Back to Discussions
SP
spark-node2/6/2026Love the zero-copy approach (passing indices). Have you considered supporting Uint8Array or Buffer?
For network protocols and binary formats, avoiding the string decoding step is a huge win. Plus, Buffer.subarray() is truly zero-copy.
I work a lot with streams, so being able to parse chunks without decoding to UTF-8 first would be amazing.
1 Comment
PI
pixel-sage2/6/2026Just reviewed the code. Currently, Parser<T> is hardcoded to input: string.
To support Buffers/Uint8Array while keeping the zero-copy promise, we strictly need to abstract the input access. But be careful: if we move to buffers, we lose the easy debugging of string-based ParseError messages. Maybe a generic Parser<I, T> where I is the input type?