inputDir)
 176                 return err
 177         }
 178         sort.Strings(infiles)
 179
 180         var refseq map[string][]tileLibRef
 181         var reftiledata = make(map[tileLibRef][]byte, 11000000)
 182         in0, err := open(infiles[0])
 183         if err != nil {
 184                 return err
 185         }
 186
 187         matchGenome, err := regexp.Compile(cmd.filter.MatchGenome)
 188         if err != nil {
 189                 err = fmt.Errorf("-match-genome: invalid regexp: %q", cmd.filter.MatchGenome)
 190                 return err
 191         }
 192
 193         if *samplesFilename != "" {
 194                 cmd.samples, err = loadSampleInfo(*samplesFilename)
 195                 if err != nil {
 196                         return err
 197                 }
 198         } else if *caseControlOnly {
 199                 return fmt.Errorf("-case-control-only does not make sense without -samples")
 200         }
 201
 202         cmd.cgnames = nil
 203         var tagset [][]byte
 204         err = DecodeLibrary(in0, strings.HasSuffix(infiles[0], ".gz"), func(ent *LibraryEntry) error {
 205                 if len(ent.TagSet) > 0 {
 206                         tagset = ent.TagSet
 207                 }
 208                 for _, cseq := range ent.CompactSequences {
 209                         if cseq.Name == *ref || *ref == "" {
 210                                 refseq = cseq.TileSequences
 211                         }
 212                 }
 213                 for _, cg := range ent.CompactGenomes {
 214                         if matchGenome.MatchString(cg.Name) {
 215                                 cmd.cgnames = append(cmd.cgnames, cg.Name)
 216                         }
 217                 }
 218                 for _, tv := range ent.TileVariants {
 219                         if tv.Ref {
 220                                 reftiledata[tileLibRef{tv.Tag, tv.Variant}] = tv.Sequence
 221                         }
 222                 }
 223                 return nil
 224         })
 225         if err != nil {
 226                 return err
 227         }
 228         in0.Close()
 229         if refseq == nil {
 230                 err = fmt.Errorf("%s: reference sequence not found", infiles[0])
 231                 return err
 232         }
 233         if len(tagset) == 0 {
 234                 err = fmt.Errorf("tagset not found")
 235                 return err
 236         }
 237
 238         taglib := &tagLibrary{}
 239         err = taglib.setTags(tagset)
 240         if err != nil {
 241                 return err
 242         }
 243         taglen := taglib.TagLen()
 244         sort.Strings(cmd.cgnames)
 245
 246         if len(cmd.cgnames) == 0 {
 247                 return fmt.Errorf("fatal: 0 matching samples in library, nothing to do")
 248         }
 249         cmd.trainingSet = make([]int, len(cmd.cgnames))
 250         if *samplesFilename == "" {
 251                 cmd.trainingSetSize = len(cmd.cgnames)
 252                 for i, name := range cmd.cgnames {
 253                         cmd.samples = append(cmd.samples, sampleInfo{
 254                                 id:         trimFilenameForLabel(name),
 255                                 isTraining: true,
 256                         })
 257                         cmd.trainingSet[i] = i
 258                 }
 259         } else if len(cmd.cgnames) != len(cmd.samples) {
 260                 return fmt.Errorf("mismatched sample list: %d samples in library, %d in %s", len(cmd.cgnames), len(cmd.samples), *samplesFilename)
 261         } else {
 262                 for i, name := range cmd.cgnames {
 263                         if s := trimFilenameForLabel(name); s != cmd.samples[i].id {
 264                                 return fmt.Errorf("mismatched sample list: sample %d is %q in library, %q in %s", i, s, cmd.samples[i].id, *samplesFilename)
 265                         }
 266                 }
 267                 if *caseControlOnly {
 268                         for i := 0; i < len(cmd.samples); i++ {
 269                                 if !cmd.samples[i].isTraining && !cmd.samples[i].isValidation {
 270                                         if i+1 < len(cmd.samples) {
 271                                                 copy(cmd.samples[i:], cmd.samples[i+1:])
 272                                                 copy(cmd.cgnames[i:], cmd.cgnames[i+1:])
 273                                         }
 274                                         cmd.samples = cmd.samples[:len(cmd.samples)-1]
 275                                         cmd.cgnames = cmd.cgnames[:len(cmd.cgnames)-1]
 276                                         i--
 277                                 }
 278                         }
 279                 }
 280                 cmd.chi2Cases = nil
 281                 cmd.trainingSetSize = 0
 282                 for i := range cmd.cgnames {
 283                         if cmd.samples[i].isTraining {
 284                                 cmd.trainingSet[i] = cmd.trainingSetSize
 285                                 cmd.trainingSetSize++
 286                                 cmd.chi2Cases = append(cmd.chi2Cases, cmd.samples[i].isCase)
 287                         } else {
 288                                 cmd.trainingSet[i] = -1
 289                         }
 290                 }
 291                 if cmd.pvalue == nil {
 292                         cmd.pvalue = func(onehot []bool) float64 {
 293                                 return pvalue(onehot, cmd.chi2Cases)
 294                         }
 295                 }
 296         }
 297         if cmd.filter.MinCoverage == 1 {
 298                 // In the generic formula below, floating point
 299                 // arithmetic can effectively push the coverage
 300                 // threshold above 1.0, which is impossible/useless.
 301                 // 1.0 needs to mean exactly 100% coverage.
 302                 cmd.minCoverage = len(cmd.cgnames)
 303         } else {
 304                 cmd.minCoverage = int(math.Ceil(cmd.filter.MinCoverage * float64(len(cmd.cgnames))))
 305         }
 306
 307         if len(cmd.samples[0].pcaComponents) > 0 {
 308                 cmd.pvalue = glmPvalueFunc(cmd.samples, cmd.pcaComponents)
 309                 // Unfortunately, statsmodel/glm lib logs stuff to
 310                 // os.Stdout when it panics on an unsolvable
 311                 // problem. We recover() from the panic in glm.go, but
 312                 // we also need to commandeer os.Stdout to avoid
 313                 // producing large quantities of logs.
 314                 stdoutWas := os.Stdout
 315                 defer func() { os.Stdout = stdoutWas }()
 316                 os.Stdout, err = os.Open(os.DevNull)
 317                 if err != nil {
 318                         return err
 319                 }
 320         }
 321
 322         // cgnamemap[name]==true for samples that we are including in
 323         // output
 324         cgnamemap := map[string]bool{}
 325         for _, name := range cmd.cgnames {
 326                 cgnamemap[name] = true
 327         }
 328
 329         err = writeSampleInfo(cmd.samples, *outputDir)
 330         if err != nil {
 331                 return err
 332         }
 333
 334         log.Info("indexing reference tiles")
 335         type reftileinfo struct {
 336                 variant  tileVariantID
 337                 seqname  string // chr1
 338                 pos      int    // distance from start of chromosome to starttag
 339                 tiledata []byte // acgtggcaa...
