breakdown by a dimension to elementary.all_columns_anomalies

[Orensarfaty]

 - elementary.all_columns_anomalies:
      severity: warn
      where_expression: " type = ('promoting_started') "
      time_bucket:
        period: day
        count: 1
      tags: ["daily"]
      
      I would like to have an option to create this test for several types and to execute a different test for each column.
      another thing is to have an option to choose only several columns that we are interested of without all columns like the name of the test because this test taking a lot of time.
      
      thanks! 

[KarthikRajashekaran]

It would be nice to have a feature with dimension anomalies which can find sum(column_value) group by a column over period of timestamp value

Example as below ,its very tedious to do each product_name value , it would be nice to have group by product_name with sum(metric_value)

Looks like dimension_anomalies lacks the same

models:
  - name: metric_product
    config:
      elementary:
        timestamp_column: 'metric_date'
    columns:
      - name: metric_value
        tests:
          - elementary.column_anomalies:
              column_anomalies:
                - sum
              time_bucket:
                period: month
                count: 1
              days_back: 365
              config:
                # optional - change severity
                severity: ERROR
              anomaly_sensitivity: 1
              where_expression: "product_name = ‘ABC'"

          - elementary.column_anomalies:
              column_anomalies:
                - sum
              time_bucket:
                period: month
                count: 1
              days_back: 365
              config:
                # optional - change severity
                severity: ERROR
              anomaly_sensitivity: 1
              where_expression: "product_name = ‘DEF' "

          - elementary.column_anomalies:
              column_anomalies:
                - sum
              time_bucket:
                period: month
                count: 1
              days_back: 365
              config:
                # optional - change severity
                severity: ERROR
              anomaly_sensitivity: 1
              where_expression: "product_name = ‘MNO' "

[haritamar]

Hi all, This is now available in the both the all_column_anomalies and column_anomalies tests! The dimensions parameter can be used to add one or more dimension to the test.