test(tests): R4.3a browser re-measure of large files, leak series, WebKit repetition (stream sink, BYOB)

Streaming into FileBytes removes the file's JavaScript copy in Chromium and Firefox
(50 MiB: 65 and 67 MiB summed growth against 56 and 79 for the TypeScript); WebKit keeps
about 180 MiB for any file read inside a worker (272 against 76), a main-thread BYOB read
feeding the worker's sink measured at 106. Corpus 666/678 on the stream path. Leak series
(3 engines × 3 documents × 3 repetitions): Chromium ≤ 0.5 MiB over the last 5, Firefox
within ±3 (idle only), WebKit copy +13 then level (12-import series −2.3). 20 dense
imports in WebKit: no worker error event.
This commit is contained in:
bchanot committed 2026-10-18 18:50:47 +02:00
1 parent 8924d50edf
commit 49591c53cd
14 files changed
+3002 -387

No files matched your search

+221
View File
@@ -0,0 +1,221 @@
import { headLines } from './r4-summary.mjs'
import { spread } from './stats.mjs'
/**
* Markdown tables of r4-large-files.mjs (TS baseline vs Rust arrayBuffer vs
* Rust stream, per file and engine) and of r4-leak-series.mjs (settled
* levels per import in one page, WebKit repetition).
*/
const MiB = 1024 * 1024
const READ_PREFIX = 'apps/app/tests/fixtures/pdf/r4/read/'
const TOLERANCE_BYTES = 10 * MiB
const orDash = (format) => (value) =>
value === null || value === undefined || Number.isNaN(value)
? '–'
: format(value)
const mb = orDash((bytes) => (bytes / MiB).toFixed(1))
const ms = orDash((value) => value.toFixed(0))
function table(header, rows) {
const line = (cells) => `| ${cells.join(' | ')} |`
const rule = line(header.map(() => '---'))
return [line(header), rule, ...rows.map(line)].join('\n')
}
const growth = (r) => r && r.rssPeakBytes - r.rssBeforeBytes
const contentGrowth = (r) =>
r && r.rssPeakContentBytes - r.rssBeforeContentBytes
const pair = (a, b) => `${mb(a)} / ${mb(b)}`
const outcomeOf = (r) => {
if (!r) return '–'
return r.resultKind ? `${r.outcome} ${r.resultKind}` : r.outcome
}
const labelOf = (r) => (r.file ?? r.class).replace(READ_PREFIX, '')
/** "byob 791 × ≤ 64 KiB" from the worker's read facts. */
function readCell(r) {
if (!r?.readVia) return r?.precheck ? 'not read (pre-check)' : '–'
const via = r.readVia.replace('stream-', '')
const kib = (r.maxChunkBytes / 1024).toFixed(0)
return `${via} ${r.chunks} × ≤ ${kib} KiB`
}
function outcomeCell(ab, stream) {
const a = outcomeOf(ab)
const s = outcomeOf(stream)
return a === s ? s : `AB ${a}; stream ${s}`
}
function largeRow(byPath, ts) {
const ab = byPath.arrayBuffer
const stream = byPath.stream
const any = stream ?? ab
const gc = (r) =>
r?.settled?.mode === 'gc' ? r.settled.contentDeltaBytes : null
return [
labelOf(any),
mb(any.fileBytes),
any.engine,
outcomeCell(ab, stream),
ts ? pair(ts.growthBytes, ts.contentGrowthBytes) : '–',
ms(ts?.durationMs),
pair(growth(ab), contentGrowth(ab)),
ms(ab?.durationMs),
pair(growth(stream), contentGrowth(stream)),
ms(stream?.durationMs),
pair(ab?.linear?.afterBytes, stream?.linear?.afterBytes),
readCell(stream),
pair(gc(ab), gc(stream)),
]
}
/** One row per (file, engine): TS, Rust arrayBuffer, Rust stream. */
export function largeFilesTable(result) {
const groups = new Map()
for (const run of result.runs) {
const key = `${labelOf(run)} ${run.engine}`
const entry = groups.get(key) ?? {}