 340                 excluded bool   // true if excluded by regions file
 341                 nexttag  tagID  // tagID of following tile (-1 for last tag of chromosome)
 342         }
 343         isdup := map[tagID]bool{}
 344         reftile := map[tagID]*reftileinfo{}
 345         for seqname, cseq := range refseq {
 346                 pos := 0
 347                 lastreftag := tagID(-1)
 348                 for _, libref := range cseq {
 349                         if cmd.filter.MaxTag >= 0 && libref.Tag > tagID(cmd.filter.MaxTag) {
 350                                 continue
 351                         }
 352                         tiledata := reftiledata[libref]
 353                         if len(tiledata) == 0 {
 354                                 err = fmt.Errorf("missing tiledata for tag %d variant %d in %s in ref", libref.Tag, libref.Variant, seqname)
 355                                 return err
 356                         }
 357                         foundthistag := false
 358                         taglib.FindAll(tiledata[:len(tiledata)-1], func(tagid tagID, offset, _ int) {
 359                                 if !foundthistag && tagid == libref.Tag {
 360                                         foundthistag = true
 361                                         return
 362                                 }
 363                                 if dupref, ok := reftile[tagid]; ok {
 364                                         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)
 365                                         delete(reftile, tagid)
 366                                 } else {
 367                                         log.Printf("found tag %d at offset %d inside tile variant %+v on %s @ %d", tagid, offset, libref, seqname, pos+offset+1)
 368                                 }
 369                                 isdup[tagid] = true
 370                         })
 371                         if isdup[libref.Tag] {
 372                                 log.Printf("dropping reference tile %+v from %s @ %d, tag not unique", libref, seqname, pos)
 373                         } else if reftile[libref.Tag] != nil {
 374                                 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)
 375                                 delete(reftile, libref.Tag)
 376                                 log.Printf("dropping reference tile %+v from %s @ %d, tag not unique", libref, seqname, pos)
 377                                 isdup[libref.Tag] = true
 378                         } else {
 379                                 reftile[libref.Tag] = &reftileinfo{
 380                                         seqname:  seqname,
 381                                         variant:  libref.Variant,
 382                                         tiledata: tiledata,
 383                                         pos:      pos,
 384                                         nexttag:  -1,
 385                                 }
 386                                 if lastreftag >= 0 {
 387                                         reftile[lastreftag].nexttag = libref.Tag
 388                                 }
 389                                 lastreftag = libref.Tag
 390                         }
 391                         pos += len(tiledata) - taglen
 392                 }
 393                 log.Printf("... %s done, len %d", seqname, pos+taglen)
 394         }
 395
 396         var mask *mask
 397         if *regionsFilename != "" {
 398                 log.Printf("loading regions from %s", *regionsFilename)
 399                 mask, err = makeMask(*regionsFilename, *expandRegions)
 400                 if err != nil {
 401                         return err
 402                 }
 403                 log.Printf("before applying mask, len(reftile) == %d", len(reftile))
 404                 log.Printf("deleting reftile entries for regions outside %d intervals", mask.Len())
 405                 for _, rt := range reftile {
 406                         if !mask.Check(strings.TrimPrefix(rt.seqname, "chr"), rt.pos, rt.pos+len(rt.tiledata)) {
 407                                 rt.excluded = true
 408                         }
 409                 }
 410                 log.Printf("after applying mask, len(reftile) == %d", len(reftile))
 411         }
 412
 413         type hgvsColSet map[hgvs.Variant][2][]int8
 414         encodeHGVS := throttle{Max: len(refseq)}
 415         encodeHGVSTodo := map[string]chan hgvsColSet{}
 416         tmpHGVSCols := map[string]*os.File{}
 417         if *hgvsChunked {
 418                 for seqname := range refseq {
 419                         var f *os.File
 420                         f, err = os.Create(*outputDir + "/tmp." + seqname + ".gob")
 421                         if err != nil {
 422                                 return err
 423                         }
 424                         defer os.Remove(f.Name())
 425                         bufw := bufio.NewWriterSize(f, 1<<24)
 426                         enc := gob.NewEncoder(bufw)
 427                         tmpHGVSCols[seqname] = f
 428                         todo := make(chan hgvsColSet, 128)
 429                         encodeHGVSTodo[seqname] = todo
 430                         encodeHGVS.Go(func() error {
 431                                 for colset := range todo {
 432                                         err := enc.Encode(colset)
 433                                         if err != nil {
 434                                                 encodeHGVS.Report(err)
 435                                                 for range todo {
 436                                                 }
 437                                                 return err
 438                                         }
 439                                 }
 440                                 return bufw.Flush()
 441                         })
 442                 }
 443         }
 444
 445         var toMerge [][]int16
 446         if *mergeOutput || *hgvsSingle {
 447                 toMerge = make([][]int16, len(infiles))
 448         }
 449         var onehotIndirect [][2][]uint32 // [chunkIndex][axis][index]
 450         var onehotChunkSize []uint32
 451         var onehotXrefs [][]onehotXref
 452         if *onehotSingle || *onlyPCA {
 453                 onehotIndirect = make([][2][]uint32, len(infiles))
 454                 onehotChunkSize = make([]uint32, len(infiles))
 455                 onehotXrefs = make([][]onehotXref, len(infiles))
 456         }
 457         chunkStartTag := make([]tagID, len(infiles))
 458
 459         throttleMem := throttle{Max: cmd.threads} // TODO: estimate using mem and data size
 460         throttleNumpyMem := throttle{Max: cmd.threads/2 + 1}
 461         log.Info("generating annotations and numpy matrix for each slice")
 462         var errSkip = errors.New("skip infile")
 463         var done int64
 464         for infileIdx, infile := range infiles {
 465                 infileIdx, infile := infileIdx, infile
 466                 throttleMem.Go(func() error {
 467                         seq := make(map[tagID][]TileVariant, 50000)
 468                         cgs := make(map[string]CompactGenome, len(cmd.cgnames))
 469                         f, err := open(infile)
 470                         if err != nil {
 471                                 return err
 472                         }
 473                         defer f.Close()
 474                         log.Infof("%04d: reading %s", infileIdx, infile)
 475                         err = DecodeLibrary(f, strings.HasSuffix(infile, ".gz"), func(ent *LibraryEntry) error {
 476                                 for _, tv := range ent.TileVariants {
 477                                         if tv.Ref {
 478                                                 continue
 479                                         }
 480                                         // Skip tile with no
 481                                         // corresponding ref tile, if
 482                                         // mask is in play (we can't
 483                                         // determine coordinates for
 484                                         // these)
 485                                         if mask != nil && reftile[tv.Tag] == nil {
 486                                                 continue
 487                                         }
 488                                         // Skip tile whose
 489                                         // corresponding ref tile is
 490                                         // outside target regions --
 491                                         // unless it's a potential
 492                                         // spanning tile.
 493                                         if mask != nil && reftile[tv.Tag].excluded &&
 494                                                 (int(tv.Tag+1) >= len(tagset) ||
 495                                                         (bytes.HasSuffix(tv.Sequence, tagset[tv.Tag+1]) && reftile[tv.Tag+1] != nil && !reftile[tv.Tag+1].excluded)) {
 496                                                 continue
 497                                         }
 498                                         if tv.Tag == cmd.debugTag {
 499                                                 log.Printf("infile %d %s tag %d variant %d hash %x", infileIdx, infile, tv.Tag, tv.Variant, tv.Blake2b[:3])
 500                                         }
 501                                         variants := seq[tv.Tag]
 502                                         if len(variants) == 0 {
 503                                                 variants = make([]TileVariant, 100)
 504                                         }
 505                                         for len(variants) <= int(tv.Variant) {
 506                                                 variants = append(variants, TileVariant{})
 507                                         }
 508                                         variants[int(tv.Variant)] = tv
 509                                         seq[tv.Tag] = variants
 510                                 }
 511                                 for _, cg := range ent.CompactGenomes {
 512                                         if cmd.filter.MaxTag >= 0 && cg.StartTag > tagID(cmd.filter.MaxTag) {
 513                                                 return errSkip
 514                                         }
 515                                         if !cgnamemap[cg.Name] {
 516                                                 continue
 517                                         }
 518                                         // pad to full slice size
 519                                         // to avoid out-of-bounds
 520                                         // checks later
 521                                         if sliceSize := 2 * int(cg.EndTag-cg.StartTag); len(cg.Variants) < sliceSize {
 522                                                 cg.Variants = append(cg.Variants, make([]tileVariantID, sliceSize-len(cg.Variants))...)
