Further improve CJK and encoded URL support #61
Conversation
There was a problem hiding this comment.
Thanks! I was thinking about this while working on b5d9e28
Can you check if it fixed it? More test cases are definitely welcome so that regressions are caught. Especially for searches like:
|
By the way, you don't have to force-push, it makes it difficult to see what changed between reviews |
sorry, out of habit. won't do it next time. |
This reverts commit 17b2826.
fregante
changed the title
see: refined-github#61 (comment) Co-authored-by: Federico <[email protected]>
didn't notice safeDecode was only used in isCustomLink
|
This seems to be unfortunately correct, however I'm not a fan of loading that library for such a rare occurrence. However I think we have an alternative:
I'll try this |
| searchParams, | ||
| hash, | ||
| } = url; | ||
| const {search, searchParams} = url; |
There was a problem hiding this comment.
This part, together with the change in applyToLink, should resolve the problem at the root: shortenRepoUrl() now works directly on what the user sees, not the result of new URL
I'll release this as a breaking version
fregante
changed the title
punycode.jsto parse punycode encoded path elements fromURLconstructorstacked on top of test: catch invalid cases #60shorten-linksencodes/mangles CJK/UTF-8 characters refined-github#7744 (comment)