Press n or j to go to the next uncovered block, b, p or k for the previous block.
| 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 | 2x 2x 2x 2x 2x 68x 68x 68x 68x 68x 68x 68x 68x 68x 68x 68x 68x 68x 68x 68x 68x 68x 185x 1x 1x 1x 184x 184x 184x 344x 344x 344x 344x 344x 344x 344x 344x 344x 344x 185x 103x 103x 103x 103x 103x 185x 81x 242x 242x 242x 241x 241x 241x 242x 242x 80x 185x 183x 185x 60x 60x 60x 123x 123x 123x 185x 185x 596458x 596458x 596458x 596458x 596458x 596458x 123x 123x 123x 123x 123x 185x 185x 185x 185x 185x 67x 67x 2x 71x 71x 71x 71x 71x 8x 71x 7x 7x 7x 7x 7x 7x 7x 2x 54x 54x 54x 1x 1x 1x 53x 54x 2x 2x 2x 54x 51x 54x 1x 1x 1x 54x 54x 54x 54x 54x 1x 1x 1x 49x 49x 49x 54x 3x 3x 3x 46x 46x 46x 46x 46x 46x 46x 46x 46x 46x 46x 46x 46x 6x 40x 54x 54x 54x 54x 54x 54x 54x 34x 12x 12x 23x 12x 12x 54x 46x 46x 54x 2x 2x 2x 2x 2x 2x 54x 54x 54x 3x 3x 2x 1x 1x 41x 41x 54x 54x 54x 54x 54x 54x 54x 44x 44x 46x 46x 4x 4x 46x 4x 4x 4x 44x 44x 44x 44x 44x 44x 44x 44x 44x 44x 44x 44x 44x 44x 44x 44x 44x 44x 44x 44x 44x 44x 44x 44x 44x 8x 8x 8x 8x 8x 36x 36x 36x 36x 36x 36x 36x 36x 36x 36x 36x 2x 81x 81x 81x 81x 230x 230x 81x 81x | /**
* The motif search, run — in the annotations worker, beside the residues.
*
* The whole shape of this is set by where the bytes are. A reference is parsed
* into the annotations worker and deliberately never leaves it whole: `bases`
* answers a window and is capped, precisely so that opening a genome does not
* put twelve megabytes on the main thread. So the scan happens here, against
* `TrackJobContext.residues`, and what goes back is the handful of coordinates
* that came out — a `FeatureSet` and a vector, both of which cross a thread
* boundary as typed arrays.
*
* ## Two outputs, because a set of sites is two different pictures
*
* `annotations/motifTrack.ts` argues it in full and both builders are its. A
* `FeatureSet` answers *where exactly*, and at chromosome scale it is a picket
* fence. A per-bin count answers *how dense here*, and at base scale it is a bar
* chart of ones. Neither is both, so one scan makes both, and the reader closes
* whichever they do not want with the same × as every other row.
*
* ## The two decisions that are biology rather than plumbing
*
* Both are inherited and neither may be lost in the move from the card:
*
* - **A reference `N` does not satisfy a pattern naming a base.** An assembly
* gap is a base nobody knows rather than a base that is any of four, so
* `annotations/motif.ts` matches on a subset test; without it a run of ten
* thousand `N`s is ten thousand fictional hits and a draft assembly is a solid
* bar of them.
* - **A palindrome searched on both strands is reported once**, with `strand: 0`
* rather than twice at identical coordinates — otherwise a reader counting the
* sites of `GAATTC` gets double the answer with nothing on screen to explain
* it.
*
* And a third that belongs to the density lane: **an empty bin is a zero, never
* `NaN`**. `NaN` in a `ProfileOutput` means *nothing is known here* and draws as
* a hole, which for a count is the opposite claim — a stretch with no sites in
* it is the most informative thing a restriction map has to say.
