Ripgrep: Raise an error on regex with backreference
I understand why rg's design fundamentally prevents support for backreferences, but it'd be nice if rg at least printed an error (warning?) message when a regex with a backreference is passed.
anntzer
All 5 comments
Doesn't it print an error today?
BurntSushi
on 5 Dec 2016
rg '(foo) \1' prints no error as of 0.3.1 (Arch Linux AUR).
anntzer
on 5 Dec 2016
Ah right. Unfortunately, \1 is actually an octal escape sequence, so I don't think that can change at the regex level. We could adopt a heuristic that emits a warning message from within ripgrep, but if someone intended to type an octal escape (which seems quite rare), then the warning would be wrong.
BurntSushi
on 5 Dec 2016
My guess is that more people are going to get burnt from trying a backreference, but I don't feel strongly about it either.
anntzer
on 5 Dec 2016
In light of https://github.com/Microsoft/vscode/issues/27802, I think I should re-open this. I've been thinking a lot about this as I rewrite the regex parser, and I regret that the regex engine supports octal escapes at all. It seems like such an incredibly niche feature that it would be much better to just not support them at all in favor of emitting a better failure mode when someone tries to use backreferences.
With that said, ripgrep could certainly decide not to support octal escapes, even if the underlying regex engine does. The current parser doesn't support this kind of introspection, but my rewrite will. Once that's ready for primetime, I'll disable octal escapes and add a better error message inside ripgrep only.
BurntSushi
on 1 Jun 2017
Related issues
kenorb
路
3Comments
chopfitzroy
路
3Comments
Limeth
路
3Comments
borekb
路
3Comments
wsdjeg
路
3Comments
Most helpful comment
In light of https://github.com/Microsoft/vscode/issues/27802, I think I should re-open this. I've been thinking a lot about this as I rewrite the regex parser, and I regret that the regex engine supports octal escapes at all. It seems like such an incredibly niche feature that it would be much better to just not support them at all in favor of emitting a better failure mode when someone tries to use backreferences.
With that said, ripgrep could certainly decide not to support octal escapes, even if the underlying regex engine does. The current parser doesn't support this kind of introspection, but my rewrite will. Once that's ready for primetime, I'll disable octal escapes and add a better error message inside ripgrep only.