Refactor the build!

[rwjblue]

• Enable easy dynamic package creation (no special sauce to add a new package, just do it...)
• Enable seamless .ts usage throughout all of packages/** (just rewrite a file from *.js to *.ts and make linting happy 😝 )
• Remove arbitrary filtering/building of packages and simplify what is included and excluded from various builds.

TODO:

• Remove lib/packages.js
• Make adding new packages/*/tests "just work" with each-package-tests
• Fix node test runs
• Add ember-utils package to ember-template-compiler.js asset

[rwjblue]

I have uploaded a tarball including dist here if anyone wants to review the build output...

[stefanpenner]

Can you share the before/after slow-tress output for cold build and js rebuild ?

[rwjblue]

The timings seem roughly unchanged (expand for specifics...)

Before

Cold Initial Build

Build successful (77367ms) – Serving on http://localhost:7020/

Slowest Nodes (totalTime => 5% )              | Total (avg)
----------------------------------------------+---------------------
Babel (57)                                    | 67061ms (1176 ms)
TypeScriptPlugin (1)                          | 4371ms

Warm Initial Build

Build successful (11229ms) – Serving on http://localhost:7020/

Slowest Nodes (totalTime => 5% )              | Total (avg)
----------------------------------------------+---------------------
TypeScriptPlugin (1)                          | 4245ms
Rollup (26)                                   | 3253ms (125 ms)
Babel (57)                                    | 1580ms (27 ms)
Concat (5)                                    | 950ms (190 ms)

JS Rebuild

Build successful (2708ms) – Serving on http://localhost:7020/

Slowest Nodes (totalTime => 5% )              | Total (avg)
----------------------------------------------+---------------------
broccoli-typescript-compiler (1)              | 2376ms

After

Cold Initial Build

Build successful (79002ms) – Serving on http://localhost:7020/

Slowest Nodes (totalTime => 5% )              | Total (avg)
----------------------------------------------+---------------------
Babel (7)                                     | 66325ms (9475 ms)
Rollup (21)                                   | 5227ms (248 ms)
TypeScriptPlugin (1)                          | 4704ms

Warm Initial Build

Build successful (12413ms) – Serving on http://localhost:7020/

Slowest Nodes (totalTime => 5% )              | Total (avg)
----------------------------------------------+---------------------
Rollup (21)                                   | 4389ms (209 ms)
TypeScriptPlugin (1)                          | 4055ms
Babel (7)                                     | 1697ms (242 ms)
Funnel (16)                                   | 898ms (56 ms)
Concat (5)                                    | 750ms (150 ms)

JS Rebuild

Build successful (2674ms) – Serving on http://localhost:7020/

Slowest Nodes (totalTime => 5% )              | Total (avg)
----------------------------------------------+---------------------
broccoli-typescript-compiler (1)              | 1578ms
Rollup: rollup ember-metal (1)                | 247ms
Rollup: rollup ember-glimmer (1)              | 240ms
Funnel (16)                                   | 229ms (14 ms)

[stefanpenner]

Warm Initial Build

before:  2708ms
after:  12413ms

this seems like a regression?

[rwjblue]

The cold initial and rebuild times are roughly the same, and I didn't compare the warm initial build times after prepping the numbers. Will dig into it a bit and see what we can do to improve.

Ultimately, IMHO the overall benefits of this PR (mentioned in the PR description) are significant and are valuable enough to justify the (hopefully temporary) warm initial build regression...

[rwjblue]

After starting to poke at that, I realized that the numbers you were comparing were the "Before: JS rebuild" time and "After: Warm Initial Build" time. This is because I had forgotten to paste in the "Before: Warm Initial Build" time 😭.

I've updated the comment above, but here is a screenshot of what it was before (where I accidentally trolled you 😢 ):

Ultimately, I still think there is a small regression for initial builds (~ 3% on initial cold, ~ 9% on initial warm) but a small improvement for rebuilds. Definitely still room for improvement!

[stefanpenner]

Ultimately, I still think there is a small regression for initial builds (~ 3% on initial cold, ~ 9% on initial warm) but a small improvement for rebuilds. Definitely still room for improvement!

😄 1

ya that type of regression isn't fatal, carry on. Thanks for digging in.