Files
orca/src/relay/fs-handler-utils.ts
T
Jinwoo HongandOrca c45712cead feat(ssh): stream fs.readFile to lift 10MB SSH preview cap (#1095) (#1676)
* feat(ssh): stream fs.readFile to lift 10MB SSH preview cap (#1095)

Replaces the single-shot fs.readFile path on the SSH relay with a
push-style stream protocol modeled on VS Code's readFileStream.

Wire shape:
- fs.readFileStream request returns metadata (streamId, totalSize,
  isBinary, mimeType, chunkEncoding, resultEncoding, optional empty)
- Relay pumps fs.streamChunk notifications (256 KB base64 chunks) and
  ends with fs.streamEnd or fs.streamError
- Client cancels via fs.cancelStream notification

Invariants:
- Max 16 concurrent streams per FsHandler (TooManyStreams)
- Client clamps totalSize against caps before allocating
- Sequence-number defense against out-of-order/missing chunks
- Subscribe-before-await with frame queueing until streamId is known
- Pump cleans up registry+handle in finally; disposeAll aborts before
  release so in-flight reads exit cleanly instead of EBADF
- Empty files short-circuit (no streamId, no handle open)

Compat:
- New client tries fs.readFileStream first, falls back to legacy
  fs.readFile on JSON-RPC -32601 (with once-per-session warn log)
- Bumps MAX_PREVIEWABLE_BINARY_SIZE 10 MB to 50 MB to match local

Tests: 91 streaming tests across relay, client, mux, integration.

Co-authored-by: Orca <help@stably.ai>

* test(ssh): wait for streamEnd instead of fixed flush() in stream test

Why: the binary-streaming test relied on 5 setImmediate ticks to drain
the pump, which is racy on slower CI runners (each handle.read is async
I/O). Swap to a deadline-bounded waitFor(streamEnd) so the test is
deterministic regardless of scheduler latency.

Co-authored-by: Orca <help@stably.ai>

* fix(ssh): preserve small binary detection in streamed reads

Co-authored-by: Orca <help@stably.ai>

* fix(ssh): rebind file watcher when connection id hydrates

Co-authored-by: Orca <help@stably.ai>

* fix(ssh): refresh explorer for update-only file creates

Co-authored-by: Orca <help@stably.ai>

* fix(ssh): recompute file watches when repo connection changes

Co-authored-by: Orca <help@stably.ai>

* Revert "fix(ssh): refresh explorer for update-only file creates"

This reverts commit 7c3c683cd0.

