eeeeeeeeh that's not very simple. Maybe, you could certainly do so with a normal worker and the html

eeeeeeeeh that's not very simple. Maybe, you could certainly do so with a normal worker and the html rewriter, but Snippets have very limited CPU (5ms) and memory (2 MB), might be cutting it close, if it's a very small html page, maybe

CChaika eeeeeeeeh that's not very simple. Maybe, you could certainly do so with a normal...

Ping for toast•6/22/23, 5:41 PM

Yeah I'm using the HTML rewriter rn

Walshy•6/22/23, 5:42 PM

that should easily be able to fit

Walshy•6/22/23, 5:42 PM

We stream through HTMLRewriter, it doesn't use a lot of CPU

Walshy•6/22/23, 5:42 PM

if you were trying to parse it all, I'd say you'd have issues. Just changing a bit of text, sounds doable to me

WWalshy if you were trying to parse it all, I'd say you'd have issues. Just changing a b...

Ping for toast•6/22/23, 5:43 PM

Yup. Just finding an id, changing the text content

PPing for toast Yeah I'm using the HTML rewriter rn

ChaikaOP•6/22/23, 5:43 PM

if you already have it deployed, you should be able to see the CPU time from the existing worker

Ping for toast•6/22/23, 5:43 PM

Cool and just to be clear, the main advantage of the snippet engine is that it's completely free?

Walshy•6/22/23, 5:43 PM

and it's got the powerful ruleset engine for specific invocation

WWalshy and it's got the powerful ruleset engine for specific invocation

Ping for toast•6/22/23, 5:44 PM

Makes sense. I'll check my stats for the worker

ChaikaOP•6/22/23, 5:45 PM

I think the main advantage is being able to use them for really customizable rules using ruleset engine to trigger, even in front of Workers themselves. Workers are pretty cheap

Ping for toast•6/22/23, 5:45 PM

This would be excellent actually. A good part of my quota is just dedicated to a high traffic static site with a very simple worker

Walshy•6/22/23, 5:46 PM

hahaha err yeah, that should work

ChaikaOP•6/22/23, 5:46 PM

is it confirmed that cache isn't an extra subrequest, since you only get one?

Walshy•6/22/23, 5:46 PM

Cache doesn't count towards a subrequest afaik

Ping for toast•6/22/23, 5:47 PM

Is there any performance difference between workers and snippets?

Walshy•6/22/23, 5:47 PM

I'd have to test it to be 100% sure but it should be fine

PPing for toast Is there any performance difference between workers and snippets?

Walshy•6/22/23, 5:47 PM

Snippets are just Workers so no

WWalshy Cache doesn't count towards a subrequest afaik

kian•6/22/23, 5:49 PM

It counts towards the 50 on a normal Worker meowthinking

kian•6/22/23, 5:49 PM

Is snippets special?

Ping for toast•6/22/23, 5:49 PM

owch. let's see if we can trim this down

Ping for toast•6/22/23, 5:50 PM

wow pages with full SSR are only a couple ms slower than this simple worker... I have some optimization to do!

Kkian It counts towards the 50 on a normal Worker <:meowthinking:350680941804388352>

Walshy•6/22/23, 5:51 PM

oh it does? damn

Walshy•6/22/23, 5:51 PM

nevermind then

kian•6/22/23, 5:51 PM

it’s not supposed to but they also didnt plan to fix it since “50 is enough”

kian•6/22/23, 5:51 PM

Walshy•6/22/23, 5:51 PM

the 1 subrequest limit is a little awkward ablobsweats

Kkian it’s not supposed to but they also didnt plan to fix it since “50 is enough”

Walshy•6/22/23, 5:52 PM

ah...

Walshy•6/22/23, 5:52 PM

that makes sense

Walshy•6/22/23, 5:52 PM

yeah we should probably fix that now for Snippets

kian•6/22/23, 5:52 PM

well, fix as in move it to be inhouse subreq like KV/etc

Ping for toast•6/22/23, 5:53 PM

is there any detailed docs about the expensive parts of HTMLRewriter? It would be interesting to know what's consuming the largest amount of CPU time

kian•6/22/23, 5:53 PM

kj::Own<CacheClient> IoContext::getCacheClient() {
  // TODO(someday): Should Cache API requests be considered in-house? They are already not counted
  //   as subrequests in metrics and logs (like in-house requests aren't), but historically the
  //   subrequest limit still applied. Since I can't currently think of a use case for more than 50
  //   cache API requests per request, I'm leaving it as-is for now.
  limitEnforcer->newSubrequest(false);
  return getIoChannelFactory().getCache();
}

kj::Own<CacheClient> IoContext::getCacheClient() {
  // TODO(someday): Should Cache API requests be considered in-house? They are already not counted
  //   as subrequests in metrics and logs (like in-house requests aren't), but historically the
  //   subrequest limit still applied. Since I can't currently think of a use case for more than 50
  //   cache API requests per request, I'm leaving it as-is for now.
  limitEnforcer->newSubrequest(false);
  return getIoChannelFactory().getCache();
}

Walshy•6/22/23, 5:53 PM

that says

false

false

though

kian•6/22/23, 5:54 PM

because that param is isInHouseisInHouse

Walshy•6/22/23, 5:54 PM

I guess the newSubrequest is what actually does it

kian•6/22/23, 5:54 PM

Walshy•6/22/23, 5:54 PM

kian•6/22/23, 5:54 PM

virtual void newSubrequest(bool isInHouse) = 0;virtual void newSubrequest(bool isInHouse) = 0;

Walshy•6/22/23, 5:54 PM

want to PR Kian? LuL

ChaikaOP•6/22/23, 5:54 PM

You can still use cache API with Snippets, just check cache, if it's not there, redirect with a query parameter to not check it /s

Walshy•6/22/23, 5:54 PM

or i guess make an issue/discussion first and see what Kenton says

CChaika You can still use cache API with Snippets, just check cache, if it's not there, ...

Walshy•6/22/23, 5:55 PM

pls no

kian•6/22/23, 5:55 PM

sgtm

kian•6/22/23, 5:55 PM

ship it

Kkian ```cpp kj::Own<CacheClient> IoContext::getCacheClient() { // TODO(someday): Sh...

Ping for toast•6/22/23, 5:55 PM

https://github.com/cloudflare/workerd/blob/e1799c742439602804b9a38cc1c01a9bff796bb7/src/workerd/io/io-context.c%2B%2B#L910-L917

bkyerv•6/22/23, 6:02 PM

is cloudflare going to launch user management service like clerk dev or supabase auth?

Walshy•6/22/23, 6:09 PM

Yeah it'd be great to have AE for sure

sakuta•6/22/23, 6:09 PM

how effective is 1.1.1.1's malware blocking? is the list updated daily, hourly? what sources does it take them from?

kj::Own<CacheClient> IoContext::getCacheClient() { // TODO(someday): Should Cache API requests be considered in-house? They are already not counted // as subrequests in metrics and logs (like in-house requests aren't), but historically the // subrequest limit still applied. Since I can't currently think of a use case for more than 50 // cache API requests per request, I'm leaving it as-is for now. limitEnforcer->newSubrequest(false); return getIoChannelFactory().getCache(); }

eeeeeeeeh that's not very simple. Maybe, you could certainly do so with a normal worker and the html

Similar Threads

Similar Threads

Similar Threads