ignore "local variable is never mutated" error if ignoring with "_ = myvar"

[silbinarywolf]

Zig Version

0.13.0

Steps to Reproduce and Observed Output

var myblah = 5; // <- "local variable is never mutated" error occurs
_ = myblah; // autofix

Expected Output

If I'm purposefully ignoring using a variable for now, I don't want it to throw an error as I then need to change it to const, then remove the ignore statement, then revert from const to var while I'm prototyping things.

var myblah = 5; // <- no error should occur
_ = myblah;

[Rexicon226]

Not a bug. The intended way to discard vars is with _ = &myblah;.

[rohlem]

(as pointed out above)
This is currently only mentioned via inline comments in the langref, in completely unrelated tests (about vectors, pointer arithmetic, slicing, allowzero).

The current compile error has a note: consider using 'const'.
We could additionally add notes use '_ = &x;' to suppress this error (and use '_ = x;' to suppress this error for the const case) to make it easier for people to find out about this mechanism.
Not sure what the opinion of maintainers on this is - we don't want to encourage unclean code, but IMO these features currently aren't very discoverable.
(We could also, or instead, add a section on "suppressing errors" to the langref.)

[mlugg]

_ = &x is a dirty secret, and not to be encouraged. There's no intended workaround in this case: just make your variable const. The only reason we ever use the _ = &x trick in tests is to ensure something is runtime-known to demonstrate runtime behaviour.

Even in prototyping, what is your use case for this?

[silbinarywolf]

The use case is usually prototyping new behaviour within a game loop, so putting a new variable at the top of the loop:

var some_variable_we_want_to_retain_across_frames: [256]u8 | i32 // could be a buffer, an int, a struct, whatever
while (!has_game_exited) {

Then deeper within this large game-loop, hundreds of lines below, I'm making this variable be used for something.

• If its a buffer, maybe its an ImGui text field I'm testing out.
• If it's an int/Timer, maybe im tracking time elapsed.
• If its a struct, maybe its a new entity, like a moving platform and I'm just testing it out before pulling it out into an ArrayList of that type/etc.

With Zig, the unused variable rule was initially not a big issue for me because ZLS could auto-generate the unused logic with _ = myvar; // autofix. This was until requiring const came in and broke that functionality.

[rohlem]

initially not a big issue for me [due to] ZLS [...] until requiring const came in and broke that functionality.

I would suggest creating an issue for an autofix _ = &x; in ZLS (surprisingly I couldn't find one already existing);
it doesn't seem like upstream Zig wants to improve usability in this regard.