Saving Metrics

Saving metrics during Validation makes it easy to construct a new data series based on observed dataset characteristics computed by Great Expectations. That data series can serve as the source for a dashboard or overall data quality metrics, for example.

Storing metrics is still a beta feature of Great Expectations, and we expect configuration and capability to evolve rapidly.

Adding a MetricsStore

A MetricStore is a special store that can store Metrics computed during Validation. A Metric store tracks the run_id of the validation and the expectation suite name in addition to the metric name and metric kwargs.

In most cases, a MetricStore will be configured as a SQL database. To add a MetricStore to your DataContext, add the following yaml block to the “stores” section:

stores:
    #  ...
    metrics_store:  # You can choose any name for your metric store
        class_name: MetricStore
        store_backend:
            class_name: DatabaseStoreBackend
            # These credentials can be the same as those used in a Datasource configuration
            credentials: ${my_store_credentials}

The next time your DataContext is loaded, it will connect to the database and initialize a table to store metrics if one has not already been created. See the Metrics Reference for more information on additional configuration options.

Configuring a Validation Action

Once a MetricStore is available, it is possible to configure a new StoreMetricsAction to save metrics during validation.

Add the following yaml block to your DataContext validation operators configuration:

validation_operators:
  # ...
  action_list_operator:
    class_name: ActionListValidationOperator
    action_list:
      # ...
      - name: store_metrics
        action:
          class_name: StoreMetricsAction
          target_store_name: metrics_store  # This should match the name of the store configured above
          # Note that the syntax for selecting requested metrics will change in a future release
          requested_metrics:
            *:  # The asterisk here matches *any* expectation suite name
              # use the 'kwargs' key to request metrics that are defined by kwargs,
              # for example because they are defined only for a particular column
              # - column:
              #     Age:
              #       - expect_column_min_to_be_between.result.observed_value
              - statistics.evaluated_expectations
              - statistics.successful_expectations

The StoreMetricsValidationAction processes an ExpectationValidationResult and stores Metrics to a configured Store. Now, when your operator is executed, the requested metrics will be available in your database!

context.run_validation_operator('action_list_operator', (batch_kwargs, expectation_suite_name))

Last updated: Aug 13, 2020