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