Restore eszett ß - which is not accented. #44
Conversation
This reverts commit 4b147a0.
|
I'm not sure how universally true this is, but... *if performing unicode decomposition of a single character yields two code points, and one of those belongs to a combining diacritical block, this is a clear candidate for diacritical removal. In fact this method can be quite effective. > 'Š'.normalize('NFKD').split('')
< ["S", "̌"] (2)
> 'Š'.normalize('NFKD').split('').map(c => c.charCodeAt(0).toString(16))
< ["53", "30c"] (2)On the other hand, if a character cannot be decomposed into two code points then I'd argue it's not 'accented', although again, I'm not sure how universally true this is. > 'ß'.normalize('NFKD').split('')
< ["ß"] (1)So yeah, 👍 Eszett is not a 'accented' IMO |
|
|
||
| // See https://github.com/tyxla/remove-accents/issues/12 | ||
| tape('ß is not accented', function(t) { | ||
| t.same(removeAccents.remove('Straße'), 'Straße'); |
There was a problem hiding this comment.
Tape needs t.end() to end the suite - see prior existing tests.
I'd very much like to refactor from tape to jest or another modern runner, but that is work for another PR.
There was a problem hiding this comment.
Oh geeze sorry. Apparently I didn't run the tests. I've fixed this (and run the tests this time).
I forgot to call `t.end()` 🤦 The now outdated "ß" -> "ss" was added to the "remove accents from string" test case as part of 1fe0b90 It seems like maybe the misunderstanding was that the string contained every to-be-sanitized character, but that's not true. Since ß now has it's own unit test, I've removed it from the "remove accents from string" test.
|
Thanks again, @michaelkirk 🙌
I guess folks can manually |
As requested at #12 (comment), ß should not be replaced.
Originally discussed in #12, this behavior regressed at #31
Since someone requested #31 in the first place, it's likely that this change will surprise at least one person (/cc @gollenia), and probably others.
Is there a recommended alternative for the behavior that people like @gollenia want? Maybe something like https://github.com/anyascii/anyascii ? (disclaimer: I've never used it)