Potential to merge effort with correctionlib parser?

[nsmith-]

I recently became aware of the effort here to make a formal grammar for TFormula. This is great!
I had made my own attempt for correctionlib using a PEG parser (see here for early thoughts in that direction), with the current grammar viewable here. As of this PR the parsed AST is also accessible from python. Perhaps it might be useful to investigate if we can merge efforts? One feature of the C++ PEG parser is it will in general be much faster than lark.

[jonas-eschle]

Hi @nsmith- thanks for bringing this up! Formulate has been a bit undermaintained (that's on me!), as it would have needed a lot of effort to be rebuilt, which @aryan26roy took upon himself and upgraded the whole library.

So of course a good moment to merge these two efforts, I agree. I think formulate should also be used by uproot.

In terms of performance, while formulate was indeed slow in the past, with Lark I have not seen any impact on performance. But having a second backend could be an option. What I find slightly discouraging is to add a C++ backend due to the compilation, converting this from a pure Python package (which just simplifies logistics tbh).

I would suggest to see if correctionlib can use formulate, what features are missing and how much the impact of the performance is. Things that are probably missing could be added to formulate. For the performance, we would need to see.

This are at least my initial thoughts, what do you think should be the best way to go? Does that diverge?