 523                                         }
 524                                         cgs[cg.Name] = cg
 525                                 }
 526                                 return nil
 527                         })
 528                         if err == errSkip {
 529                                 return nil
 530                         } else if err != nil {
 531                                 return fmt.Errorf("%04d: DecodeLibrary(%s): err", infileIdx, infile)
 532                         }
 533                         tagstart := cgs[cmd.cgnames[0]].StartTag
 534                         tagend := cgs[cmd.cgnames[0]].EndTag
 535                         chunkStartTag[infileIdx] = tagstart
 536
 537                         // TODO: filters
 538
 539                         log.Infof("%04d: renumber/dedup variants for tags %d-%d", infileIdx, tagstart, tagend)
 540                         variantRemap := make([][]tileVariantID, tagend-tagstart)
 541                         throttleCPU := throttle{Max: runtime.GOMAXPROCS(0)}
 542                         for tag, variants := range seq {
 543                                 tag, variants := tag, variants
 544                                 throttleCPU.Go(func() error {
 545                                         alleleCoverage := 0
 546                                         count := make(map[[blake2b.Size256]byte]int, len(variants))
 547
 548                                         rt := reftile[tag]
 549                                         if rt != nil {
 550                                                 count[blake2b.Sum256(rt.tiledata)] = 0
 551                                         }
 552
 553                                         for cgname, cg := range cgs {
 554                                                 idx := int(tag-tagstart) * 2
 555                                                 for allele := 0; allele < 2; allele++ {
 556                                                         v := cg.Variants[idx+allele]
 557                                                         if v > 0 && len(variants[v].Sequence) > 0 {
 558                                                                 count[variants[v].Blake2b]++
 559                                                                 alleleCoverage++
 560                                                         }
 561                                                         if v > 0 && tag == cmd.debugTag {
 562                                                                 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])
 563                                                         }
 564                                                 }
 565                                         }
 566                                         if alleleCoverage < cmd.minCoverage*2 {
 567                                                 idx := int(tag-tagstart) * 2
 568                                                 for _, cg := range cgs {
 569                                                         cg.Variants[idx] = 0
 570                                                         cg.Variants[idx+1] = 0
 571                                                 }
 572                                                 if tag == cmd.debugTag {
 573                                                         log.Printf("tag %d alleleCoverage %d < min %d, sample data wiped", tag, alleleCoverage, cmd.minCoverage*2)
 574                                                 }
 575                                                 return nil
 576                                         }
 577
 578                                         // hash[i] will be the hash of
 579                                         // the variant(s) that should
 580                                         // be at rank i (0-based).
 581                                         hash := make([][blake2b.Size256]byte, 0, len(count))
 582                                         for b := range count {
 583                                                 hash = append(hash, b)
 584                                         }
 585                                         sort.Slice(hash, func(i, j int) bool {
 586                                                 bi, bj := &hash[i], &hash[j]
 587                                                 if ci, cj := count[*bi], count[*bj]; ci != cj {
 588                                                         return ci > cj
 589                                                 } else {
 590                                                         return bytes.Compare((*bi)[:], (*bj)[:]) < 0
 591                                                 }
 592                                         })
 593                                         // rank[b] will be the 1-based
 594                                         // new variant number for
 595                                         // variants whose hash is b.
 596                                         rank := make(map[[blake2b.Size256]byte]tileVariantID, len(hash))
 597                                         for i, h := range hash {
 598                                                 rank[h] = tileVariantID(i + 1)
 599                                         }
 600                                         if tag == cmd.debugTag {
 601                                                 for h, r := range rank {
 602                                                         log.Printf("tag %d rank(%x) = %v", tag, h[:3], r)
 603                                                 }
 604                                         }
 605                                         // remap[v] will be the new
 606                                         // variant number for original
 607                                         // variant number v.
 608                                         remap := make([]tileVariantID, len(variants))
 609                                         for i, tv := range variants {
 610                                                 remap[i] = rank[tv.Blake2b]
 611                                         }
 612                                         if tag == cmd.debugTag {
 613                                                 for in, out := range remap {
 614                                                         if out > 0 {
 615                                                                 log.Printf("tag %d remap %d => %d", tag, in, out)
 616                                                         }
 617                                                 }
 618                                         }
 619                                         variantRemap[tag-tagstart] = remap
 620                                         if rt != nil {
 621                                                 refrank := rank[blake2b.Sum256(rt.tiledata)]
 622                                                 if tag == cmd.debugTag {
 623                                                         log.Printf("tag %d reftile variant %d => %d", tag, rt.variant, refrank)
 624                                                 }
 625                                                 rt.variant = refrank
 626                                         }
 627                                         return nil
 628                                 })
 629                         }
 630                         throttleCPU.Wait()
 631
 632                         var onehotChunk [][]int8
 633                         var onehotXref []onehotXref
 634
 635                         var annotationsFilename string
 636                         if *onlyPCA {
 637                                 annotationsFilename = "/dev/null"
 638                         } else {
 639                                 annotationsFilename = fmt.Sprintf("%s/matrix.%04d.annotations.csv", *outputDir, infileIdx)
 640                                 log.Infof("%04d: writing %s", infileIdx, annotationsFilename)
 641                         }
 642                         annof, err := os.Create(annotationsFilename)
 643                         if err != nil {
 644                                 return err
 645                         }
 646                         annow := bufio.NewWriterSize(annof, 1<<20)
 647                         outcol := 0
 648                         for tag := tagstart; tag < tagend; tag++ {
 649                                 rt := reftile[tag]
 650                                 if rt == nil && mask != nil {
 651                                         // With no ref tile, we don't
 652                                         // have coordinates to say
 653                                         // this is in the desired
 654                                         // regions -- so it's not.
 655                                         // TODO: handle ref spanning
 656                                         // tile case.
 657                                         continue
 658                                 }
 659                                 if rt != nil && rt.excluded {
 660                                         // TODO: don't skip yet --
 661                                         // first check for spanning
 662                                         // tile variants that
 663                                         // intersect non-excluded ref
 664                                         // tiles.
 665                                         continue
 666                                 }
 667                                 if cmd.filter.MaxTag >= 0 && tag > tagID(cmd.filter.MaxTag) {
 668                                         break
 669                                 }
 670                                 remap := variantRemap[tag-tagstart]
 671                                 if remap == nil {
 672                                         // was not assigned above,
 673                                         // because minCoverage
 674                                         outcol++
 675                                         continue
 676                                 }
 677                                 maxv := tileVariantID(0)
 678                                 for _, v := range remap {
 679                                         if maxv < v {
 680                                                 maxv = v
 681                                         }
 682                                 }
 683                                 if *onehotChunked || *onehotSingle || *onlyPCA {
 684                                         onehot, xrefs := cmd.tv2homhet(cgs, maxv, remap, tag, tagstart, seq)
 685                                         if tag == cmd.debugTag {
 686                                                 log.WithFields(logrus.Fields{
 687                                                         "onehot": onehot,
 688                                                         "xrefs":  xrefs,
 689                                                 }).Info("tv2homhet()")
 690                                         }
 691                                         onehotChunk = append(onehotChunk, onehot...)
 692                                         onehotXref = append(onehotXref, xrefs...)
 693                                 }
 694                                 if *onlyPCA {
 695                                         outcol++
 696                                         continue
 697                                 }
 698                                 if rt == nil {
 699                                         // Reference does not use any
 700                                         // variant of this tile
 701                                         //
 702                                         // TODO: diff against the
 703                                         // relevant portion of the
 704                                         // ref's spanning tile
 705                                         outcol++
 706                                         continue
 707                                 }
 708                                 fmt.Fprintf(annow, "%d,%d,%d,=,%s,%d,,,\n", tag, outcol, rt.variant, rt.seqname, rt.pos)
 709                                 variants := seq[tag]
 710                                 reftilestr := strings.ToUpper(string(rt.tiledata))
 711
 712                                 done := make([]bool, maxv+1)
 713                                 variantDiffs := make([][]hgvs.Variant, maxv+1)
 714                                 for v, tv := range variants {
 715                                         v := remap[v]
 716                                         if v == 0 || v == rt.variant || done[v] {
 717                                                 continue
 718                                         } else {
 719                                                 done[v] = true
 720                                         }
 721                                         if len(tv.Sequence) < taglen {
 722                                                 continue
 723                                         }
 724                                         // if reftilestr doesn't end
 725                                         // in the same tag as tv,
 726                                         // extend reftilestr with
 727                                         // following ref tiles until
 728                                         // it does (up to an arbitrary
 729                                         // sanity-check limit)
 730                                         reftilestr := reftilestr
 731                                         endtagstr := strings.ToUpper(string(tv.Sequence[len(tv.Sequence)-taglen:]))
 732                                         for i, rt := 0, rt; i < annotationMaxTileSpan && !strings.HasSuffix(reftilestr, endtagstr) && rt.nexttag >= 0; i++ {
 733                                                 rt = reftile[rt.nexttag]
 734                                                 if rt == nil {
 735                                                         break
 736                                                 }
 737                                                 reftilestr += strings.ToUpper(string(rt.tiledata[taglen:]))
 738                                         }
 739                                         if mask != nil && !mask.Check(strings.TrimPrefix(rt.seqname, "chr"), rt.pos, rt.pos+len(reftilestr)) {
 740                                                 continue
 741                                         }
 742                                         if !strings.HasSuffix(reftilestr, endtagstr) {
 743                                                 fmt.Fprintf(annow, "%d,%d,%d,,%s,%d,,,\n", tag, outcol, v, rt.seqname, rt.pos)
 744                                                 continue
 745                                         }
 746                                         if lendiff := len(reftilestr) - len(tv.Sequence); lendiff < -1000 || lendiff > 1000 {
 747                                                 fmt.Fprintf(annow, "%d,%d,%d,,%s,%d,,,\n", tag, outcol, v, rt.seqname, rt.pos)
 748                                                 continue
 749                                         }
 750                                         diffs, _ := hgvs.Diff(reftilestr, strings.ToUpper(string(tv.Sequence)), 0)
 751                                         for i := range diffs {
 752                                                 diffs[i].Position += rt.pos
 753                                         }
 754                                         for _, diff := range diffs {
 755                                                 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)
 756                                         }
 757                                         if *hgvsChunked {
 758                                                 variantDiffs[v] = diffs
 759                                         }
 760                                 }
 761                                 if *hgvsChunked {
 762                                         // We can now determine, for each HGVS
 763                                         // variant (diff) in this reftile
 764                                         // region, whether a given genome
 765                                         // phase/allele (1) has the variant, (0) has
 766                                         // =ref or a different variant in that
 767                                         // position, or (-1) is lacking
 768                                         // coverage / couldn't be diffed.
