Capnp-JS Tranforms

This repository exists solely as the source of truth for Capnp-JS Transform patterns.

Synchronous Iterator Transforms

Consider the SugarlessIterator interface:

interfaceSugarlessIterator<Value>{next(): SugarlessIteratorResult<Value>;}typeSugarlessIteratorResult<Value>={done: true|Error}|{done: false,value: Value};

An iterator transform exposes a sugarless iterator over some Input type as an iterator over some Output type:

typeIteratorTransform<Input,Output>=(source: SugarlessIterator<Input>)=>SugarlessIterator<Output>;

This arrangement allows each transform to reuse a single, relatively small buffer to transmit data through a chain of transformations with minimal memory use. Unless documented otherwise, every time a transform calls its source's next() method, that transform should anticipate that the input from the prior next() call has become corrupted.

Asynchronous Iterator Transforms

These are actually pull streams with naming kinda borrowed from the pull streams link and from this paper.

Consider the Source interface:

typeSource<Output>=(abort: null|true,put: (done: null|(true|Error),value: Output)=>void)=>void;

An asynchronous iterator transform exposes a source of some Input type as a source over some Output type:

typeAsyncIteratorTransform<Input,Output>=(source: Source<Input>)=>Source<Output>;

This arrangement allows each transform to reuse a single, relatively small buffer to transmit data through a chain of transformations with minimal memory use. Unless documented otherwise, every time a transform calls its source's next() method, that transform should anticipate that the input from the prior next() call has become corrupted.

Source consumers are called sinks, and they have the following shape:

typeSink<Input>=(source: Source<Input>)=>void;