Update to regex-automata v0.4 #5
Conversation
|
Are you aware of the previous pull request addeessing the same issue and the discussion there? |
|
I missed the last comments on it that it was on the latest version as well, which is why I started working on this, but yes I'm broadly aware. |
|
@hawkw: Any chance, this commit will get merged and the crate published on crates.io? This would be great. |
|
This would be lovely to see progress |
There was a problem hiding this comment.
Given that v0.4 of regex-automata actually allows constructing an automaton in a start state without providing the input, unlike v0.3, I think this update is viable. Thanks for working on this, I'll be happy to merge it.
I did leave some comments on potential changes you may want to make, but they're not blockers for moving forwards with this change.
src/lib.rs
Outdated
| let config = regex_automata::util::start::Config::new().anchored(if self.anchored { | ||
| Anchored::Yes | ||
| } else { | ||
| Anchored::No | ||
| }); |
There was a problem hiding this comment.
take it or leave it: why not just replace the anchored: bool field on Pattern with an Anchored enum value, and do this when the Pattern is constructed, instead?
|
@hawkw I think both comments are fixed now. |
|
Gentle ping on if we could review this. One nice benefit for this is that |
|
Thanks so much! Really appreciate it. |
This updates to use regex-automata v0.4, bumping the crate to 0.2 since regex-automata is a public dependency of the crate. All tests in this crate continue to pass.
This removes the
ToMatchertrait in favor of directly using the upstream Automaton trait, it didn't seem like that trait was really being used. As of the current PR, there's no way to construct Patterns with the sparse DFA (this seems to have been true of the prior version of the crate too), but all the functionality is generic over regex_automata::Automaton, so just adding constructs for the lazy DFAwould be enough to make it possible to use.
This also reduces the enabled features on regex-automata to not include unicode features. It's not obvious that downstream crates (in particular tracing-subscriber) want to pay the binary size cost for having those enabled in env-filter regexes, and this leaves the choice up to them.
Resolves #3.