[FEEDBACK] LabSDK: Use a similar bucketing mechanism for windowed-features

[AlmogBaku]

Background

Currently, feature values of a "windowed features" are calculated in a different mechanism than the way they're processed in the Natun operator:

• In Natun - we're calculating "on-the-fly" buckets of these aggregations and then performing a final aggregation when building the FeatureSet
• In LabSDK - we're calculating the rolling window based on the specified time we got,

Although the LabSDK has a much simpler approach and yields the same results, this is preventing us from pouring the historical data back to the Natun Operator and might yield inconsistencies.

What do you propose to do?

Calculate rolling features similar to the Natun implementation:

1. First, calculate the rolling data into "buckets" (retrieve the bucket name from go)
2. Then, calculate the "finalized" feature value only in the materialized view that being generated in fs.historical_get() is

What have you already tried?

No response

What else should we know?

We should set/get the bucket names via the go binding to prevent inconsistency.

[github-actions]

This issue is stale because it has been open 60 days with no activity. Remove stale label or comment or this will be closed in 14 days.

[github-actions]

This issue is stale because it has been open 60 days with no activity. Remove stale label or comment or this will be closed in 14 days.