Two implementations of DHM type inference

Here are two simple implementations of Damas-Hindley-Milner type inference.

First, is my Haskell version of a region-based optimized type checker as explained by Oleg Kiselyov, in his excellent review of the optimizations to generalization used in OCaml. Oleg gives an SML implementation, which I’ve Haskellized rather mechanically (using ST instead of mutable references, etc.) The result is a bit ugly, but it does include all the optimizations explained by Oleg above (both lambda-depth / region / level fast generalization and instantiation, plus path compression on linked variables, and not doing expensive occurs checks by delaying them to whenever we traverse the types anyway).

Second, here’s my much shorter and more elegant implementation using the neat unification-fd package by Wren Romano. It’s less optimized though – currently I’m not doing regions or other optimizations. I’m not entirely satisfied with how it looks: I’m guessing this isn’t how the author of unification-fd intended generalization to be implemented, but it works. Generalization does the expensive lookup of free metavariables in the type environment. Also the instantiate function is a bit clunky. The unification-fd package itself is doing inexpensive occurs checks as above, and path compression, but doesn’t provide an easy way to keep track of lambda-depth so I skipped that part. Perhaps the Variable class should include a parameterized payload per variable, which could be used for (among other things) keeping track of lambda-depth.

Two implementations of DHM type inference

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s