Files
faunosteo/scripts/browser/r4-large-files.mjs
T
bchanot 49591c53cd 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.
2026-10-18 18:50:47 +02:00

193 lines
6.9 KiB
JavaScript

// Large files through the Rust import worker, both read paths (R4.3a,
// BDR-058): the 10 MiB and larger inputs of the import baseline's `read`
// group, the 50 MiB file, the CAP envelope, the 16 MiB dense numbers and the
// 16 MiB rescue copy, each run once per engine and per read path:
// stream `file.stream()` into a `FileBytes` sink (BYOB reader when the
// engine offers it), `import_pdf_bytes` / `restore_copy_bytes`;
// arrayBuffer the R4.2e path: one ArrayBuffer of the file, `import_pdf` /
// the strict decode then `draft.restore`.
// Same metrics as r4-import.mjs (page ms, summed and content RSS before,
// peak and after, linear memory, chunk count and size), a fresh context per
// run, the worker terminated; the settled level before and after each run
// is a forced GC + 1 s in Chromium and 5 s idle in Firefox and WebKit (idle
// is not a GC). The two paths of one file run back to back, their order
// alternated from one file to the next. The markdown table puts the TS
// baseline's growth (peak minus before, same files) beside both paths.
//
// PLAYWRIGHT_NODE_MODULES=... node scripts/browser/r4-large-files.mjs \
// [--engines chromium,firefox,webkit] [--paths stream,arrayBuffer] \
// [--ts <ts-baseline.json>] [--scratch <dir>] [--out <file>]
import { readFileSync, readdirSync } from 'node:fs'
import { tmpdir } from 'node:os'
import { join } from 'node:path'
import { parseArgs } from 'node:util'
import { memAvailableBytes } from './lib/host.mjs'
import { importInputs, inputBytes } from './lib/import-inputs.mjs'
import { RESULTS_DIR, relativeToRepo } from './lib/paths.mjs'
import { ENGINES, loadPlaywright } from './lib/playwright.mjs'
import { largeFilesSummary } from './lib/r4-memory-summary.mjs'
import {
engineRuns,
listOf,
oneRun,
prepareModuleRun,
runConditions,
today,
writeResult,
} from './lib/r4-runs.mjs'
import { startServer } from './serve.mjs'
const MiB = 1024 * 1024
const LARGE_BYTES = 10 * MiB
const READ_PATHS = ['stream', 'arrayBuffer']
const CLASS_IDS = [
'file-50mib',
'CAP-envelope',
'dense-numbers-16mib',
'dense-numbers-copy-16mib',
]
/** The newest `<date>-ts-baseline.json` of the results directory. */
function newestTsBaseline() {
const files = readdirSync(RESULTS_DIR).filter((f) =>
f.endsWith('-ts-baseline.json'),
)
const newest = files.sort().pop()
return newest ? join(RESULTS_DIR, newest) : null
}
function options() {
const { values } = parseArgs({
options: {
engines: { type: 'string', default: ENGINES.join(',') },
paths: { type: 'string', default: READ_PATHS.join(',') },
ts: { type: 'string' },
scratch: { type: 'string', default: tmpdir() },
out: {
type: 'string',
default: join(RESULTS_DIR, `${today()}-r4-large-files.json`),
},
},
})
const engines = listOf(values.engines)
const paths = listOf(values.paths)
const bad = [...engines.filter((e) => !ENGINES.includes(e))]
bad.push(...paths.filter((p) => !READ_PATHS.includes(p)))
if (bad.length) throw new Error(`unknown engines or paths: ${bad}`)
return { ...values, engines, paths, ts: values.ts ?? newestTsBaseline() }
}
/** The `read` group's inputs of 10 MiB and more, as the baseline names them. */
function largeInputs() {
const { inputs } = importInputs(['read'])
return inputs
.map((input) => ({ input, bytes: inputBytes(input.path) }))
.filter(({ bytes }) => bytes >= LARGE_BYTES)
.map(({ input, bytes }) => ({
section: 'large',
group: input.group,
kind: input.kind ?? null,
file: relativeToRepo(input.path),
fileBytes: bytes,
input,
}))
}
const classItems = (classes) =>
classes.map((c) => ({
section: 'large',
classId: c.id,
fileBytes: c.bytes,
path: c.path,
}))
/** Each file once per path, back to back, the order alternated per file. */
function planOf(files, paths) {
return files.flatMap((file, i) => {
const order = i % 2 ? [...paths].reverse() : paths
return order.map((readPath) => ({ ...file, readPath }))
})
}
const settleOf = (engine) => (engine === 'chromium' ? 'gc' : 'idle')
/** The TS baseline's run of each large file: growth and Node-clock time. */
function tsRows(tsPath, files) {
if (!tsPath) return { file: null, rows: [] }
const ts = JSON.parse(readFileSync(tsPath, 'utf8'))
const wanted = new Set(files.map((f) => f.file).filter(Boolean))
const rows = ts.import.runs
.filter((r) => wanted.has(r.file))
.map((r) => ({
engine: r.engine,
file: r.file,
outcome: r.outcome,
durationMs: r.durationMs,
growthBytes: r.rssPeakBytes - r.rssBeforeBytes,
contentGrowthBytes: r.rssPeakContentBytes - r.rssBeforeContentBytes,
}))
return { file: relativeToRepo(tsPath), rows }
}
function conditions(server, playwrightCore, module) {
const harness = 'scripts/browser/r4-large-files.mjs'
return {
...runConditions({ harness, server, playwrightCore, module }),
page: {
worker: 'module worker /r4/worker.js, one per import, terminated',
readPaths: {
stream:
'file.stream() (BYOB reader into one 1 MiB buffer when the ' +
'engine offers it, else the default reader) pushed into ' +
'FileBytes(file.size), then import_pdf_bytes / restore_copy_bytes',
arrayBuffer:
'file.arrayBuffer(), import_pdf (the glue copies it) or the ' +
'strict TextDecoder then draft.restore (R4.2e)',
},
settle: 'Chromium: forced GC + 1 s; Firefox, WebKit: 5 s idle',
},
perimeter:
'the TS baseline ran the application (import-baseline.mjs); the Rust ' +
'runs use the measurement page; same files, RSS metrics, watchdog',
}
}
async function main() {
const opts = options()
const { pw, version } = loadPlaywright()
const prepared = await prepareModuleRun({
scratchParent: opts.scratch,
classes: CLASS_IDS,
})
const server = await startServer({ mounts: prepared.mounts })
try {
const files = [...largeInputs(), ...classItems(prepared.classes)]
const plan = planOf(files, opts.paths)
const result = conditions(server, version, prepared.module)
const context = { ...prepared, baseUrl: server.url }
const runs = []
for (const engine of opts.engines) {
const items = plan.map((item) => ({ ...item, settle: settleOf(engine) }))
const runItem = (session, item) => oneRun(session, item, context)
runs.push(...(await engineRuns(pw, engine, items, runItem)))
}
result.host.memAvailableBytesAtEnd = memAvailableBytes()
result.endedAt = new Date().toISOString()
result.engines = Object.fromEntries(
runs.map((r) => [r.engine, r.engineVersion]),
)
result.files = files.map((f) => ({
file: f.file ?? f.classId,
bytes: f.fileBytes,
}))
result.tsBaseline = tsRows(opts.ts, files)
result.runs = runs
writeResult(opts.out, result, largeFilesSummary)
} finally {
server.close()
prepared.cleanup()
}
}
await main()