stm-incremental: A library for constructing incremental computations

[ cli, concurrency, control, library, mit, options, parsing, system ] [ Propose Tags ]

A library for constructing incremental computations.


[Skip to Readme]
Versions [faq] 0.1.0.0, 0.1.0.1, 0.1.0.2
Change log CHANGELOG.md
Dependencies base (>=4.12 && <4.16), stm (>=2.1) [details]
License MIT
Copyright 2019 Samuel Schlesinger
Author Samuel Schlesinger
Maintainer sgschlesinger@gmail.com
Category Control, Concurrency
Home page https://github.com/SamuelSchlesinger/stm-incremental
Bug tracker https://github.com/SamuelSchlesinger/stm-incremental/issues
Source repo head: git clone https://github.com/samuelschlesinger/stm-incremental
Uploaded by sgschlesinger at 2020-10-15T18:12:35Z
Distributions NixOS:0.1.0.2
Downloads 93 total (93 in the last 30 days)
Rating (no votes yet) [estimated by Bayesian average]
Your Rating
  • λ
  • λ
  • λ
Status Hackage Matrix CI
Docs available [build log]
Last success reported on 2020-10-15 [all 1 reports]

Modules

[Index] [Quick Jump]

Downloads

Maintainer's Corner

For package maintainers and hackage trustees


Readme for stm-incremental-0.1.0.2

[back to package description]

stm-incremental

This library is meant to expose an interface for incremental computation using software transactional memory in Haskell.

import Control.Concurrent.STM.Incremental

main = do
  (salutation, name, greeting) <- atomically do
    salutation <- incremental "Hello"
    name <- incremental "Samuel"
    greeting <- combine (\s n -> s <> ", " n) salutation name
    pure (salutation, name, greeting)
  -- Will print "Hello, Samuel"
  atomically (observe greeting) >>= print
  atomically (set salutation "Hiya")
  -- Will print "Hiya, Samuel"
  atomically (observe greeting) >>= print

There are three operations, map, combine, and choose. They sort of correspond to the operations of fmap, liftA2, and (>>=), but not exactly, and it isn't really worth belaboring that point if it isn't clear. map allows you to construct an incremental computation depending on one other, which only ever gets updated when this single dependency does. combine allows you to construct an incremental computation depending on two others, which gets updated whenever either does. choose allows you to switch your dependency structure depending on live values in the incremental computation. In other words, this allows you to have dynamic dependencies, whereas the former two functions only allowed you to have static dependencies.

The choose combinator is the most computationally expensive, requiring in the worst case time proportional to the size of the image of the choice function passed in. That being said, it can be made logarithmic in the size of the image in a not-so-difficult way, but it must be noted that it must retain space proportional to the entire image in the heap in the worst case. In general, I would argue that this does not typically come up in the use case that I am interested in. The use case for this library is when you have a data structure you want to maintain based on some input data, but it is expensive to compute in full, and you will not be changing it extremely frequently. If you can factor your dependencies carefully, perhaps changing one piece of your computation will not require substantial work to incrementally compile the rest.