Feedback using winnow as a replacement for nom in jql

[yamafaktory]

This is basically a follow-up discussion of the initial PR I made to switch from nom to winnow after getting some good feedback on Mastodon.

I'd like to thanks @epage for taking the time to jump into this initial PR, providing very good hints on how to better use the capabilities of the library - and improving performance.

I'm not going to dive into details since most of the changes were pretty straightforward based on the direct feedback. What I really liked was:

• Parser being implemented for strings and byte strings out of the box (no need to use tag most of the time)
• Cleaner methods like separated directly taking a range as first argument
• Having a clean Parser trait compared to nom which has more cumbersome types
• Easy way to parse a str to some number with e.g. digit1.parse_to().parse_next(input)
• More natural way of calling map on a Parser compared to nom where it's a another combinator
• repeat as a way to accumulate the output of a parser into a Vec

What would be very beneficial would be to have an official how-to for users migrating from nom to winnow, condensing all the examples from:

• refactor: Switch to winnow gitext-rs/git-config-env#11
• refactor: Switch to winnow crate-ci/git-conventional#37
• refactor: Switch to winnow crate-ci/typos#664
• https://github.com/Byron/gitoxide/pulls?q=is%3Apr+winnow+is%3Aclosed

[epage]

Thanks for the feedback!

#387 adds a migration guide

[yamafaktory]

Thanks for the swift addition!