Leaking 'exports' into global about engine

[d4tocchini]

looks like an issue with Cassowary.js...

@bergie, maybe something with the component.io wrapper?

[adampilks]

Am finding this reoccurring with 1.0.0.rc1

We use requirejs for loading in js dependencies and GSS seems to conflict with the global require function.

[adampilks]

On 1.0.4-beta we can actually correctly load both GSS and requirejs as long as we load in GSS first and then let requirejs take over the global require variable. Not the case on 1.0.0.rc1 though.

[Inviz]

Rc provides require() shim to enable loading file-by-file. Cassowary provides logic to export global variable. I understand, that re-defining require function is a bad idea, and we try not to do it.

What would be the desired change? Avoid playing with require? or leaking cassowary global variable?

[paulyoung]

Hopefully there's a way to avoid playing with require and not leak a global variable.

Could we use browserify (or their technique) so that require isn't globally scoped?

[Inviz]

require() leaking shouldnt be an issue as it seems to me, because everybody who's concerned about it already has it defined. Is it correct?

The idea is to move gss.js out of repo to avoid hundred-ish mb repo as we have now. We'll only have compiled js files separately in the git. We'll have some kind of includes.js which will include each file one by one to use in tests. require() shim will just return globals.

gss.js will be built separately and put on the cdn with a script, it will use browserify thus will not mess with require(), because it'll be predefined there

Current code is set to define a single global variable GSS when you include gss.js. Would you like me to add some sniffing so when you do require('gss') it will not define a global? Both in browserify & node?

[paulyoung]

A nice advantage of browersify is not needing to compile intermediate files: https://github.com/substack/browserify-handbook#transforms

[paulyoung]

Current code is set to define a single global variable GSS when you include gss.js. Would you like me to add some sniffing so when you do require('gss') it will not define a global? Both in browserify & node?

@Inviz - I created this repo to demonstrate how I think this would be taken care of by browserify: https://github.com/paulyoung/browserify-gss-demo

[Inviz]

Thanks for example. Latests code already doesnt do anything with require() but I still would like to be able to include GSS file by file and not have gss.js in the repo - i'll figure something out

[paulyoung]

You should be able to do that based on my example. Just imagine that my demo project is responsible for only building gss.js and that GSS.coffee comes from the engine project.

[Inviz]

Though I think it plays against that other feature to update only a portion of a file on grunt watch, which would be even more desirable.

[adampilks]

I put together a very basic pen http://codepen.io/adampilks/pen/vExNZe that show the errors with requirejs. Requirejs introduces a globals of require and define which are conflicting with gss.

Is there a way to have these work side by side currently?

[paulyoung]

This was addressed in acfae7d in case that's an option for you until it gets merged to master.

[paulyoung]

Actually, the relevant commits are 8334f86 and 77e1a7e but it's all part of the move to browserify that exists on the queries branch.

[adampilks]

Great, we are in early development so will point our bower config to that branch for now.

[paulyoung]

We tagged v2.0.0 which includes those commits.