Mostly better code documentation...

[rocky]

some more lea{f,ves} -> element{,s} converted in places we are currently
trying to describe and don't want to have to explain that by X we really
mean Y.

[TiagoCavalcante]

LGTM

[TiagoCavalcante]

I also don't get it.

[mmatera]

This is just because we choose different print representations: Strings are rounded by quotes, and filenames don't.

[rocky]

Ok. I guess that makes sense, but then why isn't Filename distinct from Symbol?

Is this how Mathematica does things too? I would have hoped they would say use another pair of delimiters like <Filename>.

I'll add a comment about this upon learning that __repr__ really is the "right" thing to do.

[mmatera]

LGTM

[mmatera]

Side comment: some time ago, I noticed that in the AST module we have classes with names that appears also as BaseExpression s. For example, we have two classes called Symbol. Do we really need these two classes? Could be just one? Otherwise, shouldn't we rename the AST Symbol to avoid confussions?

[rocky]

Side comment: some time ago, I noticed that in the AST module we have classes with names that appears also as BaseExpression s. For example, we have two classes called Symbol. Do we really need these two classes? Could be just one? Otherwise, shouldn't we rename the AST Symbol to avoid confusions?

Yeah, I wondered about this too. This may be yet more sloppiness, made harder to find due to a lack of positive description of intent. Or maybe it was the intent. Like Fermat's last theorem, keeping quiet is sometimes be assumed to be brilliance.

Python's duck typing doesn't see this as an error. That the API interfaces are compatible is then probably not a coincidence. (Because if this was not the case, then Python would throw an exception.)

At the Python level, one class is parser.Atom and the other class s expression.Atom. And I could see having these two be the same because they are equivalent being intentional or being viewed as the right thing.

Renaming is a possibility and so is subclassing.

But right now though I'd rather keep this as is and come back to this after we think about it some more. I'll add a comment to note the issue and think about and possibly take action.