--all-features and --no-default-features #152
Conversation
|
Some screenshots of how this is represented in the documentation:
|
|
Thank you, this change looks very useful! Can you check the Github Action that failed above? Is it intended, or did something go wrong? |
|
Yeah, this is not exactly complete yet, but should be sooner rather than later. I think we will also want to add some API surface to builders in a followup to allow people to construct various flows with their connector (as the currently present builder entry points are not available without |
nagisa
force-pushed
the
nagisa/features
branch
2 times, most recently
from
be7d072 to
f081b9c
Compare
just to clarify - would you be ready to add these methods? Or are you in favor of first merging this PR and then figuring out the best way forward? (I don't expect you to add the methods in this PR, we don't need big PRs -- just wanted to clarify) |
|
I think its a reasonable time to consider that given that my other PR is also a breaking change (#158) and that gives more leeway to go for a no-compromise API. |
This adjusts the code and documentation for `--all-features` and `--no-default-features` to work correctly. With `--no-default-features` no `DefaultAuthenticator` is made available. Users are in control of picking the `Connector` they want to use, and are not forced to stomach a dependency on `rustls` or `hyper-tls` if their TLS implementation of choice doesn't happen to match one of the two. To indicate this, the unstable `doc_cfg` feature is used to build documentation on docs.rs. That way the generated documentation has notices on these types that look as such: > This is supported on crate features hyper-rustls or hyper-tls only. Additionally this functionality is tested via additional coverage in the Actions' CI.
|
This should be ready now. |
| //! | ||
| //! It is also a better use of resources (memory, sockets, etc.) | ||
|
|
||
| async fn r#use<C>( |
There was a problem hiding this comment.
is r#use some new syntax I'm not aware of, or did you choose this name intentionally? It reminds me of the raw string literal syntax
There was a problem hiding this comment.
It is a raw identifier, and allows using words as identifiers that are otherwise reserved, such as use. I had a hard time coming up with another name here. Open to suggestions.
|
Great work, this looks very good! |
|
Would you prefer to make this a new release (new major version) right away, or should we wait for other opinions on how this API works? |
|
I don't think this change warrants a release I think it is fine to hold off on releasing before e.g. #158 or something similar makes it in. |
This adjusts the code and documentation for
--all-featuresand--no-default-featuresto work correctly. With--no-default-featuresno
DefaultAuthenticatoris made available. Users are in control ofpicking the
Connectorthey want to use, and are not forced to stomacha dependency on
rustlsorhyper-tlsif their TLS implementation ofchoice doesn't happen to match one of the two.
To indicate this, the unstable
doc_cfgfeature is used to builddocumentation on docs.rs. That way the generated documentation has
notices on these types that look as such:
Additionally this functionality is tested via additional coverage in the
Actions' CI.
This, I believe, should be a backwards compatible change, with a very small
exception – users who have
RUSTDOCFLAGS='--cfg=yup_oauth2_docsrs'setand are building with a non-nightly compiler will observe failures to generate
documentation.