Add @kbn/analytics-client to the shared libs to test any bundle size improvements

[afharo]

Given we know that the @elastic/analytics will at least always be imported by core, should we move it to the shared libs?

Originally posted by @pgayvallet in #129927 (comment)

[elasticmachine]

Pinging @elastic/kibana-core (Team:Core)

[Bamieh]

+1 to adding it as a shared dependency. This can be done at any point so no harm holding until it is actually used in multiple places

[afharo]

AFAIK, it'll only be imported in core for the client initialization and wherever the shippers are asynchronously imported to be registered (cloud and telemetry). The shippers imports are async so they should not affect the page load bundle size. However, it's true that, in the overall picture, we are loading the library 3 times.

@elastic/kibana-operations do you think it's enough of a reason to move it to the shared library?

[pgayvallet]

The shippers imports are async so they should not affect the page load bundle size

FWIW, The shippers are registered during the setup phase, so even if that technically doesn't affect the page load bundle, it would add additional requests during the phase that already is loading the most things.

[spalger]

If it's already loaded by core on every page load then there's little reason to keep it out of the shared-ui-deps.

[afharo]

Thanks for confirming! I'll create a quick follow-up PR next week to address this issue.

[afharo]

Is this still needed? When we initially created it, it referred to @elastic/analytics. That library included the built-in shippers, so it was loaded in 3 places: core, telemetry and cloud.

As we split the package into 3: @kbn/analytics-client (only used in core) and 2 shippers (one for telemetry and another one for cloud). I think adding @kbn/analytics-client to shared-ui-deps won't bring any improvement.

The shippers depend on some types of @kbn/analytics-client, but, AFAIK, that doesn't affect the bundle size because types are removed from the end bundles.

If you think it's still worth giving it a shot, I can try to put a quick PR to test if there's any benefit. What do you think?

[pgayvallet]

It seems fine then, we can probably close this

[afharo]

I'll close then. Please, feel free to reopen if you think it's still worth trying.