slice.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         "encoding/gob"
  10         "errors"
  11         "flag"
  12         "fmt"
  13         "io"
  14         "net/http"
  15         _ "net/http/pprof"
  16         "os"
  17         "path/filepath"
  18         "runtime"
  19         "strings"
  20         "sync"
  21         "sync/atomic"
  22
  23         "git.arvados.org/arvados.git/sdk/go/arvados"
  24         "github.com/klauspost/pgzip"
  25         log "github.com/sirupsen/logrus"
  26 )
  27
  28 type slicecmd struct{}
  29
  30 func (cmd *slicecmd) RunCommand(prog string, args []string, stdin io.Reader, stdout, stderr io.Writer) int {
  31         var err error
  32         defer func() {
  33                 if err != nil {
  34                         fmt.Fprintf(stderr, "%s\n", err)
  35                 }
  36         }()
  37         flags := flag.NewFlagSet("", flag.ContinueOnError)
  38         flags.SetOutput(stderr)
  39         pprof := flags.String("pprof", "", "serve Go profile data at http://`[addr]:port`")
  40         runlocal := flags.Bool("local", false, "run on local host (default: run in an arvados container)")
  41         projectUUID := flags.String("project", "", "project `UUID` for output data")
  42         priority := flags.Int("priority", 500, "container request priority")
  43         preemptible := flags.Bool("preemptible", true, "request preemptible instance")
  44         outputDir := flags.String("output-dir", "./out", "output `directory`")
  45         tagsPerFile := flags.Int("tags-per-file", 50000, "tags per file (nfiles will be ~10M÷x)")
  46         err = flags.Parse(args)
  47         if err == flag.ErrHelp {
  48                 err = nil
  49                 return 0
  50         } else if err != nil {
  51                 return 2
  52         }
  53         inputDirs := flags.Args()
  54         if len(inputDirs) == 0 {
  55                 err = errors.New("no input dirs specified")
  56                 return 2
  57         }
  58
  59         if *pprof != "" {
  60                 go func() {
  61                         log.Println(http.ListenAndServe(*pprof, nil))
  62                 }()
  63         }
  64
  65         if !*runlocal {
  66                 runner := arvadosContainerRunner{
  67                         Name:        "lightning slice",
  68                         Client:      arvados.NewClientFromEnv(),
  69                         ProjectUUID: *projectUUID,
  70                         RAM:         500000000000,
  71                         VCPUs:       64,
  72                         Priority:    *priority,
  73                         KeepCache:   2,
  74                         APIAccess:   true,
  75                         Preemptible: *preemptible,
  76                 }
  77                 for i := range inputDirs {
  78                         err = runner.TranslatePaths(&inputDirs[i])
  79                         if err != nil {
  80                                 return 1
  81                         }
  82                 }
  83                 runner.Args = append([]string{"slice", "-local=true",
  84                         "-pprof", ":6060",
  85                         "-output-dir", "/mnt/output",
  86                 }, inputDirs...)
  87                 var output string
  88                 output, err = runner.Run()
  89                 if err != nil {
  90                         return 1
  91                 }
  92                 fmt.Fprintln(stdout, output)
  93                 return 0
  94         }
  95
  96         err = Slice(*tagsPerFile, *outputDir, inputDirs)
  97         if err != nil {
  98                 return 1
  99         }
 100         return 0
 101 }
 102
 103 // Read tags+tiles+genomes from srcdir, write to dstdir with (up to)
 104 // the specified number of tags per file.
 105 func Slice(tagsPerFile int, dstdir string, srcdirs []string) error {
 106         var infiles []string
 107         for _, srcdir := range srcdirs {
 108                 files, err := allFiles(srcdir, matchGobFile)
 109                 if err != nil {
 110                         return err
 111                 }
 112                 infiles = append(infiles, files...)
 113         }
 114         // dirNamespace[dir] is an int in [0,len(dirNamespace)), used below to
 115         // namespace variant numbers from different dirs.
 116         dirNamespace := map[string]tileVariantID{}
 117         for _, path := range infiles {
 118                 dir, _ := filepath.Split(path)
 119                 if _, ok := dirNamespace[dir]; !ok {
 120                         dirNamespace[dir] = tileVariantID(len(dirNamespace))
 121                 }
 122         }
 123         namespaces := tileVariantID(len(dirNamespace))
 124
 125         var (
 126                 tagset     [][]byte
 127                 tagsetOnce sync.Once
 128                 fs         []*os.File
 129                 bufws      []*bufio.Writer
 130                 gzws       []*pgzip.Writer
 131                 encs       []*gob.Encoder
 132
 133                 countTileVariants int64
 134                 countGenomes      int64
 135                 countReferences   int64
 136         )
 137
 138         throttle := throttle{Max: runtime.GOMAXPROCS(0)}
 139         for _, infile := range infiles {
 140                 infile := infile
 141                 throttle.Go(func() error {
 142                         f, err := open(infile)
 143                         if err != nil {
 144                                 return err
 145                         }
 146                         defer f.Close()
 147                         dir, _ := filepath.Split(infile)
 148                         namespace := dirNamespace[dir]
 149                         log.Printf("reading %s (namespace %d)", infile, namespace)
 150                         return DecodeLibrary(f, strings.HasSuffix(infile, ".gz"), func(ent *LibraryEntry) error {
 151                                 if err := throttle.Err(); err != nil {
 152                                         return err
 153                                 }
 154                                 if len(ent.TagSet) > 0 {
 155                                         tagsetOnce.Do(func() {
 156                                                 tagset = ent.TagSet
 157                                                 var err error
 158                                                 fs, bufws, gzws, encs, err = openOutFiles(dstdir, len(ent.TagSet), tagsPerFile)
 159                                                 if err != nil {
 160                                                         throttle.Report(err)
 161                                                         return
 162                                                 }
 163                                                 for _, enc := range encs {
 164                                                         err = enc.Encode(LibraryEntry{TagSet: tagset})
 165                                                         if err != nil {
 166                                                                 throttle.Report(err)
 167                                                                 return
 168                                                         }
 169                                                 }
 170                                         })
 171                                 }
 172                                 if err := throttle.Err(); err != nil {
 173                                         return err
 174                                 }
 175                                 atomic.AddInt64(&countTileVariants, int64(len(ent.TileVariants)))
 176                                 for _, tv := range ent.TileVariants {
 177                                         tv.Variant = tv.Variant*namespaces + namespace
 178                                         fileno := 0
 179                                         if !tv.Ref {
 180                                                 fileno = int(tv.Tag) / tagsPerFile
 181                                         }
 182                                         err := encs[fileno].Encode(LibraryEntry{
 183                                                 TileVariants: []TileVariant{tv},
 184                                         })
 185                                         if err != nil {
 186                                                 return err
 187                                         }
 188                                 }
 189                                 // Here, each output file gets a
 190                                 // CompactGenome entry for each
 191                                 // genome, even if there are no
 192                                 // variants in the relevant range.
