Use tree shaking

[doug-wade]

This could in theory reduce the size of the initial bundle. Webpack 2 is not out of beta, yet, though, so we'd have to migrate to rollup.

Migrated from #50

[doug-wade]

More evidence in favor of rollup instead of webpack https://nolanlawson.com/2016/08/15/the-cost-of-small-modules/

[drewpc]

Webpack 2 is out of beta now...maybe worth taking a look at migrating from Webpack 1 to Webpack 2 and using the tree shaking stuff there: https://medium.com/modus-create-front-end-development/webpack-2-tree-shaking-configuration-9f1de90f3233

[roblg]

FYI, I started this upgrade for Redfin's code internally, and one issue that I ran into that I haven't figured out yet is that the ExtractTextPlugin is still busted in watch mode, and the little hack we use in react-server-cli to "subclass" ETWP and make in non-cacheable didn't seem to work to fix it this time. :(

[karlhorky]

webpack 2 has released a final release candidate before shipping the final version 2: https://twitter.com/TheLarkInn/status/819215334590783489

[doug-wade]

Maybe even today 😏

Woah. Let's land #809 so that @drewpc doesn't have to entirely rewrite that super awesome pr, and then jump on this.

[drewpc]

@doug-wade thanks for the shout out! Unfortunately....that super awesome PR was kinda thrown out. I'll need to rewrite it anyway. There's a number of architecture decisions that need to be made first, including when/if to move to Webpack 2.

[roblg]

Oh! One other thing I just remembered about webpack 2 that we should be careful about: I can't find the exact numbers, but it slowed down our internal 90s build by >1min. For now, it's just an FYI, but I'd hate to force all users to suffer through perf problems because of us upgrading from webpack 1-2.

[karlhorky]

cc @jhnns regarding v2 performance problems

[jhnns]

I haven't heard of such a big increase in build time, but it can be. We still don't have big test projects to check performance regressions in build times 😢

One thing I wanted to point out is that tree shaking will currently have little effect because many libs are still published in CommonJS. So, this optimization will still take some time to be effective.

[drewpc]

This is addressed in #1003