feature: unary prefix/suffix operator support via PrattParser

[tomtau]

Closes #461

based on #344

Co-authored-by: @segeljakt

[tomtau]

@pest-parser/maintainers is it ok to merge and release it? It's non-breaking, but not sure if marking PrecClimber deprecated right away is ok

[NoahTheDuke]

Sorry, been busy. Is there any reason to deprecate the old parser instead of letting it live side by side?

[tomtau]

Sorry, been busy. Is there any reason to deprecate the old parser instead of letting it live side by side?

because it's a duplicate -- PrattParser can do what PrecClimber can do, but is more general (so it can handle more cases, such as unary ops)

[NoahTheDuke]

If it's purely additive, is there any reason to rename it? Or is the api fundamentally different somehow?

[tomtau]

If it's purely additive, is there any reason to rename it? Or is the api fundamentally different somehow?

The API is different. See e.g. the calculator test: https://github.com/pest-parser/pest/pull/710/files#diff-e2dbd27540c35bdae26743d1b758d99b1186ec63a0be6abc83dfd4ca70c571edR132

[tomtau]

(plus the original PR's comments: #344 (comment) )

[NoahTheDuke]

Okay, I've had a chance to read through the code. I'm glad to see that the changes to Pest internals are minimal. Overall, this is a good PR and I think it should be merged. Two thoughts:

This might be my Clojure background coming out but I think we should make it clear that we're not going to remove PrecClimber in a future version, just marking it as less powerful/useful than PrattParser. It's important that people can trust the code they write today will still work in the future, even across multiple versions.

I also think that the internal method names nud, led, and lbp are quite terse. I have no idea what they stand for, lol. Given that they're only used internally and only in one place, I think they should get more descriptive names, ideally with at least minimal docstrings.

Thanks so much for getting this cleaned up!

[tomtau]

@NoahTheDuke thanks! I extended the deprecation note.
As for the internal methods, I kept them, but added doc strings on them to explain what they stand for. While those terse names are a bit scary, they are ubiquitous in Pratt parser implementations (given that's what the original 1973 paper used).