Fix check for a global variable "module"

[timse]

The simple check for module assumes that a global module variable exists, which is not the case if the browser just has an global exports variable and not a global module variable

also the tests on master currently fail (at least for me locally)

[mroderick]

I have pushed changes to master that should make running the tests less dependent on your npm globals, and we're now also testing in both nodejs 0.8 and 0.10 on Travis.

I would suggest rebasing against master for a better experience :)

JSLint rejected your changes

src/pubsub.js:20:40 Unexpected 'typeof'. Use '===' to compare directly with undefined.

If you fix that, then I think all the tests should pass and we can merge your fix (thank you for contributing!)

[timse]

they fail because jshint expects module to be a global variable that is always available (which is not the case) and thus asks me to compare it directly with undefined. but i cant since it doesnt exist :)

[mroderick]

they fail because jshint expects module to be a global variable that is always available (which is not the case) and thus asks me to compare it directly with undefined. but i cant since it doesnt exist :)

This project uses JSLint, not JSHint. The pubsub.js file has instructions for JSLint to ignore module

You can run JSLint with

$ grunt lint

# or this one, if you don't have grunt-cli installed as a global
$ $(npm bin)/grunt lint

[timse]

same thing, you dont just tell JSLint to "ignore" the module you tell it that the module variable is an existing global variable. Which is just not the case all the time.
Thats why JSLint wants me to do a module !== undefined because it assumes that module exists but what i want to do with my check is to verify if module exists at all, assuming it may not, in which case the module !== undefined will throw ReferenceError: module is not defined.

If you still dont like it or dont understand what i'm trying to say here, just close it. This is getting too much of an effort.

[mroderick]

Thats why JSLint wants me to do a module !== undefined because it assumes that module exists but what i want to do with my check is to verify if module exists at all, assuming it may not, in which case the module !== undefined will throw ReferenceError: module is not defined.

Not exactly, JSLint tells you to use === to compare with undefined directly, and not with "undefined".

This code fails (on JSLint.com)

var module;
var exists = typeof module !== "undefined";

This code passes (on JSLint.com)

var module;
var exists = module !== undefined;

[timse]

I'm seriously not sure if you are trolling me.

you define module var module;

and then OF COURSE you can do var exists = module !== undefined;

but in your code you cant just do var module; but you want to check if it exists, you cant initialize it localy because that would override the possible global one.

I give up, let your code break.

[mroderick]

I am not trolling you.

/*global module*/
var exists = module !== undefined;

[timse]

just try it in your browser...

[timse]

can we agree on removing exports/module from the globals list and access them via root that is given into the initialization function or whatever one may call it?

if (typeof root.exports === 'object' && root.module !== undefined){

[mroderick]

The scenario described in the initial post states

The simple check for module assumes that a global module variable exists, which is not the case if the browser just has an global exports variable and not a global module variable

If I understand this correctly, all modules using most of the UMD patterns would fail in that environment.

Would it help to switch to using commonjsStrictGlobal pattern to solve this issue?

[mroderick]

This has been fixed with #63