Repr improvement

[lL1l1]

Description of the proposed changes

More FAF-specific improvements to the inspect library:

• Print the base class of a table, if any.
• Don't print __index fields if they're the same as the table. This removes a lot of common recursive clutter (it is still possible to recurse, for example the Brain.CDR.Brain or WorldView.view.WorldView in the example below).
• Print source for lua functions in a rather short local format.

Testing done on the proposed changes

Use this command on a Percival:

SimLua LOG(repr(SelectedUnit(), {meta=true}))

Example Output (Percival).txt

Use this command:

UI_Lua LOG(repr(import('/lua/ui/game/worldview.lua').GetWorldViews()['WorldCamera'], {meta = true}))

Example Output (main WorldView).txt

Additional context

The last version of the original repr printed class-specific data. I was lazy on implementing that, but it could be done by using a helper function instead of directly the names table. Still, in general these changes lay the foundations for class-specific data printing.

Checklist

• Changes are annotated, including comments where useful
• Changes are documented in the changelog for the next game version

[Garanas]

@maudlin27 / @relent0r you both use this function too. It may be interesting for you to look at these changes and share your thoughts.

[Garanas]

These changes are fine in general. It appears however that calling reprsl to a unit generates a very large print. What if we by default skip (recursively) printing tables with a metatable, with the exception of course of the table that we're printing initially?

The reason is that at the moment the Brain reference is recursively printed too. And that can hold a lot of information, especially for AIs.

[relent0r]

For the uninitiated what does this do for us exactly compared to the repr function now?

[lL1l1]

Maybe it is better to skip only brains and blueprints (if they're not the root), and create a flag for that specifically? There could be a wide variety of tables with metatables that we'd want to avoid.

Alternatively the proper solution would be to be like deepcopy and have a "backrefs" table that tells us whether we've seen a table or not before.