 769                                         hgvsCol := hgvsColSet{}
 770                                         for _, diffs := range variantDiffs {
 771                                                 for _, diff := range diffs {
 772                                                         if _, ok := hgvsCol[diff]; ok {
 773                                                                 continue
 774                                                         }
 775                                                         hgvsCol[diff] = [2][]int8{
 776                                                                 make([]int8, len(cmd.cgnames)),
 777                                                                 make([]int8, len(cmd.cgnames)),
 778                                                         }
 779                                                 }
 780                                         }
 781                                         for row, name := range cmd.cgnames {
 782                                                 variants := cgs[name].Variants[(tag-tagstart)*2:]
 783                                                 for ph := 0; ph < 2; ph++ {
 784                                                         v := variants[ph]
 785                                                         if int(v) >= len(remap) {
 786                                                                 v = 0
 787                                                         } else {
 788                                                                 v = remap[v]
 789                                                         }
 790                                                         if v == rt.variant {
 791                                                                 // hgvsCol[*][ph][row] is already 0
 792                                                         } else if len(variantDiffs[v]) == 0 {
 793                                                                 // lacking coverage / couldn't be diffed
 794                                                                 for _, col := range hgvsCol {
 795                                                                         col[ph][row] = -1
 796                                                                 }
 797                                                         } else {
 798                                                                 for _, diff := range variantDiffs[v] {
 799                                                                         hgvsCol[diff][ph][row] = 1
 800                                                                 }
 801                                                         }
 802                                                 }
 803                                         }
 804                                         for diff, colpair := range hgvsCol {
 805                                                 allele2homhet(colpair)
 806                                                 if !cmd.filterHGVScolpair(colpair) {
 807                                                         delete(hgvsCol, diff)
 808                                                 }
 809                                         }
 810                                         if len(hgvsCol) > 0 {
 811                                                 encodeHGVSTodo[rt.seqname] <- hgvsCol
 812                                         }
 813                                 }
 814                                 outcol++
 815                         }
 816                         err = annow.Flush()
 817                         if err != nil {
 818                                 return err
 819                         }
 820                         err = annof.Close()
 821                         if err != nil {
 822                                 return err
 823                         }
 824
 825                         if *onehotChunked {
 826                                 // transpose onehotChunk[col][row] to numpy[row*ncols+col]
 827                                 rows := len(cmd.cgnames)
 828                                 cols := len(onehotChunk)
 829                                 log.Infof("%04d: preparing onehot numpy (rows=%d, cols=%d, mem=%d)", infileIdx, rows, cols, rows*cols)
 830                                 throttleNumpyMem.Acquire()
 831                                 out := onehotcols2int8(onehotChunk)
 832                                 fnm := fmt.Sprintf("%s/onehot.%04d.npy", *outputDir, infileIdx)
 833                                 err = writeNumpyInt8(fnm, out, rows, cols)
 834                                 if err != nil {
 835                                         return err
 836                                 }
 837                                 fnm = fmt.Sprintf("%s/onehot-columns.%04d.npy", *outputDir, infileIdx)
 838                                 err = writeNumpyInt32(fnm, onehotXref2int32(onehotXref), 4, len(onehotXref))
 839                                 if err != nil {
 840                                         return err
 841                                 }
 842                                 debug.FreeOSMemory()
 843                                 throttleNumpyMem.Release()
 844                         }
 845                         if *onehotSingle || *onlyPCA {
 846                                 onehotIndirect[infileIdx] = onehotChunk2Indirect(onehotChunk)
 847                                 onehotChunkSize[infileIdx] = uint32(len(onehotChunk))
 848                                 onehotXrefs[infileIdx] = onehotXref
 849                                 n := len(onehotIndirect[infileIdx][0])
 850                                 log.Infof("%04d: keeping onehot coordinates in memory (n=%d, mem=%d)", infileIdx, n, n*8*2)
 851                         }
 852                         if !(*onehotSingle || *onehotChunked || *onlyPCA) || *mergeOutput || *hgvsSingle {
 853                                 log.Infof("%04d: preparing numpy (rows=%d, cols=%d)", infileIdx, len(cmd.cgnames), 2*outcol)
 854                                 throttleNumpyMem.Acquire()
 855                                 rows := len(cmd.cgnames)
 856                                 cols := 2 * outcol
 857                                 out := make([]int16, rows*cols)
 858                                 for row, name := range cmd.cgnames {
 859                                         outidx := row * cols
 860                                         for col, v := range cgs[name].Variants {
 861                                                 tag := tagstart + tagID(col/2)
 862                                                 if cmd.filter.MaxTag >= 0 && tag > tagID(cmd.filter.MaxTag) {
 863                                                         break
 864                                                 }
 865                                                 if rt := reftile[tag]; rt == nil || rt.excluded {
 866                                                         continue
 867                                                 }
 868                                                 if v == 0 {
 869                                                         out[outidx] = 0 // tag not found / spanning tile
 870                                                 } else if variants, ok := seq[tag]; ok && int(v) < len(variants) && len(variants[v].Sequence) > 0 {
 871                                                         out[outidx] = int16(variantRemap[tag-tagstart][v])
 872                                                 } else {
 873                                                         out[outidx] = -1 // low quality tile variant
 874                                                 }
 875                                                 if tag == cmd.debugTag {
 876                                                         log.Printf("tag %d row %d col %d outidx %d v %d out %d", tag, row, col, outidx, v, out[outidx])
 877                                                 }
 878                                                 outidx++
 879                                         }
 880                                 }
 881                                 seq = nil
 882                                 cgs = nil
 883                                 debug.FreeOSMemory()
 884                                 throttleNumpyMem.Release()
 885                                 if *mergeOutput || *hgvsSingle {
 886                                         log.Infof("%04d: matrix fragment %d rows x %d cols", infileIdx, rows, cols)
 887                                         toMerge[infileIdx] = out
 888                                 }
 889                                 if !*mergeOutput && !*onehotChunked && !*onehotSingle {
 890                                         fnm := fmt.Sprintf("%s/matrix.%04d.npy", *outputDir, infileIdx)
 891                                         err = writeNumpyInt16(fnm, out, rows, cols)
 892                                         if err != nil {
 893                                                 return err
