What Erlang does almost right: types can be derived from pattern matching using dialyzer. Is there a way to modify the matcher to get proper algebraic types?
Conversation
Replying to
I'm running into this with the Lure Scheme representations, which are all built out of ad-hoc s-expression matchers. How to I map those iteration patterns to a type / grammar? I'm probably at the cusp of poorly re-inventing something... (Vague recall of some nanopass framework.)
1
What I'm after: I find these ad-hoc matchers are a great exploratory tool. But at some point it's probably best to crystallize into algebraic types. Is there a way to make that less painful as some kind of overlay on the s-expressions? Some optional/gradual typing thing?
1
I mean, the match clauses _are_ the type.
1
Not quite: they can be stacked constructors... I have a feeling I'm really not seeing something obvious though.
Replying to
well i think i know what i need to do: to read the dialyzer paper and work from that understanding. tried it before but i wasn't ready. there is also a ton of work on gradual typing, and someone mentioned refinement types (which i don't know, tweet was deleted). for now, i hack.