refactor(platform-server): Add an ssr benchmark setup. #57647
Conversation
JeanMeche
force-pushed
the
ssr-benchmark
branch
4 times, most recently
from
09f7a37
to
a3d87fc
Compare
pullapprove
bot
requested review from
atscott,
crisbeto,
dylhunn and
josephperrott
|
This PR does impact all the packages because of visibility changes, @josephperrott's approval should be enough for those. |
JeanMeche
requested review from
AndrewKushnir
and removed request for
atscott,
crisbeto and
dylhunn
JeanMeche
force-pushed
the
ssr-benchmark
branch
6 times, most recently
from
df3c3a3
to
3b906fc
Compare
JeanMeche
force-pushed
the
ssr-benchmark
branch
from
3b906fc
to
15a1ec6
Compare
In order to investigate the performances of SSR, this commit introduces a benchmark suite which will measure several step of the rendering.
JeanMeche
force-pushed
the
ssr-benchmark
branch
from
15a1ec6
to
a9c2e7c
Compare
There was a problem hiding this comment.
Looks great, thanks for implementing this @JeanMeche 👍 Just left a few comments.
@alan-agius4 could you please take a look at this change when you get a chance?
| @@ -20,6 +20,9 @@ import { Version } from '@angular/core'; | |||
| // @public | |||
| export const BEFORE_APP_SERIALIZED: InjectionToken<readonly (() => void | Promise<void>)[]>; | |||
|
|
|||
| // @public (undocumented) | |||
| export const ENABLE_DOMINO: InjectionToken<boolean>; | |||
There was a problem hiding this comment.
I think we should prevent exposing those extra symbols into the public API (use the ɵ in their names).
| render, | ||
| initData, | ||
| PERFORMANCE_MARK_PREFIX, | ||
| } from '../../dist/out/darwin_arm64-fastbuild/bin/modules/ssr-benchmarks/build/server/main.server.mjs'; |
There was a problem hiding this comment.
I'm wondering if this path would always be stable (probably not, depending on OS). Is there a different option to reference that? // cc @alan-agius4
| this.finishTask = this.pendingTasks.add(); | ||
| } | ||
|
|
||
| data$ = (this.getObservableFromTransferState() ?? from(testData())).pipe( |
There was a problem hiding this comment.
Quick question: why do we need to use observables here? I think it'd be great to keep this logic synchronous, so we don't measure this extra time in benchmarks.
| import {TRANSFER_STATE_SERIALIZATION_PROVIDERS} from './transfer_state'; | ||
|
|
||
| export const INTERNAL_SERVER_PLATFORM_PROVIDERS: StaticProvider[] = [ | ||
| {provide: DOCUMENT, useFactory: _document, deps: [Injector]}, | ||
| {provide: PLATFORM_ID, useValue: PLATFORM_SERVER_ID}, | ||
| {provide: PLATFORM_ID, useValue: PLATFORM_SERVER_ID}, |
There was a problem hiding this comment.
This looks like a duplicate of a previous line.
| @@ -57,8 +63,13 @@ export const INTERNAL_SERVER_PLATFORM_PROVIDERS: StaticProvider[] = [ | |||
| ]; | |||
|
|
|||
| function initDominoAdapter() { | |||
| const enableDomino = inject(ENABLE_DOMINO, {optional: true}) ?? true; | |||
There was a problem hiding this comment.
Since we have it in 2 places and use a default value (of true), it'd be great to extract this logic into a separate function to make sure the default value remains the same in all places.
| @@ -89,11 +100,14 @@ export class ServerModule {} | |||
|
|
|||
| function _document(injector: Injector) { | |||
| const config: PlatformConfig | null = injector.get(INITIAL_CONFIG, null); | |||
| const enableDomino = injector.get(ENABLE_DOMINO, true); | |||
There was a problem hiding this comment.
Looks like another location for the default value, we can also replace that with a utility function (see previous comment).
In order to investigate the performances of SSR, this commit introduces a benchmark suite which will measure several step of the rendering.