Add 'varify' browserify transform to support IE9,10

[ronjouch]

In f4648fc I documented a way to use convict under browserify. That was fine, until trying IE<11, which currently breaks because convict uses const.

Fortunately, browserify has a feature for that, and a simple browserify transform exists for this specific const→var simplification: varify.

• The good:
  • Fixes IE9,IE10 compatibility.
  • Only impacts browserify usage (your usage of const out of browserify persists, and unit tests impact is null).
• The bad:
  • But additional work on convict using other recent features unsupported by IE might re-break it...
  • ... and I don't see an easy way to unit-test this...
  • ... but we can add more transforms when that happens.

Still, that improves the situation now, doesn't hurt, and is using the recommended browserify feature for this specific need, so I think it's worthwhile.

[ronjouch]

@madarche I confirm dependencies has to be used in this case. Fiddling with the package.json of my npm-cached convict, here is what npm installing my webapp results into:

a. With varify in convict's devDependencies

convict@0.6.1 node_modules/convict
├── depd@1.0.0
├── validator@3.26.0
├── optimist@0.6.1 (wordwrap@0.0.2, minimist@0.0.10)
├── moment@2.8.4
└── cjson@0.3.0 (jsonlint@1.6.0)

b. With varify in convict's dependencies

convict@0.6.1 node_modules/convict
├── depd@1.0.0
├── validator@3.26.0
├── optimist@0.6.1 (wordwrap@0.0.2, minimist@0.0.10)
├── varify@0.1.1 (through@2.3.6, redeyed@0.4.4)
├── moment@2.8.4
└── cjson@0.3.0 (jsonlint@1.6.0)

→ Because a browserify user of convict will want the transform by installing a package.json requiring it (even though s/he doesn't plan to work on it), we need to add the dep to dependencies, not devDependencies.

---

Now, to the concern of adding this dependency: I admit that, from a maintainers perspective, it's sad to introduce a dependency just for an (arguably small) portion of users; ideally package.json would support something like a browserDependencies, but it doesn't. The good news, though, is that the impact on node users (apart from dependency resolution) is null because the transform won't run for them.

I hope you'll see this request as a net positive (adding a minor dep, but following the major trend of making npm packages play well both in node and in the browser), but can understand you'd be reluctant to the perspective of new dependencies. Your call.

[ronjouch]

Hi. Can anyone review this? @madarche @zaach ?

[zaach]

Thanks 👍