Pattern destructing could assign values to existing variables

[FMorschel]

Today we have patterns that allow us to deconstruct a variable:

class Class {
  int myField = 0;
}

void foo() {
  var instance = Class();
  if (instance case Class(myField: var myFieldValue)) {}
}

This only creates new variables after the :. I'd like to request a way for us to assign this to an existing variable. This could only work when we have this outside if/for lines, like:

late int myValue1;
late int myValue2;
late int myValue3;
if (test) {
  Class(myField := myValue1, otherField := myValue2, thirdField := myValue3) = instance;
} else {
  myValue1 = 0;
  myValue2 = 0;
  myValue3 = 0;
}

The := is only a representation here, I'm not certain this is the best solution. This would allow us to avoid creating multiple lines here.

[lrhn]

A reason not to do that is that it makes pattern evaluation have side effects even if the pattern didn't match.

Currently patterns can only invoke ==, comparison operators and getters (including []), all of which are supposed to not have any side effects. Also switch patterns are allowed to cache the value of any such operation, so it will only be performed of the pattern matches.
That allows patterns to reorder checks.

The model for patterns is that you put a value in, and you get either "no match" or "match with these new variables" as result, but no side effects are intended, especially for the patterns that didn't match.
(You can break that with side-effecting getters or when clause, but you really shouldn't.)
If anything, a pattern could remember those variable changes, and only apply them off the pattern matched, so the := introduces a new variable that shadows the original in the test of the pattern and the when clause, but which is written back to the original variable of the pattern end up matching.

Your example doesn't require patterns - it can be written as a 1-liner:

var (myValue1, myValue2, myValue3) = test? (instance.myField, instance.otherField, instance.thirdField) : (0, 0, 0);

[FMorschel]

Thanks @tatumizer, I'm aware. I just dislike the current syntax for that, hence this issue for allowing a different one 😁.

FWIW, I prefer single-assignment (final) variables whenever possible. Simultaneous assignment is perfect for this.
If #260 is fully implemented (with "last value" rule), you will be able to write the above as

var (myValue1, myValue2, myValue3) = test? instance.{(myField, otherField, thirdField)} : (0, 0, 0);

What not to like here? :-)