corecursive-main

what if your main was also given a way to recursively call itself?

Monthly Downloads: 12

Programming language: Haskell

License: BSD 3-clause "New" or "Revised" License

Tags: Control    

README

corecursive-main

On Hackage: https://hackage.haskell.org/package/corecursive-main .

Functional programmers are used to program with (co-)recursive functions. What if the main program was itself recursive? One could benefit the underlying operating system scheduler to orchestrate resources. For instance, limiting memory, file-descriptor on a per-recursive-call basis. Another instance is supporting distributed computations across multiple machines.

This package

This package demonstrates the idea of replacing the main entry-point of the program (i.e., main :: IO ()) with a recursion-aware main. The program first pattern matches on an "argument". This argument defines which action to take for the rest of the program. The rest of the program runs with a "Self-reference", which allows the program to call itself back. Explicitly passing a Self-reference is very similar to how the fix function allows to factor recursion out of the body of a self-referential function.

Unlike fix, which executes recursive-call in the same memory-space as the call-site, developper using this package need to teach the program to transfer arguments and return values back-and-forth the call-site and the execution site. This package minimizes the developper effort to perform this translation.

Currently, one mode is supported: spawning a child process on the same OS while serializing arguments over the command-line parameters. A library user with many entry points may want to leverage the distributed-closure capability to sidestep writing many encoder/decoder.

Future versions of this package or sibling packages will likely provide:

• support for 'JSON' or 'Dhall' serialization format
• support for 'stdin' or networked serialization mechanisms

Envisionned usages

Distributed computating:

• run a same copy of the program on many places, each on a subset of the data (much like Map/Reduce)

Dev/Ops:

• use a single binary to reconfigure the whole datacenter and services after a cataclysmic outage (see #History) as well

• orchestrate processes locally

History

This package is factored out of code and findings made while writing DepTrack. If you like this package you may find other interesting ideas in DepTrack.