Jarred Sumner
|
e8b2455f11
|
feat: add Bun.JSONL.parse() for streaming newline-delimited JSON parsing (#26356)
Adds a built-in JSONL parser implemented in C++ using JavaScriptCore's
optimized JSON parser.
## API
### `Bun.JSONL.parse(input)`
Parse a complete JSONL string or `Uint8Array` and return an array of all
parsed values. Throws on invalid input.
```ts
const results = Bun.JSONL.parse('{"a":1}\n{"b":2}\n');
// [{ a: 1 }, { b: 2 }]
```
### `Bun.JSONL.parseChunk(input, start?, end?)`
Parse as many complete values as possible, returning `{ values, read,
done, error }`. Designed for streaming use cases where input arrives
incrementally.
```ts
const result = Bun.JSONL.parseChunk('{"id":1}\n{"id":2}\n{"id":3');
result.values; // [{ id: 1 }, { id: 2 }]
result.read; // 17
result.done; // false
result.error; // null
```
## Implementation Details
- C++ implementation in `BunObject.cpp` using JSC's `streamingJSONParse`
- ASCII fast path: zero-copy `StringView` for pure ASCII input
- Non-ASCII: uses `fromUTF8ReplacingInvalidSequences` with
`utf16_length_from_utf8` size check to prevent overflow
- UTF-8 BOM automatically skipped for `Uint8Array` input
- Pre-built `Structure` with fixed property offsets for fast result
object creation
- `Symbol.toStringTag = "JSONL"` on the namespace object
- `parseChunk` returns errors in `error` property instead of throwing,
preserving partial results
- Comprehensive boundary checks on start/end parameters
## Tests
234 tests covering:
- Complete and partial/streaming input scenarios
- Error handling and recovery
- UTF-8 multi-byte characters and BOM handling
- start/end boundary security (exhaustive combinations, clamping, OOB
prevention)
- 4 GB input rejection (both ASCII and non-ASCII paths)
- Edge cases (empty input, single values, whitespace, special numbers)
---------
Co-authored-by: Claude <noreply@anthropic.com>
Co-authored-by: autofix-ci[bot] <114827586+autofix-ci[bot]@users.noreply.github.com>
|
2026-01-23 00:23:25 -08:00 |
|