⚠️ No Changeset found

Latest commit: 15f508a

Merging this PR will not cause a version bump for any packages. If these changes should not result in a new version, you're good to go. If these changes should result in a version bump, you need to add a changeset.

Click here to learn what changesets are, and how to add one.

Click here if you're a maintainer who wants to add a changeset to this PR

Noting the brief discussion Rich and I had about this: The fact that you need to be very careful about which things to read in handleLoad feels like a footgun. We're not sure yet if this is an unavoidable fact and that people writing these hooks are probably more aware of the nuances of load invalidation, or if we need to tweak the design a bit (for example only tracking stuff after the inner load was called)

Is it necessary to track anything at all inside the hooks? It feels like that should happen inside the root layout instead and the hook should only be used for logging, modifying, ...

If you implement an auth guard you may want to ensure it runs on every request

Yeah sure, but would it be a problem to always run the hook when the load function chain starts? It feels more intuitive to me, very similar to how the handle hook works.

handleLoad runs for each load function, not once at the beginning for all.

Ah alright. But still, why do we need to track anything then? If your page load function depends on something that runs inside handleLoad it will always be there if it always runs. I don't see why you would need tracking. Maybe I just need an example that makes use of that functionality

I believe there was discussion around providing an untracked version of event properties alongside the main tracked one at some point?

Something like this would greatly help us at Sentry. For example to obtain route information.
Right now we rely on a brittle mechanism of avoiding to invoke the invalidation trigger.

Worth noting that we don't alter any behaviour based on reading these values so I don't think it should cause an invalidation.

@Lms24 did fly a bit under the radar, but a mechanism to not cause reruns from reading properties exists in 2.0: https://kit.svelte.dev/docs/load#rerunning-load-functions-untracking-dependencies

In fact, it's used in this PR - else handleLoad doesn't make much sense to have in many cases.

Hi, is there anything I can help with to get this merged?

Hi, is there anything I can help with to get this merged?

Hi Lukas. There's still some deliberation ongoing due to the concern mentioned in #11313 (comment) and we're not sure how to address this yet.

Gonna go ahead and close this, since a) we never found a satisfactory answer to the concerns about invalidation, b) we're moving to a world where load is deprioritised in favour of remote functions, and c) the underlying issue is solved by #13899