Infer types based on argument names?

[jdonaldson]

Really glad to stumble across this library! I'm just thinking out loud here, wanted to start a discussion.

I'm realizing that in most dynamic language libraries, the argument name winds up acting like a type of some sort. E.g. an arg of "name" is always a string type in my experience. It might be possible to define some simple rules for defining named argument types on a per-extern basis.

Some examples:

pandas : arguments with name "axis" are always Integer, "skipna" is Boolean, etc.
seaborn : "data" is usually a pandas dataframe or an ndarray type.
scipy : "border_value" is usually an integer (0)

Also, should it be possible to extract all of the base level object methods into some other extendable extern definition? It would really reduce a lot of duplication. I'm talking about init, and operators like le, etc.