bosatsu
bosatsu copied to clipboard
johnynek
A python-ish pure and total functional programming language
I'm in no rush to add typeclasses, but if I did, it might make sense to think about a design for a while. Here is a design that occurred to...
there is currently no direct syntax for type aliases. This could be a pain when you want to put type annotations. One idea would be to allow binds of capital...
Starlark allows no recursion whatsoever, and it limits it by looking at runtime if the call stack ever recurses and fails if you do. But `for i in range(10000000):` would...
I found debugging the fully typed trees to be pretty painful looking at the case classes. `TypedExpr` has a `repr` method that returns an S-expression based version of the typed...
See: https://www.haskellforall.com/2021/08/namespaced-de-bruijn-indices.html We currently do no alpha renaming but using the above trick is a way to get it: rename all variables to "a" and you have it.
https://goto.ucsd.edu/~rjhala/liquid/liquid_types.pdf this could be very powerful when combined with #735
At the end of cons we see: ```python 143 ___a0 = ___bs15 == ___bs23 144 return (1, (1, ___bs15 + ___bs23 + 1, ___bhead1, 145 ___bb13, ___bb21), 146 ___bt11) if...
I don't know if the issue is in Matchless the python generation or a combination, but I often see bits of code like: ```python 45 ___t39 = False 46 if...
Right now, you can do: ``` match foo: bar @ _: ... ``` In this case, bar is just an alias for foo. In the type system, there doesn't seem...
related to #267 #269 In this comment: https://github.com/johnynek/bosatsu/pull/269#issuecomment-497995921 While rereading the paper I am using, they state this limitation, though I had forgotten: type constructors are all monotyped in this...