Graphql object proc macro

[theduke]

This picks up form @ForsakenHarmony s work.

I squashed the commits down to one, but preserved the contribution.

Old pr: #181
( tried to push to @ForsakenHarmony s branch, but the force push apparently closed the old PR and I can't reopen it)

My commit finishes up the implementation.

Todo:

• docs
• some more tests
• deprecate graphql_object! macro and replace it everywhere

NOTE: I decided to go for a different syntax than previously discussed.
Main motivation:

• don't imply that you are actually implementing a trait
• make IDEs work/not report errors.

Check the integration_tests/src/codegen/impl_object.rs file for the tests.

It looks like this:

#[juniper::impl_object(
  context = MyContext
)]
impl Query {
  fn whatever(ctx: &MyContext, arg: bool) -> bool {
    true
  }
}

[theduke]

@ForsakenHarmony thanks for starting this, and sorry it took so long for us to get the ball rolling.

[codecov-io]

Codecov Report

Merging #333 into master will decrease coverage by 1.96%.
The diff coverage is 32.41%.

@@            Coverage Diff             @@
##           master     #333      +/-   ##
==========================================
- Coverage   88.32%   86.36%   -1.97%     
==========================================
  Files         103      105       +2     
  Lines       14922    15348     +426     
==========================================
+ Hits        13180    13255      +75     
- Misses       1742     2093     +351

Impacted Files	Coverage Δ
juniper/src/macros/tests/object.rs	92.15% <ø> (ø)	⬆️
juniper/src/macros/interface.rs	88.63% <ø> (-2.28%)	⬇️
integration_tests/juniper_tests/src/lib.rs	100% <ø> (ø)	⬆️
juniper_warp/src/lib.rs	93.17% <ø> (ø)	⬆️
juniper_iron/src/lib.rs	73.29% <ø> (ø)	⬆️
juniper/src/types/base.rs	91.01% <ø> (-0.7%)	⬇️
juniper/src/lib.rs	87.09% <ø> (ø)	⬆️
juniper/src/macros/object.rs	42.42% <ø> (-52.58%)	⬇️
juniper_codegen/src/derive_object.rs	0% <0%> (ø)	⬆️
juniper_codegen/src/impl_object.rs	0% <0%> (ø)
... and 43 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 794568e...a2b2cc3. Read the comment docs.

[mwilliammyers]

@theduke Will this work with interfaces? It is currently not very ergonomic to have to implement a trait and use graphql_object!...

I might need to open a new issue but if using this with interfaces will not work then would it be possible to use a trait + graphql_interface! + GraphQLObject instead of a trait + graphql_interface! + a struct +graphql_object!

[LegNeato]

Looks pretty good. Does this increase the minimum rust version?

• Formatting is failing
• No Changelog entry

[theduke]

@mwilliammyers interfaces are another topic, but we can definitely make these more convenient.

We can discuss in #295 .

[theduke]

@LegNeato

Alright! This was a lot of work, but it's ready now.

• impl_object now supports all features of graphql_object!
• all uses of graphql_object are replaced with impl_object, except the macro tests
• book is updated
• impl_object has decent documentation

A couple of notes:

Rust version

Sadly the Rust version must be bumped to 1.33 to allow for arbitrary tokens in a attribute.

Diagnostics

Error messages could be improved quite a bit in some cases with RFC 1566, but that doesn't seem close to stabilization.

Deprecation

#[deprecated] has no effect on macros, so I've settled on a big manual deprecation warning

Argument customization

Argument customization is pretty awkward right now.
We can substantially improve this once RFC 2565 is implemented.

For now we'll have to live with:

#[graphql(
  arguments(
    arg1(
       description = "..",
       default = true
    )
  )
)]
fn field(arg1: bool) ->  bool { arg1 }

Re-exports

I've fixed up the re-exports in juniper/src/lib.rs
We do not want to export the internal macros...

Also, I've started to manually re-export all items so rustdoc shows documentation for them.
We should actually add some documentation on them too but that's another PR.

outdated

[theduke]

@LegNeato ping

[LegNeato]

It uses a regex using "[^"]+" so I doubt that's it. Like I said I tested everything and was sure changing versions in codegen worked (even in CI) so something is wonky. As you can see on master the release job passed.

[theduke]

Alright, found it. Needed to add the replacement config in juniper/release.toml.

[LegNeato]

Arguments look pretty ugly in this new world :-/. Also not really in love with impl_object name, but can't think of anything better right now.

[theduke]

Arguments look pretty ugly in this new world

Yeah as mentioned above, arguments can become nicer once the accepted RFC for attributes on function parameters is implemented. That will enable

fn field(
  &self,
  #[graphql(default = true, description = "arg")]
  arg: bool,
) -> &str {

}

Also not really in love with impl_object name, but can't think of anything better right now.

Me neither.

I initially stuck with impl_object because of the name clash with graphql_object!, but since that is already replaced and removed from the docs, I think just object is the better choice.

[theduke]

Alright, let's do this.

PS: @LegNeato , please check gitter!