sindresorhus/normalize-url

What should happen with trailing slash plus query?

Closed this issue · 4 comments

jotto commented

Right now it looks like a normalization bug, unless these are in fact 2 different URLs:

> normalizeUrl('/index/?a=b')
'/index/?a=b'
> normalizeUrl('/index?a=b')
'/index?a=b'
sindresorhus commented

Not sure. Removing trailing slash with a query string is not always a safe change.

For example, https://sindresorhus.com/feedback/?foo works, but https://sindresorhus.com/feedback?foo does not.

SamVerschueren commented

They aren't the same thing. If I'm not mistaken, some REST APIs tend to use /resource/ to get a list of the resources and /resource to get one resource.

jotto commented

This might be a case-by-case thing. Based on @SamVerschueren's observation, the existing functionality might be wrong too:

> normalizeUrl('example.com/index/')
'http://example.com/index'

If the server is...

  • a static file store
    • /dir-or-file/ is the same as /dir-or-file and can't co-exist (and the trailing slash normalization seems correct)
  • dynamic
    • /dir-or-file/ could be different than /dir-or-file and you can't normalize a trailing slash (because you don't know if it's different or not)
stevenvachon commented

Removing a trailing slash regardless of a query is not always safe, as pointed out by @jotto. This library is already opinionated in regards to what is safe, though, so you may not always want the default settings.