 894                                         }
 895                                 }
 896                         }
 897                         debug.FreeOSMemory()
 898                         log.Infof("%s: done (%d/%d)", infile, int(atomic.AddInt64(&done, 1)), len(infiles))
 899                         return nil
 900                 })
 901         }
 902         if err = throttleMem.Wait(); err != nil {
 903                 return err
 904         }
 905
 906         if *hgvsChunked {
 907                 log.Info("flushing hgvsCols temp files")
 908                 for seqname := range refseq {
 909                         close(encodeHGVSTodo[seqname])
 910                 }
 911                 err = encodeHGVS.Wait()
 912                 if err != nil {
 913                         return err
 914                 }
 915                 for seqname := range refseq {
 916                         log.Infof("%s: reading hgvsCols from temp file", seqname)
 917                         f := tmpHGVSCols[seqname]
 918                         _, err = f.Seek(0, io.SeekStart)
 919                         if err != nil {
 920                                 return err
 921                         }
 922                         var hgvsCols hgvsColSet
 923                         dec := gob.NewDecoder(bufio.NewReaderSize(f, 1<<24))
 924                         for err == nil {
 925                                 err = dec.Decode(&hgvsCols)
 926                         }
 927                         if err != io.EOF {
 928                                 return err
 929                         }
 930                         log.Infof("%s: sorting %d hgvs variants", seqname, len(hgvsCols))
 931                         variants := make([]hgvs.Variant, 0, len(hgvsCols))
 932                         for v := range hgvsCols {
 933                                 variants = append(variants, v)
 934                         }
 935                         sort.Slice(variants, func(i, j int) bool {
 936                                 vi, vj := &variants[i], &variants[j]
 937                                 if vi.Position != vj.Position {
 938                                         return vi.Position < vj.Position
 939                                 } else if vi.Ref != vj.Ref {
 940                                         return vi.Ref < vj.Ref
 941                                 } else {
 942                                         return vi.New < vj.New
 943                                 }
 944                         })
 945                         rows := len(cmd.cgnames)
 946                         cols := len(variants) * 2
 947                         log.Infof("%s: building hgvs matrix (rows=%d, cols=%d, mem=%d)", seqname, rows, cols, rows*cols)
 948                         out := make([]int8, rows*cols)
 949                         for varIdx, variant := range variants {
 950                                 hgvsCols := hgvsCols[variant]
 951                                 for row := range cmd.cgnames {
 952                                         for ph := 0; ph < 2; ph++ {
 953                                                 out[row*cols+varIdx+ph] = hgvsCols[ph][row]
 954                                         }
 955                                 }
 956                         }
 957                         err = writeNumpyInt8(fmt.Sprintf("%s/hgvs.%s.npy", *outputDir, seqname), out, rows, cols)
 958                         if err != nil {
 959                                 return err
 960                         }
 961                         out = nil
 962
 963                         fnm := fmt.Sprintf("%s/hgvs.%s.annotations.csv", *outputDir, seqname)
 964                         log.Infof("%s: writing hgvs column labels to %s", seqname, fnm)
 965                         var hgvsLabels bytes.Buffer
 966                         for varIdx, variant := range variants {
 967                                 fmt.Fprintf(&hgvsLabels, "%d,%s:g.%s\n", varIdx, seqname, variant.String())
 968                         }
 969                         err = ioutil.WriteFile(fnm, hgvsLabels.Bytes(), 0666)
 970                         if err != nil {
 971                                 return err
 972                         }
 973                 }
 974         }
 975
 976         if *mergeOutput || *hgvsSingle {
 977                 var annow *bufio.Writer
 978                 var annof *os.File
 979                 if *mergeOutput {
 980                         annoFilename := fmt.Sprintf("%s/matrix.annotations.csv", *outputDir)
 981                         annof, err = os.Create(annoFilename)
 982                         if err != nil {
 983                                 return err
 984                         }
 985                         annow = bufio.NewWriterSize(annof, 1<<20)
 986                 }
 987
 988                 rows := len(cmd.cgnames)
 989                 cols := 0
 990                 for _, chunk := range toMerge {
 991                         cols += len(chunk) / rows
 992                 }
 993                 log.Infof("merging output matrix (rows=%d, cols=%d, mem=%d) and annotations", rows, cols, rows*cols*2)
 994                 var out []int16
 995                 if *mergeOutput {
 996                         out = make([]int16, rows*cols)
 997                 }
 998                 hgvsCols := map[string][2][]int16{} // hgvs -> [[g0,g1,g2,...], [g0,g1,g2,...]] (slice of genomes for each phase)
 999                 startcol := 0
1000                 for outIdx, chunk := range toMerge {
1001                         chunkcols := len(chunk) / rows
1002                         if *mergeOutput {
1003                                 for row := 0; row < rows; row++ {
1004                                         copy(out[row*cols+startcol:], chunk[row*chunkcols:(row+1)*chunkcols])
1005                                 }
1006                         }
1007                         toMerge[outIdx] = nil
1008
1009                         annotationsFilename := fmt.Sprintf("%s/matrix.%04d.annotations.csv", *outputDir, outIdx)
1010                         log.Infof("reading %s", annotationsFilename)
1011                         buf, err := os.ReadFile(annotationsFilename)
1012                         if err != nil {
1013                                 return err
1014                         }
1015                         if *mergeOutput {
1016                                 err = os.Remove(annotationsFilename)
1017                                 if err != nil {
1018                                         return err
1019                                 }
1020                         }
1021                         for _, line := range bytes.Split(buf, []byte{'\n'}) {
1022                                 if len(line) == 0 {
1023                                         continue
1024                                 }
1025                                 fields := bytes.SplitN(line, []byte{','}, 9)
1026                                 tag, _ := strconv.Atoi(string(fields[0]))
1027                                 incol, _ := strconv.Atoi(string(fields[1]))
1028                                 tileVariant, _ := strconv.Atoi(string(fields[2]))
1029                                 hgvsID := string(fields[3])
1030                                 seqname := string(fields[4])
1031                                 pos, _ := strconv.Atoi(string(fields[5]))
1032                                 refseq := fields[6]
1033                                 if hgvsID == "" {
1034                                         // Null entry for un-diffable
1035                                         // tile variant
1036                                         continue
1037                                 }
1038                                 if hgvsID == "=" {
1039                                         // Null entry for ref tile
1040                                         continue
1041                                 }
1042                                 if mask != nil && !mask.Check(strings.TrimPrefix(seqname, "chr"), pos, pos+len(refseq)) {
1043                                         // The tile intersects one of
1044                                         // the selected regions, but
1045                                         // this particular HGVS
1046                                         // variant does not.
1047                                         continue
1048                                 }
1049                                 hgvsColPair := hgvsCols[hgvsID]
1050                                 if hgvsColPair[0] == nil {
1051                                         // values in new columns start
1052                                         // out as -1 ("no data yet")
1053                                         // or 0 ("=ref") here, may
1054                                         // change to 1 ("hgvs variant
1055                                         // present") below, either on
1056                                         // this line or a future line.
