URI::Generic#to_s encoding incompatible

[koic]

Is this intentionally changed?

Ruby 2.2.4

> require 'uri'; URI::Generic.build(host: 'localhost').to_s.encoding.to_s
=> "UTF-8"

Ruby 2.3.0

> require 'uri'; URI::Generic.build(host: 'localhost').to_s.encoding.to_s
=> "ASCII-8BIT"

This patch is compatible with Ruby 2.2.

Thanks.

[nobu]

What problem do you have by this difference?

[koic]

Thank you for your reply.

In the deprecated method URI.decode, multi-byte characters will NOT be able to decode.

Example Code

I expect that multi-byte characters ('日本語') will be decoded.

require 'uri'

p RUBY_VERSION

str = URI::Generic.build(scheme: 'http', host: 'localhost', path: '/test', query: 'language=日本語').to_s

p URI.decode(str)

Run with Ruby 2.2.4

It's just as I expected.

"2.2.4"
"http://localhost/test?language=日本語"

Run with Ruby 2.3.0

I did not expect this.

"2.3.0"
"http://localhost/test?language=\xE6\x97\xA5\xE6\x9C\xAC\xE8\xAA\x9E"

I prefer to use non-deprecated method, for example CGI.unescape, and so on.

On the one hand, I think the problem with this change is incompatibility with existing libraries.
(e.g. https://github.com/carrierwaveuploader/carrierwave/blob/0.10-stable/lib/carrierwave/storage/fog.rb#L342)

Thanks.

[nurse]

The encoding of escaped URI should be US-ASCII, and maybe URI.decode should set the resulted encoding as UTF-8. Is this solve the issue?

Anyway fog uses the resulted string as filename. What your purpose of the decoded string?

[koic]

That makes sense. It will be no problem if URI.decode should set the resulted encoding as UTF-8.

[koic]

@nurse May I close this PR?

[nurse]

After some additional thought, I merged it to backport 2.3.
trunk will be changed described above.

[koic]

Appreciate it, many thanks.