support server side rendering

[simoneb]

When using in server side rendering scenarios (i.e using React) this package once required will throw an exception because the window object is not defined.

This fix prevents both host and guest scripts from doing anything if the window object is not defined, preserving the same behavior when used client side.

[coveralls]

Coverage increased (+0.3%) to 94.903% when pulling 38b876a on simoneb:master into 6f0a2c1 on davidjbradshaw:master.

[davidjbradshaw]

Couldn't this line now just be this?

})(window)

[simoneb]

no, it will throw an exception because the global variable window is not defined. I got inspiration from this. But I don't think we need that many checks here

[davidjbradshaw]

I've merged this in by hand, as you added it to the code in js rather than src.

Thanks.

[simoneb]

sorry but the merge you did is wrong. it really needs to be

typeof window !== 'undefined' && window

accessing the window object if it's undefined will throw and exception, whereas && will shortcircuit and won't throw an exception

[davidjbradshaw]

OK, so simplifying to just not pass in window

;(function(undefined) {
	'use strict';

	if(typeof window === 'undefined') return; // don't run for server side render

...

})()

Is that ok?

[simoneb]

it's a weird way to do that but as long as you don't try to access the window global variable before checking if its type is undefined it should be fine.

Out of curiosity, what was wrong with the implementation in my PR?

[davidjbradshaw]

You added it to the wrong files. You put it into the `js` folder instead of the `src` folder. So I had to copy it by hand and as I was doing that I tried to simplify it. The `(window || {})` in the IIFE was a previous PR from someone to work around an issue like this. If that no longer works, there doesn’t seem to be much reason to pass window in from there anymore.

[simoneb]

I would just port the change as it is, even though it was done to the wrong files its semantics still apply.