entry[run.readPath] = run
groups.set(key, entry)
}
const tsOf = (r) =>
result.tsBaseline.rows.find(
(t) => t.engine === r.engine && t.file === r.file,
)
const header = [
'File / class',
'MiB',
'Engine',
'Outcome',
'TS growth summed / content MiB',
'TS ms',
'Rust AB growth summed / content MiB',
'AB ms',
'Rust stream growth summed / content MiB',
'Stream ms',
'Linear MiB AB / stream',
'Stream read',
'Chromium GC delta MiB AB / stream',
]
const rows = [...groups.values()].map((byPath) => {
const any = byPath.stream ?? byPath.arrayBuffer
return largeRow(byPath, any.file ? tsOf(any) : null)
})
return table(header, rows)
}
export function largeFilesSummary(result) {
return [
...headLines(result),
`TS baseline: ${result.tsBaseline.file ?? 'none'}. Growth = peak minus ` +
'the level before the file is set (both sides); ms = Node clock, ' +
'file set to result (TS: to the first alert, error or navigation).',
'Large files (TS vs Rust arrayBuffer vs Rust stream):',
largeFilesTable(result),
].join('\n\n')
}
const list = (values) => values.map(mb).join(', ')
function leakRow(r) {
if (r.outcome !== 'series') {
return [r.engine, r.class, String(r.rep), r.outcome, r.error ?? '–'].concat(
Array(8).fill('–'),
)
}
const d = r.deltas.content
return [
r.engine,
r.class,
String(r.rep),
r.reading,
mb(r.levels.content[0]),
mb(d.init),
list(d.perImport),
mb(d.last5),
mb(d.finalIdle),
mb(d.last5AtFinal),
d.last5 <= TOLERANCE_BYTES && d.last5AtFinal <= TOLERANCE_BYTES
? 'yes'
: 'no',
mb(r.peakContentBytes),
outcomesCell(r.outcomes),
]
}
/** "7 × ok" or the distinct outcomes with their counts. */
function outcomesCell(outcomes) {
const counts = new Map()
for (const o of outcomes) counts.set(o, (counts.get(o) ?? 0) + 1)
return [...counts].map(([o, n]) => `${n} × ${o}`).join('; ')
}
/** Content RSS after each import of one page: init apart, then 6 more. */
export function leakTable(runs) {
const header = [
'Engine',
'Document',
'Rep',
'Reading',
'Start content MiB',
'Init Δ MiB',
'Δ per import after init MiB',
'Last 5 imports MiB',
'Final 5 s idle Δ MiB',
'Last 5 at final MiB',
'≤ 10 MiB',
'Series peak content MiB',
'Outcomes',
]
return table(header, runs.map(leakRow))
}
function eventsCell(r) {
if (!r.workerEvents?.length) return 'none'
return r.workerEvents
.map((e) => `#${e.import} ${e.kind} (${e.phase}): ${e.text}`)
.join('; ')
}
export function webkitRepeatTable(runs) {
const header = [
'Engine',
'Document',
'Imports',
'Outcomes',
'Worker error / messageerror events',
'Content after import MiB first / min / max / last',
'Series peak content MiB',
'Page ms median (min–max)',
]
const rows = runs.map((r) => {
const c = r.contentAfter ?? {}
const p = spread(r.pageMs ?? []) ?? {}
return [
r.engine,
r.class,
String(r.imports ?? '–'),
r.outcomes ? outcomesCell(r.outcomes) : r.outcome,
eventsCell(r),
[c.first, c.min, c.max, c.last].map(mb).join(' / '),
mb(r.peakContentBytes),
`${ms(p.median)} (${ms(p.min)}–${ms(p.max)})`,
]
})
return table(header, rows)
}
export function leakSummary(result) {
const parts = [...headLines(result), result.readings]
if (result.leak.runs.length) {
parts.push('Leak series (content RSS):', leakTable(result.leak.runs))
}
if (result.webkitRepeat.runs.length) {
parts.push(
'WebKit repetition, one page:',
webkitRepeatTable(result.webkitRepeat.runs),
)
}
return parts.join('\n\n')
}