feast
feast copied to clipboard
dynamodb doesn't seem to support materialization of multiple feature views having the same entity and join keys
Expected Behavior
let's say i have 2 feature views:
account_fv1 = FeatureView(
name="account_fv1",
entities=[account_entity, country_entity],
ttl=timedelta(weeks=1 * 5), # Set to be very long for example purposes only
schema=[
Field(name="feature_1", dtype=String),
Field(name="feature_2", dtype=String),
],
source=account_engagement_source,
tags={"team": "ranking"},
)
account_fv2 = FeatureView(
name="account_fv2",
entities=[account_entity, country_entity],
ttl=timedelta(weeks=1 * 5), # Set to be very long for example purposes only
schema=[
Field(name="some_other_feature", dtype=String),
Field(name="some_other_feature", dtype=String),
],
source=account_engagement_source,
tags={"team": "ranking"},
)
I expect when materializing separately, this feature view to Dynamo to be able to retrieve both of them.
Current Behavior
the last materialized feature view overrides the one materialized before because they have the same entity_id
Steps to reproduce
Specifications
- Version:latest
- Platform:linux and mac
- Subsystem:
Possible Solution
adding the feature view name as part of the entity_id or allow for concurrent or/and updates of the same key in case features have the same event_timestamp