You’re running into a known bug in the

column_anomalies

/ column monitoring query: when no

timestamp_column

is provided, the compiled SQL ends up with:

sql select   <your_columns>,   {{ elementary.null_timestamp() }} as start_bucket_in_data, from monitored_table

Note the trailing comma before

from

, which causes a syntax error. [[Redshift syntax](https://github.com/elementary-data/elementary/issues/1540); [Databricks syntax](https://github.com/elementary-data/elementary/issues/1554)] That comma is not in your

{{ elementary.null_timestamp() }}

macro; it’s in the calling macro (

column_monitoring_query.sql

) right after

start_bucket_in_data

. The relevant line in the package looks like:

sql {{ elementary.null_timestamp() }} as start_bucket_in_data,

So: - You do not need to change

null_timestamp()

itself. - The comma that needs to be removed is the one after

start_bucket_in_data

in the caller (

column_monitoring_query.sql

), in the branch that is used when

timestamp_column

is not configured. [[Databricks syntax](https://github.com/elementary-data/elementary/issues/1554)] Two workarounds until this is fixed in the package: 1. Add a

timestamp_column

in your test config so that the “else” branch (with the buggy comma) isn’t used. [[Slack Databricks thread](https://slack.com/archives/C02CTC89LAX/p1716306300184349)] 2. Locally patch the macro: copy

column_monitoring_query.sql

into your own project under

macros/

and remove the trailing comma after

start_bucket_in_data

in the “no timestamp” branch.