*/
import {
compileMotif,
findMotif,
isMotifRefusal,
motifOverlap,
type Motif,
type MotifHit,
} from '../../annotations/motif'
import {
motifDensity,
motifFeatures,
motifLabel,
type MotifHits,
type MotifScanResult,
} from '../../annotations/motifTrack'
import type { Contig } from '../../genome/assembly'
import { contigIndexOf } from '../../genome/naming'
import {
boolParam,
idParam,
numParam,
tickedSequences,
type SequenceReader,
type TrackJobContext,
type TrackTaskRunner,
} from '../job'
import { MOTIF_PIECE, MOTIF_SITE_LIMIT, motifPattern } from '../tasks/motif'
import type { IntervalsOutput, ProfileOutput, TaskOutput } from '../types'
/** What {@link scanMotif} is asked: the search, and where it may report its progress. */
export interface MotifScanOptions {
bothStrands: boolean
limit: number
mismatches?: number
/**
* The reference's sequences to search, in its own order; every one where
* absent (see {@link searchedContigs}).
*/
contigs?: readonly Contig[]
/** Residues searched so far out of all there are to search, after each piece. */
progress?: (done: number, total: number, contig: string) => void
/** Called between two pieces, so a Stop is seen without waiting for the end. */
throwIfCancelled?: () => void
/**
* Residues per piece, {@link MOTIF_PIECE} where absent: a test's way to put
* seams where it can see them without carrying sixteen megabases to do it.
*/
piece?: number
}
/**
* Every occurrence of a motif in one loaded reference.
*
* Folded in here rather than left in the worker as the `motif` request's
* handler, which is where it was: that request had exactly one caller, the
* bespoke card, and with the card gone a function in the worker called only from
* this module would be an indirection with nothing on the other end of it. What
* stays in the worker is the part only the worker can do — holding the store and
* resolving a name onto it ({@link SequenceReader}).
*
* In sequence order, one entry per contig that has a hit at all — a contig with
* none is omitted rather than carried as three empty arrays, since a draft
* assembly is mostly contigs with no hit in them and the empty entries would
* outnumber the real ones.
*
* The walk is over the reader's `contigs` rather than over its bytes, so the
* order is the reference's own; `bases` then resolves the spelling, which is
* what searches a FASTA renamed into the map's vocabulary under the names the
* rest of the application uses (§5.9).
*
* **A reference read by range is searched through `read`**, in pieces of
* {@link MOTIF_PIECE} residues that overlap by `motifOverlap` — one base short
* of the motif, which is exactly enough for a site straddling a seam to lie
* whole in one piece, and never in both: a piece reading `[a, b + overlap)`
* holds every site starting in `[a, b)` and no other. So the answer is the
* one a held reference gives, site for site, and only the memory differs.
*/
export async function scanMotif(
reader: SequenceReader,
motif: Motif,
options: MotifScanOptions,
): Promise<MotifScanResult> {
// The same ceiling {@link MOTIF_SITE_LIMIT} names and `runMotif` has already
// applied, kept here as the last line of defence for a caller that reaches
// this function directly. It must not be the *only* place it is applied: when
// it was, a run asked for two million sites capped at five hundred thousand
// and then reported the two million back, which is the one thing a truncation
// note must never do.
const cap = Math.max(1, Math.min(options.limit, MOTIF_SITE_LIMIT))
const budget = options.mismatches ?? 0
const contigs = options.contigs ?? reader.contigs
const overlap = motifOverlap(motif)
const read = reader.read?.bind(reader)
const step = Math.max(1, Math.floor(options.piece ?? MOTIF_PIECE))
const hits: MotifHits[] = []
let total = 0
let truncated = false
const all = contigs.reduce((sum, contig) => sum + contig.length, 0)
let done = 0
for (const contig of contigs) {
if (total >= cap) {
// Stopped with sequences left unread, which is a truncated answer even
// though the last contig searched finished cleanly.
