Type.to_text shall invoke instance method

[JaroslavTulach]

To turn the #6216 discussion into some practical resolution I propose following improvement:

If a method x is defined on Any, then the only "static call" syntax is Any.x. Every other form is an instance invocation. E.g. My_Type.x invokes the instance method with self=My_Type (that's actually the current behavior). The same way With_X.x should be the instance invocation (that has to change for types that override x).

The only identified drawback is that we don't have a way to invoke With_X.x statically. Let's ignore that, fix the above problem (which is really painful) and see whether we need to fix the identified drawback as well (so far we aren't aware of any such need).

[somebody1234]

should the title say "instance method", not "static method", or am i just reading it wrong?

[wdanilo]

The discussion is not finished and we are chatting about this design with Radek actively. The change proposed by Radek is incompatible with how types are intended to work and this topic needs a lot more work/thinking before making such radical changes. We should revisit this topic in approx 1 week.

[JaroslavTulach]

The change proposed by Radek is incompatible ... radical ... needs more thinking ...

@wdanilo please note that this issue isn't implementing Radek's proposal. No incompatible changes, no syntax changes, no radical changes.

This issue proposes minimal change and only affects how methods defined on Any behave. There is just a few methods on Any - e.g. the impact of this change is very limited.

We should revisit this topic .... (later)

We already have at least three duplicates and counting. The current behavior of Any-methods on Type object is contra-intuitive and we need a fix. The proposed fix isn't affecting non-Any methods - e.g. the main topic of the #6216 discussion isn't affected by the change proposed in this issue. Please let us move on, let us prepare a PR for review. Continue thinking, discussing #6216 in parallel.

[JaroslavTulach]

I've just chatted with @wdanilo and Wojciech is fine with us implementing this issue, as long as it only affects dispatch of methods defined on Any.

[enso-bot]

Hubert Plociniczak reports a new STANDUP for yesterday (2023-04-26):

Progress: Investigating special logic for methods that override Any's definitions. Draft PR is ready, most of corner cases seem to be covered. It should be finished by 2023-04-27.

Next Day: Next day I will be working on the #6300 task. Investigating a, potentially simpler solution, that involves first resolved the symbol for Any and then for self's type.

[enso-bot]

Hubert Plociniczak reports a new STANDUP for today (2023-04-27):

Progress: More testing, addressing PR review. Went back to #6257 to address PR review. Investigating various LS bugs like #6447 or #6395 It should be finished by 2023-04-27.

Next Day: Next day I will be working on the #6395 task. Debugging LS bugs.