pangloss / pattern Goto Github PK

A problem which generally exists in implementation of lazy evaluation by substitution is that of work duplication. By if we substitute unconditionally, we might end up returning a closure which contains a computation which could have been computed once before being close over, which might be called many times.
The GHC uses a combination of strategies and heuristics, including just not inlining inside closures
There is also theoretical work on optimal implementation and sample implementations which work by graph reduction.

Does Pattern have any facilities to address this issue optimally, or would a user have to go for the same heuristics as the Haskell compiler does?

Thanks

Hey Darrick!

A couple months ago you updated deps. You also changed the url for Cognitect's test-runner, which is now incorrect. It should have stayed cognitect-labs/test-runner.

It would be nice to have Github discussions enabled for this repository to be able to ask questions, clarify behavior, etc.