missing update operators (<op>=)

[Radvendii]

Zig Version

0.14.0-dev.93+ccd3cc326

Steps to Reproduce and Observed Behavior

||=, and=, or=, ++=, **= do not exist, but as far as I can tell this is just an oversight.

Expected Behavior

Any operator with a result type the same as it's first argument would have an <op>= variant with x <op>= y is the same as x = x <op> y.

[Radvendii]

I've realised ++=, and **= don't make sense because the return type is not the same as the first argument. The statement still holds for ||=, and= and or= though.

[mlugg]

[...] but as far as I can tell this is just an oversight.

What makes you think that? See also #5804 and #16511.

||= would be a weird operator because it's very rare to have a var x: type, let alone be combining error sets using a mutable variable like this. It could exist, but its use would be incredibly limited, and it would probably lead to further confusion between || and or.

++= and **= you've noted the issue with, and are already tracked by the issue linked above.

and= and or= would be quite weird operators because of the short-circuiting behavior of and and or. Zig has a rule that keyword operators are precisely the operators which perform control flow; where, then, do and= and or= lie, containing both keywords and symbols? It seems like they must surely have short-circuiting behavior (i.e. x and= expr does not evaluate expr if x is false), but this is also kind of unintuitive. Also, these operators just look weird -- Zig today doesn't have any operator which contains a mix of symbols and word characters.

Do you have a use case where these operators are particularly useful? Bear in mind that Zig is a language focused heavily on readability -- code being faster to write isn't typically a good justification for a new langauge feature, particularly if readability is harmed (and I would argue these operators absolutely harm readability).

[edqx]

Agree with mlugg about the short circuiting behaviour being unintuitive.

I think another weird part about and and or is that where it's useful in a language like JS to assign a default value for ||= (or=) or "update where exists" in the case of &&= (and=), this just isn't a thing in Zig since there's no truthiness, so I'd like to see what use case there really is for this...

Only thing I can think of is to somehow "build" boolean expressions, but I don't think I'd want to encourage that anyway..

[Radvendii]

Honestly, I came to this issue just from ++=. I was building a tuple at comptime and thought it would work.

Only then did I realise there were other operators missing and thought it might be a more general thing. But it seems like there are different reasons for different ones not being included. Thanks.

[nektro]

I've realised ++=, and **= don't make sense because the return type is not the same as the first argument

this isn't necessarily true; but also its rare enough in use that it's prolly fine to not add new syntax

[edqx]

I'll say that I've actually had quite a lot use personally for "++="-ing. For one, it makes it very easy to construct fields for reifying struct types.

[Radvendii]

Can y'all give an example of where it would be valid to use? Where today you use foo = foo ++ bar.

[mlugg]

Please read #16511 and direct further discussion of ++= / **= there.