Implement support for case statements in Ada

[NlightNFotis]

This implements support for case statements in Ada. There was already support in for case expressions, which are pretty similar but their evaluation returns a value by default. I heavily modified the support for that to get this to work.

This is still a WIP. I want to add tests (I have them already, but they need to be cleaned up), and I also want to add tests for the case expression feature which seemed to be undocumented and untested as a feature. This is just to facilitate some early reviewing for the actual functionality.

[hannes-steffenhagen-diffblue]

Does this work for when others, ranges or alternatives?

i.e.

case x of
  when others => this_always_happens;
end case;

case x of
  when 0 | 1 | 2 => valid_ternary;
  when 3 => is_a_three;
  when others => is_something_else;
end case;

case x of
  when 0..15 => can_convert_to_hex_digit;
  when others => can_not;
end case;

see here for things that can appear in a case statement

[xbauch]

This looks sensible to me.

[martin-cs]

Seems like the right kind of direction

[martin-cs]

Maybe I am just old and paranoid BUT ... if you are assuming something it is best to make it an assert rather than a comment; even if it is 'obvious'.

[hannes-steffenhagen-diffblue]

Definitely worth asserting, but the grammar does indeed guarantee this is true for now (but again, worth checking anyway, who knows what's going to happen in Ada 2023)
https://www.adaic.org/resources/add_content/standards/05rm/html/RM-3-8-1.html#S0073

[sonodtt]

Parent link of one Hannes posted (case statements)
https://www.adaic.org/resources/add_content/standards/05rm/html/RM-5-4.html

[NlightNFotis]

Just a heads up for this. It covers everything and works reliably for everything but ranges in the case statements. That fails, but it fails because Do_Expression doesn't handle ranges (this is my understanding at least, because it doesn't recognise the Node_Irep, it's falling in the when others case).

I am working on it, but it seems to be slightly expanded as scope, so I'd rather this go in as it is (it works and has tests) and then the ranges support goes in as part of a different PR.

[xbauch]

I guess this line (and 1829) could be moved after the if-then-else block.

[xbauch]

@NlightNFotis I would tend to agree: leave the ranges for another PR and just add a failing test here.

[xbauch]

Looks good.

[hannes-steffenhagen-diffblue]

why does this test not assert this function returns the right value?

[hannes-steffenhagen-diffblue]

assert result

[hannes-steffenhagen-diffblue]

assert

[hannes-steffenhagen-diffblue]

Will approve once test cases are updated

[hannes-steffenhagen-diffblue]

Cute

[hannes-steffenhagen-diffblue]

Make_Bool_Type, although even better we should probably just have a constant Bool_Type := Make_Bool_Type somewhere in this module so we don't keep recreating new ireps for all of these unnecessarily