Adds support for q-values, some refactoring #1
Conversation
adds support for q-values per spec at https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/Accept-Encoding#Directives adds extra values and potentially helpful functions to encoding new tests changes rustfmt.toml to look similar to Tide's
|
Also, someone came to me with the idea that I should also probably expose an iterator for accept-encoding, just in case they want to be responsible for deciding which encoding to use. In practice it would be a struct EDIT: It ended up being a vector, which I think is easier to work with. |
| @@ -12,8 +12,8 @@ readme = "README.md" | |||
| edition = "2018" | |||
|
|
|||
| [dependencies] | |||
| derive_is_enum_variant = "0.1.1" | |||
| failure = "0.1.3" | |||
There was a problem hiding this comment.
@fairingrey Should we be injecting failure as a dependency for library? If it is used, other users need to pull it as well. IIRC most libraries does not use failure.
There was a problem hiding this comment.
Not sure, you'll have to ask @yoshuawuyts. It was already present when I started working on this PR 😅
I would probably remark that there might be a better way to do error handling, but I don't have any strong opinions on it. FWIW, per failure's best practices we do implement Fail for our Error type so it's easier to work with: https://docs.rs/failure/0.1.5/failure/
| if (qval - 1.0f32).abs() < 0.01 { | ||
| preferred_encoding = encoding; | ||
| break; | ||
| } else if qval > 1.0 { |
There was a problem hiding this comment.
Should we move this branch as the last branch?
I think error handling branches can come at last.
There was a problem hiding this comment.
It's a little bit difficult to track the comments here since this PR was merged already, but I'll reply here anyways.
Actually, there is a slip of logic here that would require the predicate there to actually come first instead of last (since qval can technically be 1.001 and it would still pass, even though ideally it should fail since a q-value shouldn't be greater than 1.0 flat).
This is actually addressed indirectly through the refactoring on the second PR I've made: #2
Since the logic in that function runs first anyway.
Changes
This PR does a number of things:
Checklist
Context
This seeks to extract some of the Accept-Encoding parsing logic in http-rs/tide#194 to this crate such that it becomes more modular and easier to test and reason about. Not only that, but other developers might also use this crate for their own needs; it helps if it supports the major use cases in parsing
Accept-Encodingheader by parsing q-values.Semver Changes
Increment the minor version (0.1.0 -> 0.2.0) as there are breaking changes.