Syntax Error in BigQuery While Retrieving Columns that Start with a Number

[nishantgaurav-dev]

Expected Behavior

We have a feature view like this:

mbr_visits_fv_v1 = FeatureView(
	...
    schema=[
        Field(name="7_days_visits", dtype=Int64),
        Field(name="30_daysvisits", dtype=Int64),
    ],
    source=mbr_visits_bq_ds_v1
    ...
)

While querying this view as offline features I am expecting a data-frame containing desired columns without any error.

Current Behavior

Feast SDK is giving error like:
google.api_core.exceptions.BadRequest: 400 Syntax error: Missing whitespace between literal and alias at ...

I have printed some logs and observed that the generated BQ query contains a segment like this:

SELECT
  event_timestamp as event_timestamp,
  mbr_id AS mbr_id,
  7_days_visits as 7_days_visits
  30_days_visits as 30_days_visits
FROM `myproj.my_dataset.mbr_visits`
WHERE event_timestamp <= '2024-06-27T00:00:00'

This is incorrect BQ query.

It should be like this:

SELECT
  event_timestamp as event_timestamp,
  mbr_id AS mbr_id,
  `7_days_visits` as `7_days_visits`
  `30_days_visits` as `30_days_visits`
FROM `myproj.my_dataset.mbr_visits`
WHERE event_timestamp <= '2024-06-27T00:00:00'

The BQ expects that if a queried column starts with a number, the name should be enclosed in backtick (`) chars.

Steps to reproduce

Try querying a feature view with column name starting with number and use BigQuery data source.

Specifications

• Version: 0.34.1
• Platform: Mac
• Subsystem:

Possible Solution

Enclose all the BQ column names with backtick (`) chars. It will reduce the readability a little bit, but won't cause any performance impact.

[tmihalac]

"While querying this view as offline features I am expecting a data-frame containing desired columns without any error."
Can you also explain how did you query this view, which api and also a stack trace of the exception will be helpful