truncated = true
break
}
const found: MotifHit[] = []
let full = false
// One more than there is room for, and then the extra is dropped. Asking for
// exactly the remaining budget cannot tell "the cap fell on the last site
// there was" from "the cap cut the answer short" — `findMotif` stops on the
// site that fills the budget and has no way to know whether another follows
// without looking. Looking costs one more site's worth of scan and buys the
// difference between a complete answer and a claim of truncation on one.
const search = (bases: Uint8Array, offset: number): void => {
const room = cap - total - found.length
const piece = findMotif(bases, motif, {
bothStrands: options.bothStrands,
limit: room + 1,
mismatches: budget,
offset,
})
for (let at = 0; at < Math.min(piece.hits.length, room); at++) found.push(piece.hits[at])
if (piece.truncated || piece.hits.length > room) full = true
}
if (read === undefined) {
const bases = reader.bases(contig.name)
if (bases === undefined) continue
search(bases, 0)
done += contig.length
options.progress?.(done, all, contig.name)
} else {
for (let from = 0; from < contig.length && !full; from += step) {
options.throwIfCancelled?.()
const to = Math.min(contig.length, from + step)
const bases = await read(contig.name, from, Math.min(contig.length, to + overlap))
if (bases === undefined) break
search(bases, from)
done += to - from
options.progress?.(done, all, contig.name)
}
}
if (full) truncated = true
const kept = found.length
if (kept === 0) {
if (full) break
continue
}
const starts = new Int32Array(kept)
const ends = new Int32Array(kept)
const strands = new Int8Array(kept)
// Carried only where the search allowed any, so an exact search's sites
// keep the empty score column `motifFeatures` argues for.
const wrong = budget > 0 ? new Uint8Array(kept) : undefined
for (let at = 0; at < kept; at++) {
const hit = found[at]
starts[at] = hit.start
ends[at] = hit.end
strands[at] = hit.strand
if (wrong !== undefined) wrong[at] = hit.mismatches
}
hits.push({
contig: contig.name,
starts,
ends,
strands,
...(wrong === undefined ? {} : { mismatches: wrong }),
})
total += kept
if (full) break
}
return { hits, total, truncated }
}
/**
* The reference's sequences a run searches, in the reference's own order.
*
* The `sequences` parameter names them in the session's vocabulary, and the
* reference is reached through `contigIndexOf`, so a FASTA that writes `chr1`
* is searched for a map that writes `1` (§5.9). **Every sequence of the session
* ticked is the whole reference**, sequences the map has not got included —
* which is what the search did before it had a list, and what *all of them*
* means to a reader who has not unticked a box; a list short of that is exactly
* the sequences on it. Unanswered (`null`, a job that never met a form) is the
* whole reference too.
*/
export function searchedContigs(
reader: SequenceReader,
ticked: readonly string[] | null,
session: readonly string[],
): Contig[] {
if (ticked === null) return [...reader.contigs]
const chosen = new Set(ticked)
if (session.length > 0 && session.every((name) => chosen.has(name))) return [...reader.contigs]
const wanted = new Set<number>()
for (const name of ticked) {
const at = contigIndexOf(reader.contigs, name)
if (at >= 0) wanted.add(at)
}
return reader.contigs.filter((_, at) => wanted.has(at))
}
export const runMotif: TrackTaskRunner = async (job) => {
const entry = idParam(job.params, 'sequence')
// The capability, not the entry: a context with no `residues` at all is a host
// that holds no genome, which is a different sentence from a reference the
// reader has since dropped. Both are notes rather than throws — a parameter
// that cannot be honoured is a fact about the form, and §18.3's red card is
// for something having gone wrong (`analysis/run/metaplot.ts` does the same).
if (job.residues === undefined) {
job.note('This session cannot read residues, so there is no reference to search.')
return []
}
if (entry === null) {
job.note('No reference was chosen. Load a FASTA and pick it under Reference.')
return []
}
const reader = job.residues(entry)
if (reader === undefined) {
job.note('The reference chosen is no longer loaded. Drop the FASTA again, or pick another.')
