Rewrite the lexer and parser in human_encoding

[ivanlele]

This PR replaces the toolkit used in the human_encoding module with a logos lexer and a custom parser.

[ivanlele]

We used CI to scan one of our projects for license compliance, and found that a dependency of rust-simplicity has an incompatible license with the CC0 license used by rust-simplicity:
https://github.com/kamadorueda/santiago/blob/main/Cargo.toml.license

[ivanlele]

@apoelstra, In src/human_parsing/README.md, I saw that we already planned to move away from this toolkit, so it shouldn't be a major issue? License compliance might currently block few of our releases

[delta1]

ACK 10b9730; ran tests locally

[apoelstra]

@ivanlele this is not a derivative work and we are not shipping any part of our dependencies without their license. We are allowed to use GPL 3 dependencies in a CC0 project.

Is there any other motivation for this 1500-line diff that adds new dependencies and which touches code that, as you say, we don't intend to use or expand?

[ivanlele]

@apoelstra I’m not arguing that we are allowed to use GPL3 dependencies in our CC0 project. The concern is more about potential inconvenience for a user who wants to keep their code private and unknowingly uses the human_encoding module, in which case the GPL3 license would apply as it would be included in their distributing binary. human_encoding is part of the default public API, which can fasilitate such cases.

Perhaps it would make sense to put the human_encoding module behind a feature flag and add a warning to prevent accidental usage, or change it inner work, as this PR does.

as you say, we don't intend to use or expand?

I didn’t say that. human_encoding is useful for tests and potentially for prototyping combinators. What I was referring to in the README is the intention to eventually switch the underlying toolkit, and not intent to to discard or archive it.

[ivanlele]

the intention to eventually switch the underlying toolkit

Which was motivation for creating this PR

[apoelstra]

Perhaps it would make sense to put the human_encoding module behind a feature flag and add a warning to prevent accidental usage, or change it inner work, as this PR does.

I wouldn't mind feature-gating it. We should not put a 'GPL3" warning on it though. People who want to be antisocial can pay their own lawyers to do the technical work.

I also wouldn't mind going in another direction and just relicensing this whole project as GPLv3.

[apoelstra]

concept ack 10b9730. I read through the code and other than the lockfile noise I'm happy with this:

• logos is pretty low-dep (nearly as low-dependency as santiago) and no worse as far as macro dependence; the logos code is at least readable since it's made out of attributes on enum attributes, rather than making us implement the parser inside a giant macro which is throwing closures all over the place
• ...relatedly, this hand-rolled parser is much nicer than the santiago one and (probably) has better error messages
• it does not increase our MSRV at all

I would also like to feature-gate this whole module since it's kinda niche and involves a whole extra dependency. We can do that in a followup if you want.

[ivanlele]

I would also like to feature-gate this whole module since it's kinda niche and involves a whole extra dependency. We can do that in a followup if you want.

Sure, I'll roll this into the next PR. human_encoding should be an appropriate name for this feature, I think