1057                                         hgvsColPair = [2][]int16{make([]int16, len(cmd.cgnames)), make([]int16, len(cmd.cgnames))}
1058                                         rt, ok := reftile[tagID(tag)]
1059                                         if !ok {
1060                                                 err = fmt.Errorf("bug: seeing annotations for tag %d, but it has no reftile entry", tag)
1061                                                 return err
1062                                         }
1063                                         for ph := 0; ph < 2; ph++ {
1064                                                 for row := 0; row < rows; row++ {
1065                                                         v := chunk[row*chunkcols+incol*2+ph]
1066                                                         if tileVariantID(v) == rt.variant {
1067                                                                 hgvsColPair[ph][row] = 0
1068                                                         } else {
1069                                                                 hgvsColPair[ph][row] = -1
1070                                                         }
1071                                                 }
1072                                         }
1073                                         hgvsCols[hgvsID] = hgvsColPair
1074                                         if annow != nil {
1075                                                 hgvsref := hgvs.Variant{
1076                                                         Position: pos,
1077                                                         Ref:      string(refseq),
1078                                                         New:      string(refseq),
1079                                                 }
1080                                                 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])
1081                                         }
1082                                 }
1083                                 if annow != nil {
1084                                         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])
1085                                 }
1086                                 for ph := 0; ph < 2; ph++ {
1087                                         for row := 0; row < rows; row++ {
1088                                                 v := chunk[row*chunkcols+incol*2+ph]
1089                                                 if int(v) == tileVariant {
1090                                                         hgvsColPair[ph][row] = 1
1091                                                 }
1092                                         }
1093                                 }
1094                         }
1095
1096                         startcol += chunkcols
1097                 }
1098                 if *mergeOutput {
1099                         err = annow.Flush()
1100                         if err != nil {
1101                                 return err
1102                         }
1103                         err = annof.Close()
1104                         if err != nil {
1105                                 return err
1106                         }
1107                         err = writeNumpyInt16(fmt.Sprintf("%s/matrix.npy", *outputDir), out, rows, cols)
1108                         if err != nil {
1109                                 return err
1110                         }
1111                 }
1112                 out = nil
1113
1114                 if *hgvsSingle {
1115                         cols = len(hgvsCols) * 2
1116                         log.Printf("building hgvs-based matrix: %d rows x %d cols", rows, cols)
1117                         out = make([]int16, rows*cols)
1118                         hgvsIDs := make([]string, 0, cols/2)
1119                         for hgvsID := range hgvsCols {
1120                                 hgvsIDs = append(hgvsIDs, hgvsID)
1121                         }
1122                         sort.Strings(hgvsIDs)
1123                         var hgvsLabels bytes.Buffer
1124                         for idx, hgvsID := range hgvsIDs {
1125                                 fmt.Fprintf(&hgvsLabels, "%d,%s\n", idx, hgvsID)
1126                                 for ph := 0; ph < 2; ph++ {
1127                                         hgvscol := hgvsCols[hgvsID][ph]
1128                                         for row, val := range hgvscol {
1129                                                 out[row*cols+idx*2+ph] = val
1130                                         }
1131                                 }
1132                         }
1133                         err = writeNumpyInt16(fmt.Sprintf("%s/hgvs.npy", *outputDir), out, rows, cols)
1134                         if err != nil {
1135                                 return err
1136                         }
1137
1138                         fnm := fmt.Sprintf("%s/hgvs.annotations.csv", *outputDir)
1139                         log.Printf("writing hgvs labels: %s", fnm)
1140                         err = ioutil.WriteFile(fnm, hgvsLabels.Bytes(), 0777)
1141                         if err != nil {
1142                                 return err
1143                         }
1144                 }
1145         }
1146         if *onehotSingle || *onlyPCA {
1147                 nzCount := 0
1148                 for _, part := range onehotIndirect {
1149                         nzCount += len(part[0])
1150                 }
1151                 onehot := make([]uint32, nzCount*2) // [r,r,r,...,c,c,c,...]
1152                 var xrefs []onehotXref
1153                 chunkOffset := uint32(0)
1154                 outcol := 0
1155                 for i, part := range onehotIndirect {
1156                         for i := range part[1] {
1157                                 part[1][i] += chunkOffset
1158                         }
1159                         copy(onehot[outcol:], part[0])
1160                         copy(onehot[outcol+nzCount:], part[1])
1161                         xrefs = append(xrefs, onehotXrefs[i]...)
1162
1163                         outcol += len(part[0])
1164                         chunkOffset += onehotChunkSize[i]
1165
1166                         part[0] = nil
1167                         part[1] = nil
1168                         onehotXrefs[i] = nil
1169                         debug.FreeOSMemory()
1170                 }
1171                 if *onehotSingle {
1172                         fnm := fmt.Sprintf("%s/onehot.npy", *outputDir)
1173                         err = writeNumpyUint32(fnm, onehot, 2, nzCount)
1174                         if err != nil {
1175                                 return err
1176                         }
1177                         fnm = fmt.Sprintf("%s/onehot-columns.npy", *outputDir)
1178                         err = writeNumpyInt32(fnm, onehotXref2int32(xrefs), 5, len(xrefs))
1179                         if err != nil {
1180                                 return err
1181                         }
1182                         fnm = fmt.Sprintf("%s/stats.json", *outputDir)
1183                         j, err := json.Marshal(map[string]interface{}{
1184                                 "pvalueCallCount": cmd.pvalueCallCount,
1185                         })
1186                         if err != nil {
1187                                 return err
1188                         }
1189                         err = os.WriteFile(fnm, j, 0777)
1190                         if err != nil {
1191                                 return err
1192                         }
1193                 }
1194                 if *onlyPCA {
1195                         cols := 0
1196                         for _, c := range onehot[nzCount:] {
1197                                 if int(c) >= cols {
1198                                         cols = int(c) + 1
1199                                 }
1200                         }
1201                         if cols == 0 {
1202                                 return fmt.Errorf("cannot do PCA: one-hot matrix is empty")
1203                         }
1204                         log.Printf("have %d one-hot cols", cols)
1205                         stride := 1
1206                         for *maxPCATiles > 0 && cols > *maxPCATiles*2 {
1207                                 cols = (cols + 1) / 2
1208                                 stride = stride * 2
1209                         }
1210                         if cols%2 == 1 {
1211                                 // we work with pairs of columns
1212                                 cols++
1213                         }
1214                         log.Printf("creating full matrix (%d rows) and training matrix (%d rows) with %d cols, stride %d", len(cmd.cgnames), cmd.trainingSetSize, cols, stride)
1215                         mtxFull := mat.NewDense(len(cmd.cgnames), cols, nil)
1216                         mtxTrain := mat.NewDense(cmd.trainingSetSize, cols, nil)
1217                         for i, c := range onehot[nzCount:] {
1218                                 if int(c/2)%stride == 0 {
1219                                         outcol := int(c/2)/stride*2 + int(c)%2
1220                                         mtxFull.Set(int(onehot[i]), outcol, 1)
1221                                         if trainRow := cmd.trainingSet[int(onehot[i])]; trainRow >= 0 {
1222                                                 mtxTrain.Set(trainRow, outcol, 1)
1223                                         }
1224                                 }
1225                         }
1226                         log.Print("fitting")
1227                         transformer := nlp.NewPCA(cmd.pcaComponents)
1228                         transformer.Fit(mtxTrain.T())
1229                         log.Printf("transforming")
1230                         pca, err := transformer.Transform(mtxFull.T())
1231                         if err != nil {
1232                                 return err
1233                         }
1234                         pca = pca.T()
1235                         outrows, outcols := pca.Dims()
1236                         log.Printf("copying result to numpy output array: %d rows, %d cols", outrows, outcols)
1237                         out := make([]float64, outrows*outcols)
1238                         for i := 0; i < outrows; i++ {
1239                                 for j := 0; j < outcols; j++ {
1240                                         out[i*outcols+j] = pca.At(i, j)
1241                                 }
1242                         }
1243                         fnm := fmt.Sprintf("%s/pca.npy", *outputDir)
1244                         log.Printf("writing numpy: %s", fnm)
1245                         output, err := os.OpenFile(fnm, os.O_CREATE|os.O_TRUNC|os.O_WRONLY, 0777)
1246                         if err != nil {
1247                                 return err
1248                         }
1249                         npw, err := gonpy.NewWriter(nopCloser{output})
1250                         if err != nil {
1251                                 return fmt.Errorf("gonpy.NewWriter: %w", err)
1252                         }
1253                         npw.Shape = []int{outrows, outcols}
1254                         err = npw.WriteFloat64(out)
1255                         if err != nil {
1256                                 return fmt.Errorf("WriteFloat64: %w", err)
1257                         }
1258                         err = output.Close()
1259                         if err != nil {
1260                                 return err
1261                         }
1262                         log.Print("done")
1263
1264                         log.Print("copying pca components to sampleInfo")
1265                         for i := range cmd.samples {
1266                                 cmd.samples[i].pcaComponents = make([]float64, outcols)
1267                                 for c := 0; c < outcols; c++ {
1268                                         cmd.samples[i].pcaComponents[i] = pca.At(i, c)
1269                                 }
1270                         }
1271                         log.Print("done")
1272
1273                         err = writeSampleInfo(cmd.samples, *outputDir)
1274                         if err != nil {
1275                                 return err
1276                         }
1277                 }
1278         }
1279         if !*mergeOutput && !*onehotChunked && !*onehotSingle && !*onlyPCA {
1280                 tagoffsetFilename := *outputDir + "/chunk-tag-offset.csv"
1281                 log.Infof("writing tag offsets to %s", tagoffsetFilename)
1282                 var f *os.File
1283                 f, err = os.Create(tagoffsetFilename)
1284                 if err != nil {
1285                         return err
1286                 }
1287                 defer f.Close()
1288                 for idx, offset := range chunkStartTag {
1289                         _, err = fmt.Fprintf(f, "%q,%d\n", fmt.Sprintf("matrix.%04d.npy", idx), offset)
1290                         if err != nil {
1291                                 err = fmt.Errorf("write %s: %w", tagoffsetFilename, err)
1292                                 return err
1293                         }
1294                 }
1295                 err = f.Close()
1296                 if err != nil {
1297                         err = fmt.Errorf("close %s: %w", tagoffsetFilename, err)
1298                         return err
1299                 }
1300         }
1301
1302         return nil
1303 }
1304
1305 type sampleInfo struct {
1306         id            string
1307         isCase        bool
1308         isControl     bool
1309         isTraining    bool
1310         isValidation  bool
1311         pcaComponents []float64
1312 }
1313
1314 // Read samples.csv file with case/control and training/validation
1315 // flags.
