Value::get: return a Result to account for type mismatch...

[fengalin]

... and introduce get_some for types implementing FromValue.

First commit is for non-regression validation.

I'll update the other crates (gtk-rs & gstreamer-rs) when this is merged.

Fixes #510

[fengalin]

Not ready yet! I still hope to find a good name for get_some methods :)

[fengalin]

I think we should keep get_some eventually. I can't find a better short replacement.

I am worried about the CI error on Windows 32bits: it seems like this platform doesn't like #[should_panic] attributes in tests. Any of you guys seen this before?

[sdroege]

Probably related to using the mingw toolchain. Not sure what to do about that, the Windows CI is generally borderline broken (see also GTK failing).

[sdroege]

Maybe blacklist that test on Windows for the time being.

[EPashkin]

There also error on installing mingw-w64-i686-glib2, but it was not passed build too:
https://ci.appveyor.com/project/GuillaumeGomez/glib/builds/26555598/job/f90x0sqsfokbj90i#L103

[fengalin]

Maybe blacklist that test on Windows for the time being

I'm not sure if there is a better way to do this: I added a #[cfg(..)] clause before the 2 tests that #[should_panic].

There also error on installing mingw-w64-i686-glib2

Yup! It seems kind of broken ATM.

[fengalin]

CI on Windows 32bits is now aligned with the other platforms.

I think the next step would be to merge gtk-rs/gir#825 so that I can regen the depending crates.

[sdroege]

This PR here looks good to me in any case, thanks!

If you're interested you could also update all the BoolError in our manual code (like in the set_property() function) to proper error types :)

[fengalin]

If you're interested you could also update all the BoolError in our manual code (like in the set_property() function) to proper error types :)

Ok, I'll add it to my todo list :) But I had other glib-related plans actually. I'll post a proposal when I'm done with the Value::get changes.

[sdroege]

Ok, I'll add it to my todo list :) But I had other glib-related plans actually. I'll post a proposal when I'm done with the Value::get changes.

No problem :)

[EPashkin]

As result code looks like this:

let s = model
                    .get_value(&iter, 1)
                    .get::<String>()
                    .expect("Value is not string"),
                    .expect("Couldn't get string value"),

Maybe we try include None to error?

pub enum GetError {
    Nothing,
    WrongType{requested: Type, actual: Type},
}

[fengalin]

Maybe we try include None to error?

The None case is not necessarily an error. There are cases for which you will want an optional String. See for instance this test case in subclass/object.rs.

For types that don't allow storing a None variant, we can use get_some.

[EPashkin]

Yes, with my proposal most of code will be use .ok() to get Option<T> so type error will be not handled correctly.
👍 for current version.
@fengalin Thanks

[fengalin]

gtk-rs/examples#257 all green!

I think we can merge this one and I'll proceed with the direct dependents (gio, atk, gdk-pixbuf), then the others...

[sdroege]

All good for me, but needs @GuillaumeGomez to merge all the trivial regen PRs :)