builderio / framework-benchmarks Goto Github PK

Hello,

I think that this is a great project, Angular Universal has lacked public performance benchmark data since it's so different to other frameworks and Universal has been difficult to set up until quite recently. I think there are a few issues affecting the performance of the Angular Universal example which is currently the slowest SSR example.

Things like:

• Dynamically running the critical CSS inliner at runtime (an unsafe default): angular/universal#2106
• Extra change detection cycles by not explicitly using onPush change detection: https://angular.io/api/core/ChangeDetectionStrategy
• Missing a track by function to keep track of individual rows and not re-rendering the whole table: https://angular.io/api/core/TrackByFunction
• Calling methods from inside templates which are invoked on each change detection cycle: https://blog.devgenius.io/avoid-calling-expressions-in-template-of-angular-39439c547653

If we make some PRs for that is there a contribution process or should we just make a bunch of PRs?

Thanks!

Here's what I did:

1. I made sure I'm using Node 16 with nvm (it would be cool if you could fix the exact version of Node this way)
2. npm install
3. npm run install
4. cd apps/components
5. npm install
6. npm run build
7. cd ../.. - Back to project root
8. npm run build

It's compiling a lot and then failing with this:

Error: src/app/generated-components/components/todo-item.ts:5:9 - error TS2304: Cannot find name 'Todo'.

5   item: Todo;
          ~~~~

It's possible I'm missing something obvious.

It looks like that as a side effect of those npm installs, it modified package locks:

> git st                                                  ✔  4s  16.15.0 Node
On branch main
Your branch is up to date with 'origin/main'.

Changes not staged for commit:
	modified:   frameworks/astro/package-lock.json
	modified:   frameworks/next/package-lock.json
	modified:   frameworks/next/yarn.lock
	modified:   frameworks/nuxt3/package-lock.json
	modified:   frameworks/solid/package-lock.json
	modified:   frameworks/svelte/package-lock.json

Untracked files:
	frameworks/fresh/package-lock.json

That might have something to do with the failure case.

So they can run in a consistent remote environment that anyone can reproduce results on

Hi, in the React SSR bench there's:

React SSR times:
Test Node vs Bun vs Deno at React SSR of a non trivial (the dashboard) app

Note: currently having an issue with Deno, looking into

What's the issue? How can we help to fix it?

Two frameworks or libs I see much less mentions on social media are:

• https://capri.build/ - afaik came out around the same time as Fresh, and the funny part is, the logos are even similar. This one is definitely comparable to the other frameworks in the test (e.g. you can also choose Solid to render interactive parts)
• https://turbo.hotwired.dev/ - came out a few years ago, which was the first new framework to my knowledge to make a turn to the 'send HTML and not only JavaScript bundles'-era we're entering now (if we exclude marko)

Hey it would be great to have CSV files for benchmarks, so that we could put it into a spreadsheet and do different kinds of stuff.