return []
}
// A reference read by range refuses to be read *whole*, and a search does not
// need it whole: it reads a piece at a time through `read`. The refusal is
// for a host that offers neither.
if (reader.read === undefined && reader.refusal !== undefined) {
job.note(reader.refusal)
return []
}
// Unanswered is every sequence and an empty list is nothing (§18.2), as for
// every task with this list; the estimate refuses the second before it is
// sent, and this is what one that arrives anyway hears.
const ticked = tickedSequences(job.params, 'sequences')
if (ticked !== null && ticked.length === 0) {
job.note('No sequence is ticked, so there is nothing to search.')
return []
}
const session = Object.entries(job.inputs.contigLengths ?? {})
.filter(([, length]) => length > 0)
.map(([name]) => name)
const searched = searchedContigs(reader, ticked, session)
if (searched.length === 0) {
job.note('None of the ticked sequences is in this reference, so there is nothing to search.')
return []
}
const whole = searched.length === reader.contigs.length
const pattern = motifPattern(job.params)
const compiled = compileMotif(pattern)
// **A pattern that will not compile is not an exception.** `compileMotif`
// hands back a sentence about what the reader typed — "“X” is not a IUPAC
// base" — and a note is where a sentence like that belongs: the drawer lists
// the notes under the run it finished, where a throw would show a red card for
// a typo the reader can fix in the box above it (§18.3, and `run/metaplot.ts`
// answers a missing input the same way). The *form* has usually said it
// already, through this same function in the task's estimate, with **Run**
// greyed out; this is the path a stored recipe and a batch arrive by, where
// there was no form to say it.
if (isMotifRefusal(compiled)) {
job.note(compiled.reason)
return []
}
const bothStrands = boolParam(job.params, 'bothStrands', true)
// Clamped to the ceiling here and not only inside `scanMotif`, because this is
// the number every sentence below quotes. The spec carries the same `max`, so
// a form can no longer submit a larger one — but a stored recipe written
// before it did, a batch and the CLI never pass through a form, and a note
// naming a limit the scan never used is a note that sends the reader to raise
// a control which cannot help them.
const limit = Math.min(
MOTIF_SITE_LIMIT,
Math.max(1, Math.round(numParam(job.params, 'limit', MOTIF_SITE_LIMIT))),
)
// Never as many as the motif is long, which would match every window; the
// matcher holds the same line, and this is the number the label quotes.
const mismatches = Math.max(
0,
Math.min(compiled.length - 1, Math.round(numParam(job.params, 'mismatches', 0))),
)
job.progress(
0,
1,
searched.length === 1
? `searching ${searched[0].name}`
: `searching ${searched.length.toLocaleString()} sequences`,
)
const result = await scanMotif(reader, compiled, {
bothStrands,
limit,
mismatches,
contigs: searched,
// Only a reference read by range reports as it goes: a held one is one
// pass over memory, done before a bar could be drawn.
...(reader.read === undefined
? {}
: {
progress: (done: number, total: number, contig: string) =>
job.progress(done, total, `searching ${contig}`),
throwIfCancelled: () => job.throwIfCancelled(),
}),
})
job.progress(1, 1, 'done')
const label = scopedLabel(motifLabel(pattern, bothStrands, mismatches), searched, whole)
if (result.total === 0) {
job.note(
whole
? `No match for ${label} anywhere in this reference.`
: `No match for ${label}.`,
)
return []
}
const sites = `${result.total.toLocaleString()} site${result.total === 1 ? '' : 's'}`
// Two remedies, because only one of them is ever true. Below the ceiling the
// reader owns the limit and raising it finds more; *at* the ceiling there is
// nothing to raise — 500,000 is the most one run reports — and the old
// sentence prescribed the raise anyway, so a reader who followed it got the
// identical answer back with the identical note under it and no way to learn
// why. `limit` is the clamped number, so the figure quoted is the one the
// scan actually stopped at.
job.note(
result.truncated
? `${sites} — the search stopped at the limit of ${limit.toLocaleString()} and there are ` +
'more. ' +
(limit < MOTIF_SITE_LIMIT
? 'Raise “Most sites”, or narrow the pattern.'
