feat(replay): Move sample rate tags into event context

[billyvg]

Move sample rate from tags into context as that is the more appropriate location for this type of data. This data is more for us to collect to debug rather than having as a top level field on the replay event.

[github-actions]

size-limit report 📦

Path	Size
@sentry/browser - ES5 CDN Bundle (gzipped + minified)	19.83 KB (-0.01% 🔽)
@sentry/browser - ES5 CDN Bundle (minified)	61.48 KB (0%)
@sentry/browser - ES6 CDN Bundle (gzipped + minified)	18.5 KB (-0.01% 🔽)
@sentry/browser - ES6 CDN Bundle (minified)	54.79 KB (0%)
@sentry/browser - Webpack (gzipped + minified)	20.22 KB (0%)
@sentry/browser - Webpack (minified)	66.19 KB (0%)
@sentry/react - Webpack (gzipped + minified)	20.25 KB (0%)
@sentry/nextjs Client - Webpack (gzipped + minified)	47.54 KB (0%)
@sentry/browser + @sentry/tracing - ES5 CDN Bundle (gzipped + minified)	26.75 KB (-0.01% 🔽)
@sentry/browser + @sentry/tracing - ES6 CDN Bundle (gzipped + minified)	25.03 KB (-0.01% 🔽)
@sentry/replay ES6 CDN Bundle (gzipped + minified)	43.7 KB (+0.04% 🔺)
@sentry/replay - Webpack (gzipped + minified)	38.48 KB (+0.06% 🔺)
@sentry/browser + @sentry/tracing + @sentry/replay - ES6 CDN Bundle (gzipped + minified)	61.03 KB (+0.04% 🔺)

[Lms24]

I'm wondering if there's a specific reason to set the sample rates as "top-level" fields rather than e.g. storing them in event.contexts.replay for example? Does it have something to do with contexts being hard to query for data analysis?
Just asking because usually we put stuff into contexts when collecting user options or additional information that's relevant for us to gather insights rather than relevant for the actual event.
Not saying this has to be changed, just curious ;)

[billyvg]

Ah didn't realize that. Storing in context makes more sense then. What's the best way to set context while we are preparing the event?

[Lms24]

I think we have two options:

1. Use the SDK to set the context. I think this is what we want to do eventually anyway, as outlined in Modify Sentry SDK to have a "replay" context, add replay_id to this context for transactions and errors #6741. So for example, when setupOnce() is called, we could call getCurrentHub().setContext('replay', {...}) and set the rates there. This however, will add this context to all events, including errors and transactions. I think relay-wise this is fine; just something we should keep in mind.

2. Alternatively, we can just append the context in prepareReplayEvent. This doesn't exclude doing option 1 in the future so it might be a good place to start. Plus, it'll only apply the fields to replay events. Something along these lines

  preparedEvent.contexts = {
    ...preparedEvent.contexts,
    replay: {
      ...(preparedEvent.contexts && preparedEvent.contexts.replay),
      replaysOnErrorSampleRate: options.replaysOnErrorSampleRate
      replaysSessionSampleRate: options.replaysSessionSampleRate
    }
  }

wdyt?

[billyvg]

I like the second option for now so that we're not attaching replay context to other events.

[Lms24]

Nice!