This library contains both basic convenience functions such as readToEnd
, concat
, slice
, clone
, webToNode
and nodeToWeb
, and more complex functions for transforming and parsing streams. Examples of the latter can be found below.
npm install --save web-stream-tools
import stream from 'web-stream-tools';
See the documentation for a full list of functions.
In this example we're encrypting a stream using an imaginary API which has process
and finish
methods.
const encryptor = new Encryptor();
const encrypted = stream.transform(input, function process(chunk) {
return encryptor.process(chunk);
}, function finish() {
return encryptor.finish();
});
Both the process
and finish
functions:
- are optional (by default no data is written to the transformed stream)
- may be asynchronous
- may throw (in which case the error is forwarded to the transformed stream)
input
can be a stream containing anything, or it can be a plain value (Uint8Array or String) in which case transform()
will simply return process(input)
and finish()
concatenated together.
In this example we're encrypting a stream using an imaginary API which has a process
method that requires us to pass in chunks of size 1024 (unless it's the last chunk).
const encrypted = stream.transformPair(input, async (readable, writable) => {
const reader = stream.getReader(readable);
const writer = stream.getWriter(writable);
try {
while (true) {
await writer.ready;
const chunk = await reader.readBytes(1024);
// The above will return 1024 bytes unless the stream closed before that, in which
// case it either returns fewer bytes or undefined if no data is available.
if (chunk === undefined) {
await writer.close();
break;
}
await writer.write(encryptor.process(chunk));
}
} catch(e) {
await writer.abort(e);
}
});
The above example may seem more complicated than necessary, but it correctly handles:
- Backpressure (if
encrypted
gets read slowly,input
gets read slowly as well) - Cancellation (if
encrypted
gets canceled,input
gets cancelled as well) - Erroring (if
input
errors,encrypted
gets errored as well)
Unlike transform
, transformPair
will always return a stream, even if input
is not.
There are also helper functions for reading a specific number of bytes, or a single line, etc:
stream.parse(input, reader => {
const byte = await reader.readByte(); // Single byte or undefined
const bytes = await reader.readBytes(n); // Uint8Array of up to n bytes, or undefined
const line = await reader.readLine(); // Returns String up to and including the first \n, or undefined. This function is specifically for a stream of Strings.
// There's also peekBytes() and unshift(), which you can use to look ahead in the stream.
const stream = reader.remainder(); // New stream containing the remainder of the original stream. Only available when using a Reader from stream.parse()
});
Most of the functions above are also available when getting a reader using stream.getReader()
instead of stream.parse()
.
All of the functions above also work when reading a stream containing Strings instead of a Uint8Arrays, and will return Strings in that case.
There are also a few functions not for reading the stream, but for manipulating the stream for another function to read:
stream.slice(input, begin, end); // Returns a stream pointing to part of the original stream, or a Uint8Array
stream.clone(input); // Returns a copy of the stream so that two functions can read it. Note: this does *not* clone a Uint8Array, since this function is only meant for reading the same data twice.
stream.passiveClone(input); // Also returns a copy of the stream, but doesn't return data immediately when you read from it, only returns data when you read from the original stream. This is meant for respecting backpressure.
Note: these three functions do not work well with Node streams. Please convert Node streams to Web streams with stream.nodeToWeb
first before using them.