Channel-based control-flow for parallel tasks with concurrency control
async.parallel with concurrency control that returns results in the correct order. Or a transform stream that executes multiple "transforms" at once.
The general use-case is concatenating files (as the example above).
var chanel = require'chanel'covar ch = chanel// only two file descriptors open at a timechconcurrency = 2var files ='1.txt''2.txt''3.txt''4.txt''5.txt''6.txt'// create and push the functions or "tasks"filesforEachchpushfsreadFilefilename 'utf8' donewhile chreadable// write each file to stdout in orderprocessstdoutwriteyield ch// exit the process (unnecessary)processexit
or concatenate them all with:
var results = yield chtruevar string = resultsjoin''
For this specific example, you're better off using combine-streams, but there are use-cases where you need to buffer the entire file such as build systems with compilation steps.
concurrency- maximum number of concurrent callbacks
discard- discard the results of the callbacks. Will only throw errors, if any, if
open- by default, the channel is closed, meaning
yield ch(true)will flush only the remaining callbacks. If opened,
yield ch(true)will not yield until the channel is closed.
Push a thunk to the channel. If the thunk returns multiple arguments, it will be sliced into an array (just like co).
Pull the next value in the channel. This waits for the next result in the channel indefinitely whether or not the channel is closed.
If an error was thrown, this function will return that error, and no more additional callbacks will be executed. To continue executing callbacks, just
yield ch again.
ch.discard === false, errors will be returned in the correct order. Otherwise, errors will be returned ASAP.
Note that this is a function. If you are not using generators, you can do something recursive like:
chif err throw err;if res console.logres;if chreadable chnext;processexit;;
co generators, you can just
cowhile chreadable console.logyield ch;
Waits until all the pending callbacks are completed. I personally call this "flushing the channel". Unless you are discarding data, all the results of the callbacks will be returned as the array
This is essentially the "give me everything method". Without generators, you can do:
Number of results waiting to be read.
A boolean to check whether you can
yield* ch.read(). Otherwise, a
yield* ch.read() call may never be yielded.
Whether the channel is closed.
true by default.
Opens the channel. When the channel is opened,
yield ch(true) will not
yield until the channel is closed.
Closes the channel.
If the channel is closed,
If the channel is opened, returns
true if data has been pushed and returns
false if the channel has been closed, which means there is no data to read.
If you're using an open channel, you want to wait until
yield ch.pushed before you start reading the data.