: searched.length > 1
? 'That is the most one run reports, so narrow the pattern or tick fewer sequences.'
: 'That is the most one run reports, so narrow the pattern.')
: `${sites}, on ${result.hits.length.toLocaleString()} ` +
`${result.hits.length === 1 ? 'sequence' : 'sequences'}.`,
)
// One id for the pair, minted once. It carries the pattern **and** the strand
// because two searches in one session are two different answers, and one id
// for both would make the second silently replace the first — the workspace
// keeps one row per output id, on purpose. Re-running the *same* search still
// lands on the rows it made last time, which is what a reader refining a
// pattern wants and what **Recompute** depends on. Letters and digits only by
// construction: `motifPattern` dropped the whitespace and upper-cased it, and
// `compileMotif` accepted every character that is left. A search over some of
// the sequences carries them too, for the same reason: chr1 searched and then
// chr2 are two answers, and the second must not take the first one's rows.
const id =
`motif-${pattern}${bothStrands ? '-both' : ''}${mismatches > 0 ? `-mm${mismatches}` : ''}` +
(whole ? '' : scopeId(searched))
const outputs: TaskOutput[] = [sitesOutput(result, pattern, label, id)]
const density = densityOutput(job, result, label, id)
if (density !== null) outputs.push(density)
return outputs
}
/**
* A run's label, with the sequences it searched where they were not all of them.
*
* Named outright up to three — *GATC (both strands) on chr1* is the row a
* reader searching chromosome by chromosome needs to tell from the next one —
* and counted past that, where a list would be longer than the lane's name.
*/
function scopedLabel(label: string, searched: readonly Contig[], whole: boolean): string {
if (whole) return label
return searched.length <= 3
? `${label} on ${searched.map((contig) => contig.name).join(', ')}`
: `${label} on ${searched.length.toLocaleString()} sequences`
}
/**
* The sequences a narrowed search covered, as a suffix of its output id.
*
* Letters and digits from each name up to three, as the label names them, and
* past that the count and a checksum of the list — two different sets of five
* sequences are two answers, and the count alone would file them as one.
*/
function scopeId(searched: readonly Contig[]): string {
const names = searched.map((contig) => contig.name)
if (names.length <= 3) return `-on-${names.map((name) => name.replace(/[^A-Za-z0-9]/g, '')).join('-')}`
let sum = 5381
for (const char of names.join('\n')) sum = ((sum * 33) ^ char.charCodeAt(0)) >>> 0
return `-on${names.length}-${sum.toString(36)}`
}
/**
* The sites, as an interval set.
*
* Straight onto an entry from here: a `FeatureSet` is what a parsed BED would
* have produced, so it becomes an ordinary row of the Tracks card — a 1D track,
* or a set of points on the map — with the same colour, the same controls and
* the same **Save as BED** as a file the reader dropped
* (`AnalysisWorkspace.adoptIntervals`). The names and the empty score column are
* `motifFeatures`' decisions and are argued there.
*/
function sitesOutput(
result: MotifScanResult,
pattern: string,
label: string,
id: string,
): IntervalsOutput {
return {
kind: 'intervals',
id,
label,
payload: { kind: 'features', data: motifFeatures(result, pattern) },
noun: { one: 'site', many: 'sites' },
// Places: a track beside the map, and a point on it where the reader puts
// them there (§18.7).
display: 'lane',
look: 'point',
}
}
/**
* Sites per bin, on the session's **own** bin grid.
*
* This is the one decision in the task that is not obvious, and getting it wrong
* produces a lane that is drawn in the wrong places rather than one that is
* missing. The bespoke card chose a bin size from the *size of the genome* — a
* `motifBinsize` that aimed at twenty thousand bins and snapped to a round
* number — and for the card that was right: it laid the vector out over the
* assembly itself, so it owned both halves of the grid and they agreed by
* construction. That function is gone with the card, because a task has no use
* for it.
