Send PayPal correlation_id in analytics payload

[scannillo]

Summary

Currently, we cannot tie our analytics to a PayPal correlationID (aka clientMetadataID) because we are not sending the value in our analytics batch payload. We recently had a Slack inbound where this data would be extremely helpful for handoff to the underlying PP Web or PP NXO team to do further investigation on the specific PayPal transaction.

Changes

• Update FPTIBatchData to send correlation_id at the event level (I verified that I can see these values populate in FPTI event explorer)
• Update BTAnalyticsService to accept an optional correlationID param
• Send already-existant clientMetadataID in BTPayPalWebClient
• Collect clientMetadataID & send in analytics for BTPayPalNativeCheckoutClient

Feedback/Questions

• We don't have a testing infrastructure for BTPayPalNativeCheckoutClient set up right now, which I left a TODO for. It will be a handful of work, since MXO is hard to test (ticket DTBTSDK-3076)
• We don't have testing in place for our analytics (ticket DTBTSDK-2441)
• We could block this PR until the above ^ are taken care of, if folks feel strongly? What do you think?

Checklist

• Added a changelog entry

Authors

@scannillo

[jaxdesmarais]

We could block this PR until the above ^ are taken care of, if folks feel strongly? What do you think?

I personally don't think we need to wait on these. I feel like MXO being easily testable is a long ways off. With the analytics ticket maybe we can prioritize that in the next few sprints? I feel like that one may be a bit easier to tackle sooner but still not a blocker.

[scannillo]

I personally don't think we need to wait on these. I feel like MXO being easily testable is a long ways off. With the analytics ticket maybe we can prioritize that in the next few sprints? I feel like that one may be a bit easier to tackle sooner but still not a blocker.

Cool, yeh I'm down for not blocking this!

The testing story on MXO won't change, realistically. So we'll have to do something like PPCP with a wrapper protocol. It's possible, but a pain :/. We could also explore if OCMock makes things easier too (I captured this in the ticket).

[scannillo]

UI Test failures are unrelated :/

[jaxdesmarais]

🔥