Implement basic vfunc support

[bilelmoussaoui]

Currently, the parsing part ignores the virtual functions definitions. Ideally we would need:

• Parse the XML for all the various vfuncs

• Have a [object.subclass] configuration that would allow you to

  • Modify the subclassing trait, e.g it would be by default called TypeImpl & TypeImplExt for chaining up
  • Modify the [object.subclass.vfunc] by marking at either ignore = true or rename = "some_different_vfunc_name"
  • In a far feature, configure the vfuncs once we can auto generate them

• Embed the docs for the TypeImpl::some_vfunc from the corresponding TypeClass::some_vfunc

• Generate working docs links for the vfuncs

[hohilwik]

Hi, I would like to work on this. It would take me at least a week to go through the codebase before I can start contributing though, is that okay?

[bilelmoussaoui]

Sure, make sure to check #604 for inspiration as well.

[hohilwik]

Could you give me some examples of the kinds of parsing for virtual-method parser.rs needs to support? I am not entirely sure of the inputs

[bilelmoussaoui]

For example

      <virtual-method name="get_action_name" invoker="get_action_name">
        <attribute name="org.gtk.Property.get" value="action-name"/>
        <doc xml:space="preserve">Gets the action name for @actionable.</doc>
        <return-value transfer-ownership="none" nullable="1">
          <doc xml:space="preserve">the action name</doc>
          <type name="utf8" c:type="const char*"/>
        </return-value>
        <parameters>
          <instance-parameter name="actionable" transfer-ownership="none">
            <doc xml:space="preserve">a `GtkActionable` widget</doc>
            <type name="Actionable" c:type="GtkActionable*"/>
          </instance-parameter>
        </parameters>
      </virtual-method>

Look at the files in the gir-files repository.

[hohilwik]

Thank you. I will try to get the first task done by 16th

[hohilwik]

Hello, I have slowly realized that this issue would require hundreds of lines of code so, if possible could you give me some guidance on the ways you want all this implemented? I am not very familiar with the codebase so don't know how gobject and gir internals work.

There was an old commit from vhdirk reviewed by @sdroege which did a few things with regards to vfunc parsing, but the review said that it still had a lot of assumptions about crate name, return value, etc. Some info on the shortcomings in that commit would help a lot, and if there is some doc on the expected behavior of vfuncs, that would help too

[hohilwik]

Hi, so far I have understood that just the virtual-method parsing (line 606 to 610 in https://github.com/gtk-rs/gir/pull/604/files) and read_virtual_method function(line 1017 to 1091) with some modification to include annotations for parameters and return values is what is required.

Is that correct?

[bilelmoussaoui]

The whole subclassing code (which uses virtual functions that you have to implement to change a behaviour) is all manual for now but we want to change that somehow. The goal of this issue is to start by doing the parsing of virtual-functions from the GIR file and include that information in Class/Interface types so we can use them for generating the documentation.

Instead of trying to do everything at once, I would suggest doing multiple smaller PRs that follows more or less what I mentioned in the issue and allows ticking those items.

[bilelmoussaoui]

I believe the scope of the issue is now implemented. An issue to discuss generating subclassing code could be opened if someone want to work on it.