Erik Rose
Erik Rose
Also, you might want to call out to an expression of a grammar other than the default one.
I think we have a failure to communicate. :-) Some clarifications: - `grammar['rule']` returns an Expression subclass and always has. - Originally, we accepted custom rules as `**kwargs`. That was...
Do you expect that you'll use custom rules much of the time? I am assuming they'd be fairly rare.
Your multiple constructors idea brought to mind one alternative we hadn't considered: chaining methods. ``` python g = Grammar(""" a = b / c b = c d """ ).more(d=lambda...
Realized that abstract supergrammars are inherently incomplete grammars and will need to be representable somehow. Rethinking based on that. `+` may return.
Okay, let's think about going back to kwarg-style custom rules. They're nice because they (a) have less syntax and (b) are more dict-like. The problem with them is that they...
If you want to whitelist, @mvcisback's solution is the thing. But if you want to blacklist, I think the ultimate solution is to make a subgrammar (once #30 is implemented)...
It's coming! It'll probably look something like this: https://github.com/erikrose/parsimonious/pull/23#issuecomment-16073245.
Suppress and Select are two pieces of sugar that would be nice. Perhaps a recursive flattener like in PyPy would help as well. I'll do a survey of Parsimonious grammars...
Transforms should also (do I even need to mention it?) be heritable, so we can suppress whitespace once and have it stick elsewhere: ``` _ = >" "+< things =...