Regular expression literal can have its whitespace removed

[li-a]

Consider this legal JS (found in highcharts 4.1.4, already apparently minified by Closure Compiler):

{dSetter:function(a,b,c){a&&a.join&&(a=a.join(" "));/(NaN| {2}|^$)/.test(a)&&(a="M 0 0");c.setAttribute(b,a);this[b]=a}}

JSMin outputs:

{dSetter:function(a,b,c){a&&a.join&&(a=a.join(" "));/(NaN|{2}|^$)/.test(a)&&(a="M 0 0");c.setAttribute(b,a);this[b]=a}}

The RegExp literal has been modified (| { --> |{, which also happens to be invalid syntax).

Seems to be distinct from #11 because testing across history indicates the problem was only introduced by cec896f (which is newer than #11).

[douglascrockford]

A problem I have had historically is that if one of my tools accepts code that is badly written, then it is assumed that 'I' am approving, which I am not. That is why made that change in 2012. Up until now, no one has complained.

There is a lot of stupid stuff happening in that highcharts code, but I am past caring about it. JSMin will now accept it.

[li-a]

Thanks for looking. The "badly written" code here is output from minification, not meant for human consumption. Highcharts source happens to look friendlier:

if (/(NaN| {2}|^$)/.test(value)) {
	value = 'M 0 0';
}

(and minifies fine w/ JsMin).

Ugliness or other badness aside, as far as I can tell, the JsMin input here is valid ES 3 and the output is not. We have an easy workaround for this one, but we believe this hits many codebase files that are not highcharts.

I didn't realize JsMin was being maintained, and we have seen the usage warning/caveats, but I filed this in case someone else is searching around. Will post a fix if we make one.

(Of course we shouldn't be passing pre-minified code through JsMin anyway; that's a future fix.)

[li-a]

Some simplified failure cases:

• ;/a# b/; --> ;/a#b/;
• ;/a| b/; --> ;/a|b/;
• ;/# /; --> ;/#/;
• ;/| /; --> Error: Unterminated regular expression at line 0 and column 7

The leading semicolon seems to be significant.

[douglascrockford]

I think this is fixed in the latest.

Why are you minifying minified code?

[li-a]

Confirmed all our cases are resolved. I didn't notice changes went in, thanks for looking.

The double minification fix is planned.