perf(runtime): improve rendering performance - #17857
Conversation
🦋 Changeset detectedLatest commit: 8514d2b The changes in this PR will be included in the next version bump. Not sure what this means? Click here to learn what changesets are. Click here if you're a maintainer who wants to add another changeset to this PR |
github-actions
Bot
added
the
pkg: astro
Merging this PR will improve performance by 30.93%
|
| Benchmark | BASE |
HEAD |
Efficiency | |
|---|---|---|---|---|
| ⚡ | large-array (BufferedRenderer per child) |
194.9 ms | 121 ms | +61.05% |
| ⚡ | many-slots (eager slot prerendering) |
6.5 ms | 4.4 ms | +49.51% |
| ⚡ | large-array [streaming] |
164.4 ms | 112.7 ms | +45.88% |
| ⚡ | many-components [streaming] |
8.2 ms | 5.7 ms | +43.54% |
| ⚡ | many-components (markHTMLString, isHTMLString, validateProps) |
9.6 ms | 7.5 ms | +28.15% |
| ⚡ | Rendering: streaming [true], .astro file |
208.8 ms | 165.7 ms | +25.98% |
| ⚡ | many-expressions [streaming] |
13.7 ms | 12.2 ms | +12.28% |
| ⚡ | Rendering: streaming [false], .astro file |
198.8 ms | 177.7 ms | +11.89% |
| ⚡ | many-expressions (renderChild dispatch, escapeHTML) |
14.8 ms | 13.4 ms | +10.58% |
Tip
Curious why performance improved? Comment @codspeedbot explain why performance improved on this PR, or directly use the CodSpeed MCP with your agent.
Comparing perf/server-render-allocations (8514d2b) with main (2fc7ce9)1
Footnotes
400137f to
e609ba6
Compare
| @@ -1,8 +1,35 @@ | |||
| import { escape } from 'html-escaper'; | |||
There was a problem hiding this comment.
I know, I know, this is scary, however the code here is incredibly hot (🔥 🔥), and as such it needs a very fast HTML escape solution. This code is between 2 to 3 times as fast depending on the workload and is really similar (albeit slightly tuned to Astro's use case) to rising stars libraries like https://github.com/SukkaW/fast-escape-html
In a further PR, I'd like to perhaps remove that dep completely and have our own escape.ts somewhere that we re-use.
There was a problem hiding this comment.
I am not worried about "being scary", but maybe this change should be part of its own PR:
- if we break something, we can revert only this change
- this PR doesn't have a test suite to cover the dep we're removing (we assume it's tested)
Princesseuh
changed the title
astro-factory
Bot
removed
the
pr: astro-review
ematipico
left a comment
ematipico
left a comment
There was a problem hiding this comment.
The changes make sense. I suggested to move the HTML escape into its own PR. Let's wait for when Matthew comes back, I would like his review too
| @@ -1,8 +1,35 @@ | |||
| import { escape } from 'html-escaper'; | |||
There was a problem hiding this comment.
I am not worried about "being scary", but maybe this change should be part of its own PR:
- if we break something, we can revert only this change
- this PR doesn't have a test suite to cover the dep we're removing (we assume it's tested)
![astro-factory[bot]](https://avatars.githubusercontent.com/u/44914786?s=60&v=4){kind=small}
There was a problem hiding this comment.
No findings.
Merge Requirement Assessment
Status: satisfied
- Improve SSR rendering performance by reducing allocations, avoiding V8 deopts, and using faster rendering paths — implemented across
packages/astro/src/runtime/server/(escape, render-child dispatch, template-result rendering, streaming engine, attribute spreading, slot handling, fetch-state slots) and covered by dedicated regression tests for frozen Astro globals, frozen render instructions, hydration attribute escaping, thenable-versus-render-instance precedence, and sync subtree expansion.
Questions
None.
Review Status
Verdict: ready to merge based on static review
Context: PR #17857 (#17857)
Scope: full PR diff (22 files) at head 70824a3c928e4421190d71d4d79d00fe423ab41d against base 92f23cc121d3a1a03c6fb901a519309662b333de
Changeset: present and covers astro
Limitations: none
Validation: Static review only; no project code, tests, builds, or checks were run.
GitHub context: collected
Git scope: PR base/head verified via read-only GitHub tools; 0 unresolved prior review threads.
This review was made by an LLM. The analysis may be wrong, and reports might be incorrect.
|
@ematipico We have several tests for HTML escaping, https://github.com/withastro/astro/blob/9870f95601690d9d98799b6fa78a0bc76165ee06/packages/astro/test/html-escape.test.ts https://github.com/withastro/astro/blob/9870f95601690d9d98799b6fa78a0bc76165ee06/packages/astro/test/units/render/escape.test.ts and a few more elsewhere. The code there has already changed through times between a custom version vs the dep vs other stuff |
|
LOL the package has like 4 tests 🤣 |
| return markHTMLString(content); | ||
| } | ||
|
|
||
| const HEAD_INSTRUCTION = Object.freeze( |
There was a problem hiding this comment.
bonus points for using Object.freeze
ab3f880 to
dc81a5a
Compare
3 participants
Changes
This PR improves the overall performance of our rendering pipeline, without changing the entire structure or anything, just mostly focused on reducing allocations, preventing V8 de-opt, and/or just faster code in some places.
This of course mostly improves SSR performance, but nonetheless, in local this makes the docs build 2s faster on my computer. In SSR stress benchmarks, this has resulted in a up to 70% improvement in raw ops/s, but it's of course not representing real work.
Testing
Tests should pass. Also added some tests for behavior that I found easy to regress by accident while working on this.
Docs
N/A