Support dot-like method calls; alternative to #19384 #19919; fixes #19392

[ringabout]

closes #19384
closes #19919
fixes #19392

tests are coming ...

[ringabout]

I'm checking #19806

[ringabout]

Hi @markspanbroek. I tested the tests of question, all of them pass. I would like to add questionable to important packages in case of regression. However it is blocked by mratsim/weave#182

[ringabout]

for i in x .++ (if true: 12):
  collector.add i

if expr gives the same error like before:

Error: expression '12' is of type 'int literal(12)' and has to be used (or discarded)

[Araq]

Please explain your solution here.

[ringabout]

Sorry for the delay. My PR has the same problem as #8759 ([parser] a => counter+=1 parsed as (a => counter) += 1, violating the spec, and causing issues with => )

[Menduist]

What's the status on this PR :) ? Looking forward for questionable to work again on devel

[ringabout]

@Menduist, Sorry for the inconveneience. I cannot think of a good solution.

#19919 can be merged now which reverts enabling the experiemental feature, we can embrace dot-like method calls in the future.

[Menduist]

Maybe it's the best solution indeed, we can keep it as an optional flag until we find a solution / give up & remove it

[ringabout]

I will revisit it later.