build: upgrade winnow to 0.7 #26
Conversation
Signed-off-by: tison <wander4096@gmail.com>
| F: ModalParser<I, O, E>, | ||
| G: FnMut(O) -> Result<O2, E2>, | ||
| I: Stream, | ||
| E: FromExternalError<I, E2>, | ||
| { | ||
| #[inline] | ||
| fn parse_next(&mut self, input: &mut I) -> PResult<O2, E> { | ||
| fn parse_next(&mut self, input: &mut I) -> ModalResult<O2, E> { | ||
| let start = input.checkpoint(); | ||
| let o = self.parser.parse_next(input)?; | ||
|
|
||
| (self.map)(o).map_err(|err| { | ||
| input.reset(&start); | ||
| ErrMode::from_external_error(input, ErrorKind::Verify, err).cut() | ||
| ErrMode::from_external_error(input, err).cut() |
There was a problem hiding this comment.
@epage FYI how this crate upgrades to winnow 0.7 and retain the try_map_cut combinator.
Not sure if we can achieve the same behavior without Cut. From the changelog, it seems cut is secondary to use?
There was a problem hiding this comment.
Making ErrMode optional was about paying for what you use. toml_edit still uses ModalResult and cut_err.
To not use cut_err, you need to find ways to restructure a parser to reduce the use of opt/alt to where you don't need it (e.g. using dispatch! instead).
There was a problem hiding this comment.
Thanks for your information! Then I'd merge this patch as is.
I'd investigate if dispatch! can be used here later.
There was a problem hiding this comment.
winnow-rs/winnow#692 is an example of that. I only did that so I wouldn't force ErrMode on users.
There was a problem hiding this comment.
Thanks for the reference.
After a closer look it seems cronexpr wants to terminate immediately if a part match is success but the value is an unexpected one (mainly for better error reporting), thus try_map_cut. No backtrace is expected here.
So I'd continue the current approach unless I find a way to express the semantic better.
No description provided.