Esm build

[cromefire]

For size (#1552) and speed in modern browsers

• If you've added code that should be tested, please add tests.
• Ensure your code lints and the test suite passes (yarn lint) & (yarn test).

[cromefire]

utils are not esm, because there is no single entry point.

The es5 and esm build are side by side in dist and esm.

[cromefire]

@kamilogorek Are you okay with me rewriting the imports in the esm modules, so utils can also be esm?

[cromefire]

There are multiple travis errors that don't seem like I caused them and also travis has problems because this is a PR and so certain env variables are not there.

[cromefire]

I just have to get the rewrite script running under node 6

[cromefire]

This is ready from my site, but there are still some weird travis errors

[HazAT]

Thanks for this amazing PR 💯
At first glance this really looks solid, the CI errors are because it's a "third party" PR.

We will test it locally and merge it soon.

utils do not have a single entry point because we wanted to harvest "deep imports" to use different files in browser and node.

Edit: Something that's not necessary but, do you already have some number of how much we size we save using esm over the es5 build?

[cromefire]

We will test it locally and merge it soon.

I still have to turn off side-effects I remembered, so please wait until I did that.

Does any of your files inject something into a global local object, just by importing it?

At first glance this really looks solid, the CI errors are because it's a "third party" PR.

You can turn off certain builds for 3rd party PRs

utils do not have a single entry point because we wanted to harvest "deep imports" to use different files in browser and node.

Yes I fixed it, I always import things like that, it just took some more effort.

Edit: Something that's not necessary but, do you already have some number of how much we size we save using esm over the es5 build?

I can calculate that

[cromefire]

If you want I can also do a PR after that to optimize the existing code like introduce es6 classes

[HazAT]

Would love that, like I said before the code changes you did so far look really solid :)

[cromefire]

Normal:
1300KB

ESM, but splattered (not bundled):
1200KB
min: 220KB
min, gzip (cli defaults): 212KB

Warning: This values are not very good.
Better values when it's released or if I test it via yarn link (maybe tomorrow) and it's not yet optimized for esnext

[cromefire]

My application shows:
default: 22.33KB
esm: 15.48KB
(30.7% less; gzip, min w/ terser)

[HazAT]

Thank you for your work 🙂👍

[HazAT]

Hey, sorry for the delay. We did not forget about this but we still need some time to merge this.
We want to release a patch release first, the next minor bump will include this.

[cromefire]

I was thinking about making this more opt in, because right now webpack will automatically choose the module files, which point to esm, what could be a problem for people who already use this with webpack, but don't have babel in place to transpile the files.

[kamilogorek]

Sorry for such a long response time, I was over 4 weeks on a vacations.

The code looks nice and should work just fine. However build process is already complex for us and we had some accidents along the way with it.
We need to write some assertions that the build is actually correct and working as well as some assertions around rewrite script before we merge this.

Also packages/raven-js/dist/raven.js has to be restored. It's legacy code and it's used by some people directly through services like rawgit etc. (yup, we had reports like this, so let's just keep it there).

[kamilogorek]

This sentence (and the one below) sounds kinda confusing. I'd reword it somehow.

[cromefire]

Sure I didn't come up with a better one, but if you have a better idea

[cromefire]

Also I thought some time ago, if the esm should be the default module build ("module": "esm/index.js"), because that could break some things for some people using webpack, but not having their code transpiled using babel

[cromefire]

All the build errors are coming from either this being a 3rd party PR or from the node module (which I didn't modify)

[cromefire]

I think you should skip certain builds in 3rd party PRs so the build does not error. To do that:

• change matrix to jobs
• Add a conditional (type != pull_request) to the jobs that will error in 3rd party PRs

[HazAT]

@cromefire Thanks for reacting so quickly, could you please just restore the deleted raven-js file and merge master in. All tests should work then.

[HazAT]

@cromefire

[cromefire]

Oh my bad obviously there was a problem, I'll look into what went wrong

[cromefire]

Yes I'm sorry, as I already said there had to be some problem, while rolling back

[cromefire]

Fixed, it was ignored by the .gitignore (that's also, why it got removed)

[cromefire]

Wait, I forgot to put things in .npmignore

[HazAT]

Thanks again, awesome work!