on otp-21+ use uri_string to parse uri

[tsloughter]

This also fixes shit because currently it seems to always be returning undefined for scheme and other properties.

I'd also like to extend the request to include the rest of the information uri_string:parse can give -- like userinfo.

We should re-evaluate what OTP versions to retain support for.

[coveralls]

Coverage increased (+0.07%) to 75.62% when pulling d920130 on tsloughter:use-uri-string into 581ca7a on elli-lib:main.

[coveralls]

Coverage decreased (-0.4%) to 75.135% when pulling b99032d on tsloughter:use-uri-string into 581ca7a on elli-lib:main.

[yurrriq]

lgtm except for https://travis-ci.org/github/elli-lib/elli/jobs/743673725. will plan to check on this tomorrow too

[yurrriq]

Won't it always be the case that {undefined, undefined, undefined} = {Scheme, Host, Port} for {abs_path, FullPath}? 👍 to this PR still, and for adding stuff like userinfo and query to the request.

[yurrriq]

Looks like uri_string:dissect_query/1 is failing on name=knut%3D&foo with {error,missing_value,<<>>}

[yurrriq]

Seems like RFC 3986 doesn't specify, but that uri_string:dissect_query/1 doesn't support keys without values, at least on 21.3.

[yurrriq]

aaand. you already patched it 😄 erlang/otp#1840

[tsloughter]

Hah, I don't remember patching that 👍

Did you notice what is the earliest OTP with that patch? We should probably only use uri_string from that version and above.

[yurrriq]

Did you notice what is the earliest OTP with that patch?

Looks like 22.0

We should probably only use uri_string from that version and above.

👍