Parinfer Does Not Ignore Parens in String Literals [Spacemacs/Emacs]

[hrestey]

I first noticed this with a regex expression that was getting messed up in Clojure, but it appears to be the case for normal string literals as well.

Type "()", delete closing parens, and parinfer edits it to "("). It's not often that I have to use parens inside of strings, but for a regex it is obviously something that happens frequently. Regexes of course do not follow the same syntax as a lisp, so this is obviously quite undesirable.

[eraserhd]

Hrmm... it is definitely supposed to, and this doesn't happen for me. Is it possible there's another plugin doing something weird?

[eraserhd]

To be clear, I tested in a different editor because I don't have a Spacemacs setup at the moment.

[garrett-hopper]

I am able to reproduce this on the latest master branch using the parinfer-rust emacs plugin.

It does not occur when switching to v0.4.3

[garrett-hopper]

Appears to be a regression between 0789c485 and 60204aba

Last buildable commit that doesn't have issue: 0789c4852d09d51ad66b81c44ce2575f421cd031
(There's a few unbuildable commits in between)
First buildable commit that introduces issue: 60204aba3ce375b9ec1f9b90756f62746b41419c

[gagbo]

I just started noticing this as well in Emacs, it messes up a lot of files, thanks for finding the commit to revert to.

[gagbo]

FTR, the only meaningful difference I saw at first glance is the pattern matching order changing in the src/parinfer.rs:944 0789c48...60204ab#diff-12c8375c7f73fa70fdbbaa8f23ed708403b0199da0779cf58c4a83d348370449R946

What I don't get, is how can we be hit by a regression now while the change is 2 years old.

[garrett-hopper]

@gagbo, I believe the issue was just that the Emacs plugin isn't providing a string_delimiter argument which was previously set by default.
justinbarclay/parinfer-rust-mode#70

[gagbo]

The I suppose the issue is stemming from here not parinfer-rust-mode.

The commits in the span you found seem to indicate that string_delimiters is supposed to have a default value of ", but the wrappers defined in this repo seem to ignore this part and force an empty list of delimiters instead on all exposed FFI bindings