Stabilize state aggregate

[WireBaron]

State Aggregate

What evidence do we have the feature is being used

We see over 50 instance using state aggregates in the past 6 months in function telemetry.

Why do we feel this feature is ready to be stable

This feature has been around for a long time and is being used by many customers. We did make one recent change to allow users to more efficiently create state aggregates using integer states, but the change was fairly minor. We also want to make use of this aggregate in some upcoming user facing examples and would prefer it stabilized for that.

Is there any known further work needed on this feature after stabilization

There is some concern that the transient state we use during aggregation can grow too large, especially with string states. We may want to improve this to not store multiple copies of the same string.

Are there any compatibility concerns that may arise during future work on this feature

None known

Feature History

• Experimental release version: 1.5.0
• Last version modifying on-disk format: 1.13.0
• Target stabilization version: 1.14.0

Stabilization checklist:

• Ensure tests exist for all public API
• Ensure API documentation exists and is accurate
• Remove toolkit_experimental tags and update test usages
• Add arrow operators for accessors if applicable
• Ensure arrow operators have test coverage
• If present, ensure combine and rollup are tested
• Add serialization tests for text format
• Add serialization tests for on disk format
• Add upgrade tests
• Add continuous aggregate test
• Add feature level documentation

[WireBaron]

As the timeline aggregate has all the functionality of the state aggregate and then some, we've decided to replace state_agg with the timeline_agg version. The current state_agg will be renamed (TDB, potentially `simple_state_agg) and left experimental while we determine if there is a good use case for it.