feat: inline Response.arrayBuffer inside load functions during ssr #10535
Conversation
🦋 Changeset detectedLatest commit: 5d1e3cf The changes in this PR will be included in the next version bump. This PR includes changesets to release 1 package
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 |
|
Are there no native web APIs that we can use for this? I'd really like to not have to pull in a library to do this (which, by the look of it, I'm guessing is also going to be slower than doing this natively). It's not a whole lot, but it looks like this would also ship the |
|
|
||
| return u8.buffer; | ||
| } | ||
|
|
There was a problem hiding this comment.
Uint8Array.from could be optimized under the hood?
I think this is equivalent Uint8Array.from(atob(text), (c) => c.charCodeAt(0))
There was a problem hiding this comment.
I did some testing on node v20.5.1 Ryzen 9 5950X 32GB
Uint8Array.from(atob(text), (c) => c.charCodeAt(0)) takes ~74ms to decode 1MB (out) of data
const d = atob(text);
return Uint8Array.from({length: d.length}, (_, i) => d.charCodeAt(i));takes ~50ms
And using the for loop, ~17ms
Uint8Array.from(atob(text), (c) => c.charCodeAt(0)) is the only one with any major gc
| // The Uint16Arrray(Uint8Array(..)) is done on purpose | ||
| // this way the code points are padded with 0's | ||
| return btoa(new TextDecoder('utf-16').decode(new Uint16Array(new Uint8Array(buff)))); | ||
| } |
There was a problem hiding this comment.
in my own tests using kit's existing base64 worked well without needing all this. But maybe I was missing a use case.
https://github.com/sveltejs/kit/blob/master/packages/kit/src/runtime/server/page/crypto.js#L210-L239
There was a problem hiding this comment.
Node 20.6.1
In my testing kit's version is faster (~5x) for smaller sizes, at 1kib they are the same, at 32kib TextDecoder is 1.5x faster,
at 64kib TextDecoder is 2.5x and at 512kib is 10x.
for smaller sizes i don't think it matters if encoding a request takes 0.0006ms instead of 0.0001ms. But encoding 128kib is 0.5ms instead of 1.7ms
|
this looks great, thank you! |
closes #9672
Please don't delete this checklist! Before submitting the PR, please make sure you do the following:
Tests
pnpm testand lint the project withpnpm lintandpnpm checkChangesets
pnpm changesetand following the prompts. Changesets that add features should beminorand those that fix bugs should bepatch. Please prefix changeset messages withfeat:,fix:, orchore:.