* fix(ssh): install relay watcher dependency

Co-authored-by: Orca <help@stably.ai>

---------

Co-authored-by: Orca <help@stably.ai>
2026-05-11 15:29:15 -07:00

193 lines
6.4 KiB
TypeScript

/**
* Pure helpers and child-process search utilities extracted from fs-handler.ts.
*
* Why: oxlint max-lines requires .ts files to stay under 300 lines.
* These functions depend only on their arguments (plus `rg` being on PATH),
* so they are straightforward to test independently.
*/
import { spawn, execFile } from 'child_process'
import { open } from 'fs/promises'
import {
buildRgArgs,
createAccumulator,
finalize,
ingestRgJsonLine,
SEARCH_TIMEOUT_MS as SHARED_SEARCH_TIMEOUT_MS
} from '../shared/text-search'
import type { SearchResult as SharedSearchResult } from '../shared/types'
// ─── Constants ───────────────────────────────────────────────────────
// Why: remote reads still travel through bounded JSON-RPC frames, but matching
// the old 5MB search cap would block common JSON/log files before Monaco's
// large-file optimizations can handle them.
export const MAX_TEXT_FILE_SIZE = 10 * 1024 * 1024
// Why: matches the local cap (src/main/ipc/filesystem.ts MAX_PREVIEWABLE_BINARY_SIZE).
// Reads above the legacy 16MB single-frame budget go through fs.readFileStream,
// which chunks at STREAM_CHUNK_SIZE; see docs/relay-file-stream-design.md.
export const MAX_PREVIEWABLE_BINARY_SIZE = 50 * 1024 * 1024
export const BINARY_PROBE_BYTES = 8192
export const SEARCH_TIMEOUT_MS = SHARED_SEARCH_TIMEOUT_MS
export const DEFAULT_MAX_RESULTS = 2000
export const IMAGE_MIME_TYPES: Record<string, string> = {
'.png': 'image/png',
'.jpg': 'image/jpeg',
'.jpeg': 'image/jpeg',
'.gif': 'image/gif',
'.svg': 'image/svg+xml',
'.webp': 'image/webp',
'.bmp': 'image/bmp',
'.ico': 'image/x-icon',
'.pdf': 'application/pdf'
}
// ─── Binary detection ────────────────────────────────────────────────
export function isBinaryBuffer(buffer: Buffer): boolean {
const len = Math.min(buffer.length, 8192)
for (let i = 0; i < len; i++) {
if (buffer[i] === 0) {
return true
}
}
return false
}
export async function isBinaryFilePrefix(filePath: string): Promise<boolean> {
const handle = await open(filePath, 'r')
try {
const probe = Buffer.alloc(BINARY_PROBE_BYTES)
const { bytesRead } = await handle.read(probe, 0, probe.length, 0)
return isBinaryBuffer(probe.subarray(0, bytesRead))
} finally {
await handle.close()
}
}
// ─── Search types ────────────────────────────────────────────────────
export type SearchOptions = {
caseSensitive?: boolean
wholeWord?: boolean
useRegex?: boolean
includePattern?: string
excludePattern?: string
maxResults: number
}
export type SearchResult = SharedSearchResult
// ─── rg-based search ─────────────────────────────────────────────────
/**
* Run ripgrep (`rg`) with JSON output to collect text matches.
*
* Why `spawn` and not `execFile`: `execFile` buffers stdout internally and
* kills the child when `maxBuffer` is exceeded, even when 'data' listeners
* are attached. Under rg's verbose `--json` output, a 50MB buffer fills
* well before the match cap in large folders, and `execFile`'s silent
* buffer-exceeded error resolves the result as `truncated: false` despite
* dropping matches. See docs/design/share-text-search.md.
*/
export function searchWithRg(
rootPath: string,
query: string,
opts: SearchOptions
): Promise<SearchResult> {
return new Promise((resolve) => {
const rgArgs = buildRgArgs(query, rootPath, opts)
const acc = createAccumulator()
let buffer = ''
let resolved = false
// Why: spawn can throw synchronously on invalid options (e.g. bad cwd),
// which would leak out of the `new Promise` executor and leave the
// promise forever pending. Treat a synchronous throw as a clean
// "no results" fallback, the same way an async 'error' event is handled.
let child: ReturnType<typeof spawn>
try {
child = spawn('rg', rgArgs, {
cwd: rootPath,
stdio: ['ignore', 'pipe', 'pipe']
})
} catch {
resolve(finalize(acc))
return
}
const resolveOnce = (): void => {
if (resolved) {
return
}
resolved = true
clearTimeout(killTimeout)
resolve(finalize(acc))
}
const processLine = (line: string): void => {
const verdict = ingestRgJsonLine(line, rootPath, acc, opts.maxResults)
if (verdict === 'stop') {
child.kill()
}
}
child.stdout!.setEncoding('utf-8')
child.stdout!.on('data', (chunk: string) => {
buffer += chunk
const lines = buffer.split('\n')
buffer = lines.pop() ?? ''
for (const line of lines) {
processLine(line)
}
})
child.stderr!.on('data', () => {
/* drain */
})
child.once('error', () => resolveOnce())
child.once('close', () => {
if (buffer) {
processLine(buffer)
}
resolveOnce()
})
const killTimeout = setTimeout(() => {
acc.truncated = true
child.kill()
}, SEARCH_TIMEOUT_MS)
})
}
// ─── rg availability check ──────────────────────────────────────────
// Why no cache: `rg --version` is a sub-10ms local spawn, and caching the
// result caused a footgun — a negative cache persisted across rg installs
// (forcing a relay restart), while a positive cache could mask an rg that
// was uninstalled or broken mid-session. The `settled` flag below closes
// the original race between 'error' and 'close' that the cache was added
// to paper over, so re-checking per call is both simpler and safer.
export function checkRgAvailable(): Promise<boolean> {
return new Promise((resolve) => {
let settled = false
const child = execFile('rg', ['--version'])
child.once('error', () => {
if (settled) {
return
}
settled = true
resolve(false)
})
child.once('close', (code) => {
if (settled) {
return
}
settled = true
resolve(code === 0)
})
})
}
// Moved to fs-handler-list-files.ts to keep this file under 300 lines (oxlint)
export { listFilesWithRg, LIST_FILES_TIMEOUT_MS } from './fs-handler-list-files'