GROK: Haxl from Facebook

Sun, Jul 3, 2016

Two here that are close enough to the same: a github repo and an article on its origins

GROK: Open-sourcing Haxl: A library for Haskell

GROK: facebook/Haxl (github repo)

Goal

Lately I’m interested in separation of concerns at the app level, specifically distinguishing task/instruction implementation from a domain-specific compiler for optimizing batches of those operations (grouping requests, rate-limiting, etc). This is especially useful in a functional programming mindset because of the typical emphasis on defining stateless functions.

Facebook has proved itself to be a powerful engineering force (GraphQL, React, etc.). I’ve been quite impressed with the literature Facebook’s engineering provides around its projects as well. I’m pushing the functional stuff for myself lately (Haskell, Elm). So this seems like a great project to explore.

Resource

Media Outlet

code.facebook.com (facebook dev blog), published June 10, 2014

Authors:

Simon Marlow
Jon Purdy

Discovered:

I found a reference to this while looking up a Github issue for graphql. I was actually looking for information related to merging graphql schemas. Found and pocketed this along the way. Issue link.

Offload

Haxl

layer between application code and ‘data sources’

can automatically:

- group multiple requests into a single one
- concurrently request data from multiple
- cache previous requests

benefit: sep of concerns: moving optimization out of the application logic

accomplished via:

- implicit concurrency via Applicative functors
- implicit caching means soundness and modularity for free

Implicit Concurrency

numCommonFriends x y = length <$> (intersect <$> friendsOf x <*> friendsOf y)
- <*> allows for concurrent function application

Implicit Caching

soundness - ensures that changing resources behind the haxl layer don’t lead to unpredictable results
modularity - call sites don’t need to know anything about how other call sites work

Kickers

TODO: put together/link well-written FB Tech posts
TODO: pursue Simon Marlow, Jon Purdy for background, other tech writing, github projects
TODO: Read source for Haxl
- learn how soundness is ensured
TODO: handful of haskell concepts to read about/understand

Russ Matney