Make sure operator precedence is right

[erikrose]

IIRC, the original paper had / as extremely low precedence. Make sure we got it all right.

[keleshev]

In my PEG implementation I just used Ford's grammar, and am very happy about it. I found it very easy to bootstrap the grammar too. Ford's grammar defines character ranges, which for me eliminated any need to have embedded regular expressions.

[erikrose]

Yes, I want to deprecate regexes and build them myself out of Ford-style syntax. Regexes were just a way to get to a working product faster.

[erikrose]

I'm probably going to keep the = signs, though; they're easier to type and, I think, easier to read for Python people. I did a survey of PEG libs before I designed Parsimonious's syntax, and none of them could agree on anything (Pijnu uses :=, for example), so there wasn't the interoperability gain I hoped for from following some consensus.

[keleshev]

Well, there is a chance to interoperate with my halst/peg.rb library :-). Also, I usually follow this rule: if I don't know any better—just pick something existing. That's how docopt became.

[erikrose]

I'd consider supporting <- as an alternative syntax, though I like = as the recommended one: more like a Python assignment, easier to type, more compact, etc.

[keleshev]

If you can make Parsimonious' syntax a strict superset of Fords'—that would make me really happy.

[ramunas]

It would be great if you could lower the precedence of the '/' operator. This
would allow to avoid quite a few superfluous parenthesis. It is awkward to
group choices in longer grammars, as in the following trivial example

ex = ("x" "y" "z")
   / ("u" "w")

Note the grammar without parenthesis is not accepted by parsimonious.

[erikrose]

Working on this right now. :-)

[erikrose]

Ended up hitting a mysterious non-parsing grammar while writing the first test case and then spending the rest of the day working on a graphical grammer debugger. But I guess that's good, too. :-)

[Botwise-Operator]

I just ran into the whole "a = b c / d" thing. Anything that can be done about it?

[erikrose]

I keep coming back to it but haven't had any luck yet; it basically amounts to rewriting the grammar grammar. Someday I intend to get it right, but I can't give an estimate of when. Help is welcome, of course.

[erikrose]

In the meantime, you can certainly work around it with parentheses.

[erikrose]

Here, I've pushed the branch with my work so far: https://github.com/erikrose/parsimonious/tree/lower-precedence-ors. I don't promise it makes sense, but, if anyone wants to tinker with it, you're welcome. As the 2nd-last commit message says, see test_something in test_grammar.py for an explanation of my current problem.