1316 func loadSampleInfo(samplesFilename string) ([]sampleInfo, error) {
1317         var si []sampleInfo
1318         f, err := open(samplesFilename)
1319         if err != nil {
1320                 return nil, err
1321         }
1322         buf, err := io.ReadAll(f)
1323         f.Close()
1324         if err != nil {
1325                 return nil, err
1326         }
1327         lineNum := 0
1328         for _, csv := range bytes.Split(buf, []byte{'\n'}) {
1329                 lineNum++
1330                 if len(csv) == 0 {
1331                         continue
1332                 }
1333                 split := strings.Split(string(csv), ",")
1334                 if len(split) < 4 {
1335                         return nil, fmt.Errorf("%d fields < 4 in %s line %d: %q", len(split), samplesFilename, lineNum, csv)
1336                 }
1337                 if split[0] == "Index" && split[1] == "SampleID" && split[2] == "CaseControl" && split[3] == "TrainingValidation" {
1338                         continue
1339                 }
1340                 idx, err := strconv.Atoi(split[0])
1341                 if err != nil {
1342                         if lineNum == 1 {
1343                                 return nil, fmt.Errorf("header does not look right: %q", csv)
1344                         }
1345                         return nil, fmt.Errorf("%s line %d: index: %s", samplesFilename, lineNum, err)
1346                 }
1347                 if idx != len(si) {
1348                         return nil, fmt.Errorf("%s line %d: index %d out of order", samplesFilename, lineNum, idx)
1349                 }
1350                 var pcaComponents []float64
1351                 if len(split) > 4 {
1352                         for _, s := range split[4:] {
1353                                 f, err := strconv.ParseFloat(s, 64)
1354                                 if err != nil {
1355                                         return nil, fmt.Errorf("%s line %d: cannot parse float %q: %s", samplesFilename, lineNum, s, err)
1356                                 }
1357                                 pcaComponents = append(pcaComponents, f)
1358                         }
1359                 }
1360                 si = append(si, sampleInfo{
1361                         id:            split[1],
1362                         isCase:        split[2] == "1",
1363                         isControl:     split[2] == "0",
1364                         isTraining:    split[3] == "1",
1365                         isValidation:  split[3] == "0" && len(split[2]) > 0, // fix errant 0s in input
1366                         pcaComponents: pcaComponents,
1367                 })
1368         }
1369         return si, nil
1370 }
1371
1372 func writeSampleInfo(samples []sampleInfo, outputDir string) error {
1373         fnm := outputDir + "/samples.csv"
1374         log.Infof("writing sample metadata to %s", fnm)
1375         f, err := os.Create(fnm)
1376         if err != nil {
1377                 return err
1378         }
1379         defer f.Close()
1380         pcaLabels := ""
1381         if len(samples) > 0 {
1382                 for i := range samples[0].pcaComponents {
1383                         pcaLabels += fmt.Sprintf(",PCA%d", i)
1384                 }
1385         }
1386         _, err = fmt.Fprintf(f, "Index,SampleID,CaseControl,TrainingValidation%s\n", pcaLabels)
1387         if err != nil {
1388                 return err
1389         }
1390         for i, si := range samples {
1391                 var cc, tv string
1392                 if si.isCase {
1393                         cc = "1"
1394                 } else if si.isControl {
1395                         cc = "0"
1396                 }
1397                 if si.isTraining {
1398                         tv = "1"
1399                 } else if si.isValidation {
1400                         tv = "0"
1401                 }
1402                 var pcavals string
1403                 for _, pcaval := range si.pcaComponents {
1404                         pcavals += fmt.Sprintf(",%f", pcaval)
1405                 }
1406                 _, err = fmt.Fprintf(f, "%d,%s,%s,%s%s\n", i, si.id, cc, tv, pcavals)
1407                 if err != nil {
1408                         return fmt.Errorf("write %s: %w", fnm, err)
1409                 }
1410         }
1411         err = f.Close()
1412         if err != nil {
1413                 return fmt.Errorf("close %s: %w", fnm, err)
1414         }
1415         log.Print("done")
1416         return nil
1417 }
1418
1419 func (cmd *sliceNumpy) filterHGVScolpair(colpair [2][]int8) bool {
1420         if cmd.chi2PValue >= 1 {
1421                 return true
1422         }
1423         col0 := make([]bool, 0, len(cmd.chi2Cases))
1424         col1 := make([]bool, 0, len(cmd.chi2Cases))
1425         cases := make([]bool, 0, len(cmd.chi2Cases))
1426         for i, c := range cmd.chi2Cases {
1427                 if colpair[0][i] < 0 {
1428                         continue
1429                 }
1430                 col0 = append(col0, colpair[0][i] != 0)
1431                 col1 = append(col1, colpair[1][i] != 0)
1432                 cases = append(cases, c)
1433         }
1434         return len(cases) >= cmd.minCoverage &&
1435                 (pvalue(col0, cases) <= cmd.chi2PValue || pvalue(col1, cases) <= cmd.chi2PValue)
1436 }
1437
1438 func writeNumpyUint32(fnm string, out []uint32, rows, cols int) error {
1439         output, err := os.Create(fnm)
1440         if err != nil {
1441                 return err
1442         }
1443         defer output.Close()
1444         bufw := bufio.NewWriterSize(output, 1<<26)
1445         npw, err := gonpy.NewWriter(nopCloser{bufw})
1446         if err != nil {
1447                 return err
1448         }
1449         log.WithFields(log.Fields{
1450                 "filename": fnm,
1451                 "rows":     rows,
1452                 "cols":     cols,
1453                 "bytes":    rows * cols * 4,
1454         }).Infof("writing numpy: %s", fnm)
1455         npw.Shape = []int{rows, cols}
1456         npw.WriteUint32(out)
1457         err = bufw.Flush()
1458         if err != nil {
1459                 return err
1460         }
1461         return output.Close()
1462 }
1463
1464 func writeNumpyInt32(fnm string, out []int32, rows, cols int) error {
1465         output, err := os.Create(fnm)
1466         if err != nil {
1467                 return err
1468         }
1469         defer output.Close()
1470         bufw := bufio.NewWriterSize(output, 1<<26)
1471         npw, err := gonpy.NewWriter(nopCloser{bufw})
1472         if err != nil {
1473                 return err
1474         }
1475         log.WithFields(log.Fields{
1476                 "filename": fnm,
1477                 "rows":     rows,
1478                 "cols":     cols,
1479                 "bytes":    rows * cols * 4,
1480         }).Infof("writing numpy: %s", fnm)
1481         npw.Shape = []int{rows, cols}
1482         npw.WriteInt32(out)
1483         err = bufw.Flush()
1484         if err != nil {
1485                 return err
1486         }
1487         return output.Close()
1488 }
1489
1490 func writeNumpyInt16(fnm string, out []int16, rows, cols int) error {
1491         output, err := os.Create(fnm)
1492         if err != nil {
1493                 return err
1494         }
1495         defer output.Close()
1496         bufw := bufio.NewWriterSize(output, 1<<26)
1497         npw, err := gonpy.NewWriter(nopCloser{bufw})
1498         if err != nil {
1499                 return err
1500         }
1501         log.WithFields(log.Fields{
1502                 "filename": fnm,
1503                 "rows":     rows,
1504                 "cols":     cols,
1505                 "bytes":    rows * cols * 2,
1506         }).Infof("writing numpy: %s", fnm)
1507         npw.Shape = []int{rows, cols}
1508         npw.WriteInt16(out)
1509         err = bufw.Flush()
1510         if err != nil {
1511                 return err
1512         }
1513         return output.Close()
1514 }
1515
1516 func writeNumpyInt8(fnm string, out []int8, rows, cols int) error {
1517         output, err := os.Create(fnm)
1518         if err != nil {
1519                 return err
1520         }
1521         defer output.Close()
1522         bufw := bufio.NewWriterSize(output, 1<<26)
1523         npw, err := gonpy.NewWriter(nopCloser{bufw})
1524         if err != nil {
1525                 return err
1526         }
1527         log.WithFields(log.Fields{
1528                 "filename": fnm,
1529                 "rows":     rows,
1530                 "cols":     cols,
1531                 "bytes":    rows * cols,
1532         }).Infof("writing numpy: %s", fnm)
1533         npw.Shape = []int{rows, cols}
1534         npw.WriteInt8(out)
1535         err = bufw.Flush()
1536         if err != nil {
1537                 return err
1538         }
1539         return output.Close()
1540 }
1541
1542 func allele2homhet(colpair [2][]int8) {
1543         a, b := colpair[0], colpair[1]
1544         for i, av := range a {
1545                 bv := b[i]
1546                 if av < 0 || bv < 0 {
1547                         // no-call
1548                         a[i], b[i] = -1, -1
1549                 } else if av > 0 && bv > 0 {
1550                         // hom
1551                         a[i], b[i] = 1, 0
1552                 } else if av > 0 || bv > 0 {
1553                         // het
1554                         a[i], b[i] = 0, 1
1555                 } else {
1556                         // ref (or a different variant in same position)
1557                         // (this is a no-op) a[i], b[i] = 0, 0
1558                 }
1559         }
1560 }
1561
1562 type onehotXref struct {
1563         tag     tagID
1564         variant tileVariantID
1565         hom     bool
1566         pvalue  float64
1567         maf     float64
1568 }
1569
1570 const onehotXrefSize = unsafe.Sizeof(onehotXref{})
1571
1572 // Build onehot matrix (m[tileVariantIndex][genome] == 0 or 1) for all
1573 // variants of a single tile/tag#.
