Support semicolon as URL query separator

[pencil]

Addressable does not support semicolon as query separator. Using semicolon instead of the ampersand is recommended by W3C (Wikipedia).

uri = Addressable::URI.parse 'http://www.google.com/search?q=nemo;oe=utf-8'
puts uri.query_values.inspect

Expected output:

{"q"=>"nemo", "oe"=>"utf-8"}

Actual output:

{"q"=>"nemo;oe=utf-8"} 

[sporkmonger]

I'm tempted to say, "Let me know when the IETF recommends this." Closing for now, but could possibly be convinced to support it if someone gave actual real-world examples of it being used in the wild on the basis of liberal parsing & Postel's Law.

[sporkmonger]

Thought about it more and decided it was more dangerous than I originally was thinking. This would introduce substantial additional ambiguity around ";" character escaping.

[luxflux]

I just know one software which does this.... OTRS. It generates URLs like this:
https://otrs.example.org/otrs/index.pl?Action=AgentTicketZoom;TicketID=1337

Thanks for implementing!

[matiasfrndz]

Section 3.3 in RFC 3986 mentions:

semicolon (";") […] are often used to delimit parameters and parameter values

Also read this answer on stackoverflow:
http://stackoverflow.com/a/2163885

And last but not least the section on path and parameters in the following article:
What Every Developer Should Know About URLs

Seems like support for semicolons is really a must in respect to standards compliance.

[sporkmonger]

It's actually a lot more complicated than everyone is making it out to be here. Despite the fact that section 3.3 does talk about the purpose of reserving the ';' character (notably in the path section, not the query section), the justification is essentially that "it's commonly done this way" and I'm hoping that it's abundantly clear to everyone that that particular phrasing is... somewhat inaccurate. Not only that, but in that particular example, the ';' character is being used in a way that is more analogous to the '?' character than the '&' character. There are also significant problems with the approach being advocated. By far the biggest is this one. If we assume the behavior you're asking for:

uri = Addressable::URI.parse("http://example.com/?param1=one;param2=two")
uri.query_values
# => {"param1" => "one", "param2" => "two"}
uri.query_values = uri.query_values.merge("param3" => "three")
# => {"param1" => "one", "param2" => "two", "param3" => "three"}
uri.to_s
# => "http://example.com/?param1=one&param2=two&param3=three"

I hope you can see why this might not be desirable behavior.

Ultimately, the query_values method is attempting to emulate the application/x-www-form-urlencoded content type, poorly specified though it may be. If I implement something other than what's already there, it'll be this.