Proposal: Require whitespace on both sides of - or neither

[SpexGuy]

This may be out of scope for the language, but it's in line with other accepted proposals like #35 and #114 which aim to avoid misinterpretation.
Here's an example of the problem this prevents:

const directions = [_]isize{
    -pitch-1,
    -pitch,
    -pitch+1
    -1,
    1,
    pitch-1,
    pitch,
    pitch+1,
};

See the bug? It took me a minute.

View Answer

Here the `,` is missing from the third item in the list. This makes the list 7 items long instead of 8, and the third item `-pitch+1-1`. If you look closely, you can see that GitHub's syntax highlighting parser actually identifies the bug, and gives the `-` on line 4 a different color to indicate that it's not unary.

My proposal to solve this is to require whitespace either both before and after -, or neither before nor after. We may also want to apply this to other operators for consistency, but I think - is the one that really causes problems. Here are some examples under this rule:

_ = a - b; // fine
_ = a-b; // fine
_ = a -b; // error
_ = a- b; // error
_ = some_really_long_thing -
    some_other_really_long_thing; // ok
_ = some_really_long_thing
    - some_other_really_long_thing; // ok
_ = some_really_long_thing-
    some_other_really_long_thing; // error
_ = some_really_long_thing
    -some_other_really_long_thing; // error

[SpexGuy]

That was fast 😄
Is the accepted version to apply this to all binary operators or just -?

[andrewrk]

That was fast smile

Yep! Easiest way to get a proposal accepted is to

(1) solve a footgun
(2) make the language accept fewer permutations of source code rather than more

Is the accepted version to apply this to all binary operators or just -?

I'd say all, for consistency, but feel free to make a case for just -.

---

Tasks to close the issue:

• Update the language spec / grammar
• Stage1 parser
• Stage2 parser

[ghost]

I have a question related to this and #35, is zig fmt expected to be able to run on code that doesn't compile? Would zig fmt help to "fix" bad whitespace, or will it just die because the code doesn't compile?

[frmdstryr]

Linear / polynomial equations are often written using + to make the coefficients clear...

So what about x*x + -2*x + c0? Is that invalid?

[SpexGuy]

That's fine, the * operators have whitespace on neither side and the + operators have whitespace on both sides. Unary - is not affected by this rule.

[andrewrk]

I have a question related to this and #35, is zig fmt expected to be able to run on code that doesn't compile? Would zig fmt help to "fix" bad whitespace, or will it just die because the code doesn't compile?

No, zig fmt would fail to parse it. If it "fixed" it then it would re-introduce the footgun.

[PavelVozenilek]

If all operators are required to have mandatory space around, it would increase readability and also allow to use - (minus) as name character. this-style may feel more readable than this_style.

Where such requirement would really shine are long names: some_long_name-other_long_name vs some_long_name - other_long_name or even some-long-name - other-long-name.