1574 //
1575 // Return nil if no tile variant passes Χ² filter.
1576 func (cmd *sliceNumpy) tv2homhet(cgs map[string]CompactGenome, maxv tileVariantID, remap []tileVariantID, tag, chunkstarttag tagID, seq map[tagID][]TileVariant) ([][]int8, []onehotXref) {
1577         if tag == cmd.debugTag {
1578                 tv := make([]tileVariantID, len(cmd.cgnames)*2)
1579                 for i, name := range cmd.cgnames {
1580                         copy(tv[i*2:(i+1)*2], cgs[name].Variants[(tag-chunkstarttag)*2:])
1581                 }
1582                 log.WithFields(logrus.Fields{
1583                         "cgs[i].Variants[tag*2+j]": tv,
1584                         "maxv":                     maxv,
1585                         "remap":                    remap,
1586                         "tag":                      tag,
1587                         "chunkstarttag":            chunkstarttag,
1588                 }).Info("tv2homhet()")
1589         }
1590         if maxv < 1 || (maxv < 2 && !cmd.includeVariant1) {
1591                 // everyone has the most common variant (of the variants we don't drop)
1592                 return nil, nil
1593         }
1594         tagoffset := tag - chunkstarttag
1595         coverage := 0
1596         for _, cg := range cgs {
1597                 alleles := 0
1598                 for _, v := range cg.Variants[tagoffset*2 : tagoffset*2+2] {
1599                         if v > 0 && int(v) < len(seq[tag]) && len(seq[tag][v].Sequence) > 0 {
1600                                 alleles++
1601                         }
1602                 }
1603                 if alleles == 2 {
1604                         coverage++
1605                 }
1606         }
1607         if coverage < cmd.minCoverage {
1608                 return nil, nil
1609         }
1610         // "observed" array for p-value calculation (training set
1611         // only)
1612         obs := make([][]bool, (maxv+1)*2) // 2 slices (hom + het) for each variant#
1613         // one-hot output (all samples)
1614         outcols := make([][]int8, (maxv+1)*2)
1615         for i := range obs {
1616                 obs[i] = make([]bool, cmd.trainingSetSize)
1617                 outcols[i] = make([]int8, len(cmd.cgnames))
1618         }
1619         for cgid, name := range cmd.cgnames {
1620                 tsid := cmd.trainingSet[cgid]
1621                 cgvars := cgs[name].Variants[tagoffset*2:]
1622                 tv0, tv1 := remap[cgvars[0]], remap[cgvars[1]]
1623                 for v := tileVariantID(1); v <= maxv; v++ {
1624                         if tv0 == v && tv1 == v {
1625                                 if tsid >= 0 {
1626                                         obs[v*2][tsid] = true
1627                                 }
1628                                 outcols[v*2][cgid] = 1
1629                         } else if tv0 == v || tv1 == v {
1630                                 if tsid >= 0 {
1631                                         obs[v*2+1][tsid] = true
1632                                 }
1633                                 outcols[v*2+1][cgid] = 1
1634                         }
1635                 }
1636         }
1637         var onehot [][]int8
1638         var xref []onehotXref
1639         var maf float64
1640         for col := 2; col < len(obs); col++ {
1641                 // col 0,1 correspond to tile variant 0, i.e.,
1642                 // no-call; col 2,3 correspond to the most common
1643                 // variant; so we (normally) start at col 4.
1644                 if col < 4 && !cmd.includeVariant1 {
1645                         continue
1646                 }
1647                 if col&1 == 0 {
1648                         maf = homhet2maf(obs[col : col+2])
1649                         if cmd.pvalueMinFrequency < 1 && maf < cmd.pvalueMinFrequency {
1650                                 // Skip both columns (hom and het) if
1651                                 // allele frequency is below threshold
1652                                 col++
1653                                 continue
1654                         }
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                         maf:     maf,
1668                 })
1669         }
1670         return onehot, xref
1671 }
1672
1673 func homhet2maf(onehot [][]bool) float64 {
1674         if len(onehot[0]) == 0 {
1675                 return 0
1676         }
1677         n := 0
1678         for i := range onehot[0] {
1679                 if onehot[0][i] {
1680                         // hom
1681                         n += 2
1682                 } else if onehot[1][i] {
1683                         // het
1684                         n += 1
1685                 }
1686         }
1687         return float64(n) / float64(len(onehot[0])*2)
1688 }
1689
1690 // convert a []onehotXref with length N to a numpy-style []int32
1691 // matrix with N columns, one row per field of onehotXref struct.
1692 //
1693 // Hom/het row contains hom=0, het=1.
1694 //
1695 // P-value row contains 1000000x actual p-value.
1696 func onehotXref2int32(xrefs []onehotXref) []int32 {
1697         xcols := len(xrefs)
1698         xdata := make([]int32, 6*xcols)
1699         for i, xref := range xrefs {
1700                 xdata[i] = int32(xref.tag)
1701                 xdata[xcols+i] = int32(xref.variant)
1702                 if xref.hom {
1703                         xdata[xcols*2+i] = 1
1704                 }
1705                 xdata[xcols*3+i] = int32(xref.pvalue * 1000000)
1706                 xdata[xcols*4+i] = int32(-math.Log10(xref.pvalue) * 1000000)
1707                 xdata[xcols*5+i] = int32(xref.maf * 1000000)
1708         }
1709         return xdata
1710 }
1711
1712 // transpose onehot data from in[col][row] to numpy-style
1713 // out[row*cols+col].
1714 func onehotcols2int8(in [][]int8) []int8 {
1715         if len(in) == 0 {
1716                 return nil
1717         }
1718         cols := len(in)
1719         rows := len(in[0])
1720         out := make([]int8, rows*cols)
1721         for row := 0; row < rows; row++ {
1722                 outrow := out[row*cols:]
1723                 for col, incol := range in {
1724                         outrow[col] = incol[row]
1725                 }
1726         }
1727         return out
1728 }
1729
1730 // Return [2][]uint32{rowIndices, colIndices} indicating which
1731 // elements of matrixT[c][r] have non-zero values.
1732 func onehotChunk2Indirect(matrixT [][]int8) [2][]uint32 {
1733         var nz [2][]uint32
1734         for c, col := range matrixT {
1735                 for r, val := range col {
1736                         if val != 0 {
1737                                 nz[0] = append(nz[0], uint32(r))
1738                                 nz[1] = append(nz[1], uint32(c))
1739                         }
1740                 }
1741         }
1742         return nz
1743 }