[roadmap] Streaming query for large JSON files

[Cryptoteep]

Goal

Currently query() loads the entire document into memory. For very large JSON files (logs, exports), we should support a streaming query mode that parses incrementally and yields matches as they are found.

Proposed API

import { queryStream } from 'jsoncraft';
import { createReadStream } from 'node:fs';

for await (const match of queryStream(createReadStream('big.json'), '$..events[?(@.level == "error")]')) {
  console.log(match);
}

Implementation notes

• A streaming JSON parser (SAX-style) would be needed. To preserve the zero-dependency core, this could live in a separate jsoncraft/stream entry point.
• The query engine already walks the tree recursively; adapting it to emit matches from a streaming parser is the main work.

Acceptance

• queryStream() works on Node.js readable streams
• Memory usage stays flat regardless of file size
• Supports the same JSONPath subset as query()