Trouble defining new filtering rules with diacritics

[hieutkt]

I am trying to make a new "matching style" (not sure if I'm using the correct jargon) to make filtering in orderless behaves well with diacritics. For example typing resume would match résumé, etc...

I have written a small function that would convert résumé to resume, using the built-in ucs-normalize.el:

(defun remove-diacritics (s)
    (replace-regexp-in-string
     ucs-normalize-nfd-quick-check-regexp ""
     (ucs-normalize-NFD-string s)))

However, adding this function to either orderless-matching-styles or orderless-style-dispatchers doesn't seem to work. I read the README but haven't been able to wrap my head around this. Any help is appreciated. Thanks!

[oantolin]

You may be thinking backwards about how to do this. A matching style gets called on the pattern you type to turn it into a regexp. So here's what happens with the remove-diacritics matching style: say you have "résumé" among the candidates and you type "resume" at the prompt, then remove-diacritics gets called with the pattern you typed, namely "resume", and turns it into the regexp "resume" which, of course, does not match "résumé".

What you need instead of remove-diacritics is a function that turns "resume" into some regexp that does match "résumé". The built-in function char-fold-to-regexp fits the bill, so you can delete your remove-diacritics and add char-fold-to-regexp directly to orderless-matching-styles.

[hieutkt]

Perfect! char-fold-to-regexp solves some of the other edge-cases that I'm thinking about, too!

Thanks a lot. I'm closing this issue.