Topple

Toolkit for rate-limiting nodejs processes that make so many asynchronous requests so fast that node runs out of memory. This frequently happens when you're using Node for things that don't really need to be async, and thus should have been written in a synchronous environment. Why would anyone do this? Two reasons: first, the adage about hammers and nails. Second, the desire to keep your technology stack as homogenous as possible.

What's with the name?

Visualize the pending nodejs asynchrounous calls queue like a physical stack that you pile calls onto. If the pile gets to high, it topples. There's a natural limit to how high that pile can grow (in this case, the memory limit is reached). Letting the stack topple naturally can be catastrophic: your program crashes. Topple will let you set an artificial upper bound on the height of the pile, and will topple it gracefully when that limit is hit.

How does it work?

There are two operations that Topple has to know about: firing asynchronous functions that you want rate-limited, and when those functions get resolved (i.e., their callback fires). So, you let Topple take care of those operations, like so (coffeescript example):

Topple = require 'topple'
topple = new Topple 1000, 0, 100  # ceiling, floor, polling frequency (delay between polls, in ms)
                                  # all but the first of these is optional. Default values shown.
                                  # The polling frequency specifies the how often Topple should check to resume processing after toppling

# topple.limitedCallback wraps the given function with code for managing the rate limiting mechanism. The result can be used
# exactly like the given function would be.
myAsyncFunctionCallback = topple.limitedCallback (arg1, arg2) ->
  doSpiffyStuff()
  
# topple.dispatch calls an asynchrounous function, blocking if the stack is full
# callback - will be called when the stack is empty and we are ready to resume processing.
             If the stack is under the limit, callback is called immediately
# context - the context (value of 'this') to give to the function. May be null if you don't care.
# asyncFunction - the function to call
# arg list - pass the function's arguments here. Not a single parameter; pass them just as you would pass them to the function.
topple.dispatch callback, context, asyncFunction, argList

Events

Topple extends eventEmitter and emits the following events:

1. stackFull - emitted when we topple the stack
2. stackEmpty - emitted when the backlog of asynchrounous functions has cleared (as in, the stack count <= the floor)

These events allow you to do special processing (like, for example, pausing and resuming a stream) when the stack topples and when it bottoms out.