 193                                 // Easier for downstream code.
 194                                 atomic.AddInt64(&countGenomes, int64(len(ent.CompactGenomes)))
 195                                 for _, cg := range ent.CompactGenomes {
 196                                         for i, v := range cg.Variants {
 197                                                 if v > 0 {
 198                                                         cg.Variants[i] = v*namespaces + namespace
 199                                                 }
 200                                         }
 201                                         for i, enc := range encs {
 202                                                 start := i * tagsPerFile
 203                                                 end := start + tagsPerFile
 204                                                 if max := len(cg.Variants)/2 + int(cg.StartTag); end > max {
 205                                                         end = max
 206                                                 }
 207                                                 if start < int(cg.StartTag) {
 208                                                         start = int(cg.StartTag)
 209                                                 }
 210                                                 var variants []tileVariantID
 211                                                 if start < end {
 212                                                         variants = cg.Variants[(start-int(cg.StartTag))*2 : (end-int(cg.StartTag))*2]
 213                                                 }
 214                                                 err := enc.Encode(LibraryEntry{CompactGenomes: []CompactGenome{{
 215                                                         Name:     cg.Name,
 216                                                         Variants: variants,
 217                                                         StartTag: tagID(start),
 218                                                         EndTag:   tagID(start + tagsPerFile),
 219                                                 }}})
 220                                                 if err != nil {
 221                                                         return err
 222                                                 }
 223                                         }
 224                                 }
 225                                 // Write all ref seqs to the first
 226                                 // slice. Easier for downstream code.
 227                                 atomic.AddInt64(&countReferences, int64(len(ent.CompactSequences)))
 228                                 if len(ent.CompactSequences) > 0 {
 229                                         for _, cs := range ent.CompactSequences {
 230                                                 for _, tseq := range cs.TileSequences {
 231                                                         for i, libref := range tseq {
 232                                                                 tseq[i].Variant = libref.Variant*namespaces + namespace
 233                                                         }
 234                                                 }
 235                                         }
 236                                         err := encs[0].Encode(LibraryEntry{CompactSequences: ent.CompactSequences})
 237                                         if err != nil {
 238                                                 return err
 239                                         }
 240                                 }
 241                                 return nil
 242                         })
 243                 })
 244         }
 245         throttle.Wait()
 246         if throttle.Err() != nil {
 247                 closeOutFiles(fs, bufws, gzws, encs)
 248                 return throttle.Err()
 249         }
 250         defer log.Printf("Total %d tile variants, %d genomes, %d reference sequences", countTileVariants, countGenomes, countReferences)
 251         return closeOutFiles(fs, bufws, gzws, encs)
 252 }
 253
 254 func openOutFiles(dstdir string, tags, tagsPerFile int) (fs []*os.File, bufws []*bufio.Writer, gzws []*pgzip.Writer, encs []*gob.Encoder, err error) {
 255         nfiles := (tags + tagsPerFile - 1) / tagsPerFile
 256         fs = make([]*os.File, nfiles)
 257         bufws = make([]*bufio.Writer, nfiles)
 258         gzws = make([]*pgzip.Writer, nfiles)
 259         encs = make([]*gob.Encoder, nfiles)
 260         for i := 0; i*tagsPerFile < tags; i++ {
 261                 fs[i], err = os.Create(dstdir + fmt.Sprintf("/library%04d.gob.gz", i))
 262                 if err != nil {
 263                         return
 264                 }
 265                 bufws[i] = bufio.NewWriterSize(fs[i], 1<<26)
 266                 gzws[i] = pgzip.NewWriter(bufws[i])
 267                 encs[i] = gob.NewEncoder(gzws[i])
 268         }
 269         return
 270 }
 271
 272 func closeOutFiles(fs []*os.File, bufws []*bufio.Writer, gzws []*pgzip.Writer, encs []*gob.Encoder) error {
 273         var firstErr error
 274         for _, gzw := range gzws {
 275                 if gzw != nil {
 276                         err := gzw.Close()
 277                         if err != nil && firstErr == nil {
 278                                 firstErr = err
 279                         }
 280                 }
 281         }
 282         for _, bufw := range bufws {
 283                 if bufw != nil {
 284                         err := bufw.Flush()
 285                         if err != nil && firstErr == nil {
 286                                 firstErr = err
 287                         }
 288                 }
 289         }
 290         for _, f := range fs {
 291                 if f != nil {
 292                         err := f.Close()
 293                         if err != nil && firstErr == nil {
 294                                 firstErr = err
 295                         }
 296                 }
 297         }
 298         return firstErr
 299 }