*
* A task does not own both halves. `AnalysisWorkspace.register` takes the
* `binsize` off this output and pairs it with the *session's* contig grid —
* `startBin` and `bins` per sequence, at the open file's resolution — and
* `binVectorSamples` then reads a value at `contig.startBin + floor(bp /
* binsize)`. Hand it a bin size the grid was not built with and every value
* lands at the wrong multiple of it: a 1 kb density registered against a 5 kb
* grid draws each chromosome's sites compressed into the first fifth of it, on
* the wrong chromosome for all but the first. So the bin size **is the
* framing's**, always, and the grid below is rebuilt on exactly the rule
* `analysis/context.ts` builds the session's with.
*
* The cost is a coarser picture than the card's on a coarse map, and that is the
* right trade: a lane on a private grid is a lane that cannot be compared with
* the insulation track above it or the coverage track below it, which is most of
* what a lane is for (§18.5).
*
* **Where there is no grid there is no lane, and the run says so.** A 1D browser
* session has no map, so `AnalysisFraming.binsize` is `0` — there is no
* resolution to be on — and the sites lane above, which carries its own
* coordinates, is the whole of what a browser session gets.
*/
function densityOutput(
job: TrackJobContext,
result: MotifScanResult,
label: string,
id: string,
): ProfileOutput | null {
const binsize = job.framing.binsize
const contigs = gridContigs(job)
if (!(binsize > 0) || contigs.length === 0) {
job.note(
'The sites are a lane; the density is not. A per-bin track needs the bin grid of an open ' +
'contact map, and this session has none — so there is nothing for a density to be ' +
'per. Open the map and run it again for that half.',
)
return null
}
// `motifDensity` lays the grid out itself, on the same rule and in the same
// order — a bin per `ceil(length / binsize)`, cumulative — so the vector and
// the grid the workspace registers it against agree bin for bin.
const vector = motifDensity(result, contigs, binsize)
return {
kind: 'profile',
id: `${id}-density`,
label: `${label} density`,
values: vector.values,
extent: [0, vector.values.length],
binsize,
// Sites, and a count rather than a rate: the bin size is fixed across the
// lane, so dividing by it would scale every value by the same number and
// lose the property that the lane's values sum to the site count.
unit: 'sites',
}
}
/**
* The sequences the session's bin grid is laid out over, in its order.
*
* From `inputs.contigLengths`, which is the only assembly this venue is given:
* the annotations worker holds handles and parsed sets, and neither of those is
* an assembly. It is built by `AnalysisWorkspace` from the very list its `grid`
* walks, so the order here is the order the startBins were computed in — which
* is the property this whole function turns on, and the reason it is not
* rebuilt from the reference's own contigs: a FASTA lists its sequences in its
* own order, and a grid built in that order would put every contig after the
* first disagreement at the wrong absolute bin.
*
* Zero-length sequences are dropped, because the session's grid gives them no
* bins at all (`binsForLength`) while `motifDensity` gives every sequence at
* least one — a divergence of one bin that would shift everything after them.
*
* The one thing a record cannot promise is its own order for a **numeric** name:
* `Object.entries` puts canonical integer keys — `1`, `2`, `10`, which is how a
* `.hic` spells its chromosomes — ascending and ahead of the rest, whatever
* order they were inserted in. That is the same order every such file lists them
* in, so the two agree in practice; a file that listed `10` before `2` would be
* the one shape this cannot follow, and the density lane is what would be wrong
* rather than the sites. Written down rather than guarded, because the guard
* would need a fact this venue is not given: the startBins themselves.
*/
export function gridContigs(job: TrackJobContext): Contig[] {
const lengths = job.inputs.contigLengths
if (lengths === undefined) return []
const out: Contig[] = []
for (const [name, length] of Object.entries(lengths)) {
if (length > 0) out.push({ name, length })
}
return out
}
|