Repository navigation
Handle Empty Dataframe write to feature store #5428
Description
Problem Statement:
For now, if you do provide the empty dataframe to the write to the online store, the feast does not throws exception nor it rejects the empty dataframe.
its not failing not only for remote store but for any. I think it's good to identify the empty data at the higher level than at the store class level.
Originally posted by @jyejare in #5422 (comment)
Example:
event_df = pd.DataFrame(
{
"driver_id": [1000, 1001],
"conv_rate": [],
"acc_rate": [],
"avg_daily_trips": [],
"event_timestamp": [],
"created": [],
}
)
# OR event a complete empty Dataframe like pd.DataFrame({})
# Perform the online write push
client_store.push(
push_source_name="driver_stats_push_source", df=event_df, to=PushMode.ONLINE
)
Activity
aniketpalu commentedon Dec 2, 2025
We have addressed this issue & this can be closed.
Reference PR: #5497
CC: @jyejare @ntkathole