GDScript: Fix get_*_list() methods return incorrect info

[dalexeev]

• Closes "class_name" field of property list item is wrong for use defined classes and for at least CanvasItem.material #25432.
  • See also Add enum/flags info for properties and checks for property bindings #77618.
• Closes Object.get_method_list()  #31181.
• Closes get_method_list() returns an incorrect info about custom functions #33624.
• Closes get_script_method_list() output differs from get_method_list() on same script and neither return 'default_args[]' #55482.
• Closes GDScript 2.0: get_method_list() and class_get_method_list() return an incorrect function description #67544.
• Closes Add an indicator to method metadata to indicate static methods godot-proposals#152.
  • The METHOD_FLAG_STATIC flag was added by Add static method support to core Variant types #46378, this PR adds the flag for GDScript static methods.
• Supersedes Fix get_method_list of GDScript to properly fill the returning array #50425.
• See also [3.x] GDScript: Fix get_method_list for custom functions #68454.
• See also Editor: Improve signal callback generation #79366.

[adamscott]

Looks fine! And even simplify a lot the code. Thanks as always, @dalexeev

[bitsawer]

Do you know if this helps or affects #66218 in any way?

[dalexeev]

Do you know if this helps or affects #66218 in any way?

In the case of non-constant expressions (as well as arrays and dictionaries), the default value is unknown. We can only insert null instead.

[bitsawer]

I guess null could be a reasonable choice, at least that way it would be possible to figure out at least some default arguments. If understood that issue issue right, it's not currently possible in all cases. Then again maybe we could use some magic sentinel value instead of null, like return certain type of object etc. to mark that the value could not be used. That way it would also be possible to tell a difference between an actual null default vs non-constant expression.

Not really relevant to this PR, but thanks for answering!

Too early approval.

[adamscott]

@dalexeev If you could comment the changes you made in this PR, as discussed in the GDScript meeting, it would be very appreciated!

[dalexeev]

This PR fixes GDScript "reflection" of properties (static and non-static variables, the latter can optionally be exported), methods and signals. There are many open duplicate issues, but in each of them there are some different details that I tried to cover everything.

1. Prior to this PR, there was no single method for converting a GDScript data type into a PropertyInfo that best matches it. Unfortunately, it is not always possible to ideally represent a GDScript type as a PropertyInfo (for example, in the case of non-global classes, inner classes). This is a core issue that is still needs to be unified.

The GDScriptDataType class (which represents runtime type) had an operator PropertyInfo(), but it a. was not used everywhere (for example, signals) and b. loses some information (for example, enums). This is now the to_property_info() method in the GDScriptParser::DataType class (which represents static type) and its implementation has been improved.

2. There was an issue with Object.get_*_list() and Script.get_script_*_list() methods returning different data due to duplicate implementations. Now this data is stored in GDScript (MemberInfo::property_info) and GDScriptFunction (method_info) structures. They are generated once in the compiler (in the analyzer for signals) and the GDScript/GDScriptInstance methods simply return the data, rather than generating it in different places.

3. This PR fixes the inconsistent use of the GDScript::name property in some cases as a global name (in the Script::get_global_name() override) and in others as a local (inner class identifier).

4. Minor reorganization of GDScriptFunction: dead code removed, properties and methods arranged more conveniently.

5. Fixed incorrect function parameter names for get_method_list(), argument names and default argument values are now available in release builds too.

6. Fixed bug with missing PROPERTY_USAGE_NIL_IS_VARIANT for Variant variables.

7. Fixed bug with missing METHOD_FLAG_STATIC for static functions.

[adamscott]

Approved by the GDScript team in meeting. Great work @dalexeev, and especially great explanation of what you did. Magnificent.

[akien-mga]

Thanks!