Repository navigation
error: "Snowflake Data Type is not supported" on "NUMBER" with non-zero scale #3146
Description
Expected Behavior
feast plan/apply completed without error in version 0.23.0 and I was able to query with get_historical_features.
Current Behavior
With the snowflake datatype NUMBER(38,6) I get:
NotImplementedError: The following Snowflake Data Type is not supported: DECIMAL -- Convert to DOUBLE
Steps to reproduce
$ feast init -t snowflake- Alter the table like so:
alter table SNOWY_FEAST.PUBLIC."amazed_dane_feast_driver_hourly_stats" add column temp_conv_rate NUMBER(38,6);
update SNOWY_FEAST.PUBLIC."amazed_dane_feast_driver_hourly_stats" set temp_conv_rate = "conv_rate";
alter table SNOWY_FEAST.PUBLIC."amazed_dane_feast_driver_hourly_stats" drop column "conv_rate";
alter table SNOWY_FEAST.PUBLIC."amazed_dane_feast_driver_hourly_stats" rename column temp_conv_rate to "conv_rate";
$ feast apply
Traceback (most recent call last):
...
raise NotImplementedError(
NotImplementedError: The following Snowflake Data Type is not supported: DECIMAL -- Convert to DOUBLE
Specifications
- Version:
Feast SDK Version: "feast 0.24.0" - Platform: x86_64 on Python 3.10.5
- Subsystem: Linux 5.4.203-
Activity
achals commentedon Aug 26, 2022
I think this is intentional cc @sfc-gh-madkins
sfc-gh-madkins commentedon Aug 27, 2022
@cburroughs my advice for right now would be to cast this to type double/float, as the error recommends. Let me see if I can get NUMBER with scale to work.
The issue stems from afew things. One protobuf doesnt support fixed point numbers, with scale. Two, how NUMBER is passed to pandas types.
cburroughs commentedon Aug 30, 2022
I installed from
masterto test it out. Looks good; thank you!