Formatting can incorrectly break interval literals

[bbannier]

The formatter always prefers to format interval literals with a space between value and unit, e.g., 1 min instead of 1min. It then considers the whitespace a potential line break, even though this produces invalid source code, e.g.,

$ cat bla.zeek
event zeek_init()
	{
	local aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa = 1min;
	vector(aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa, 5 min);
	}

$ zeek-format bla.zeek
event zeek_init()
	{
	local aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa = 1
	    min;
	vector(aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa, 5
	    min);
	}

$ zeek bla_formatted.zeek
error in ./bla_formatted.zeek, line 4: syntax error, at or near "min"

[ckreibich]

I think @stevesmoot also encountered this, and I believe @Mohan-Dhawan did too (a good while ago). I fully agree that we should of course never format this way.

I took a quick look at Zeek's scanner and it treats this similarly to subnet ranges, which actually surprised me — it's happy to parse "192.168.0.0 /24", for example.

[stevesmoot]

or format them with an unbreakable space,if thats supported.

[bbannier]

I took a quick look at Zeek's scanner and it treats this similarly to subnet ranges, which actually surprised me — it's happy to parse "192.168.0.0 /24", for example.

We currently have #29 open for exactly that, but I believe the issue is different there; I left a comment over there.