Validity metrics

Last modified on 24-Jul-24

Use a validity metric in a check to surface invalid or unexpected values in your dataset.

checks for dim_customer:
# Check for valid values
  - invalid_count(email_address) = 0:
      valid format: email
  - invalid_percent(english_education) = 0:
      valid length: 100
  - invalid_percent(total_children) <= 2:
      valid max: 6
  - invalid_percent(marital_status) = 0:
      valid max length: 10
  - invalid_count(number_cars_owned) = 0:
      valid min: 1
  - invalid_percent(marital_status) = 0:
      valid min length: 1
  - invalid_percent(last_name) < 5%:
      invalid regex: (?:XX)
  - invalid_count(house_owner_flag) = 0:
      valid values: [0, 1]

checks for dim_customer:
# Check for invalid values
  - invalid_count(first_name) = 0:
      invalid values: [Antonio]
  - invalid_count(number_cars_owned) = 0:
      invalid values: [0, 3] 

✖️    Requires Soda Core Scientific (included in a Soda Agent)
✔️    Supported in Soda Core
✔️    Supported in Soda Library + Soda Cloud
✔️    Supported in Soda Cloud Agreements + Soda Agent
✔️    Available as a no-code check with a self-hosted Soda Agent connected to any
        Soda-supported data source, except Spark, and Dask and Pandas
        OR
        with a Soda-hosted Agent connected to a BigQuery, Databricks SQL, MS SQL Server,
        MySQL, PostgreSQL, Redshift, or Snowflake data source

Define checks with validity metrics
    Specify valid or invalid values
    Specify valid format
    Troubleshoot valid format and values
    Failed row samples
Optional check configurations
List of validity metrics
List of configuration keys
List of valid formats
List of comparison symbols and phrases
Go further

Define checks with validity metrics

In the context of SodaCL check types, you use validity metrics in standard checks. Refer to Standard check types for exhaustive configuration details.

You can use all validity metrics in checks that apply to individual columns in a dataset; you cannot use validity metrics in checks that apply to entire datasets. Identify the column by adding a value in the argument between brackets in the check.

• You must use a configuration key:value pair to define what qualifies as an valid value or invalid value.
• If you wish, you can add a % character to the threshold for a invalid_percent metric for improved readability. This character does not behave as a wildard in this context.

checks for dim_customer
  - invalid_count(number_cars_owned) = 0:
      valid min: 1

You can use validity metrics in checks with fixed thresholds, or relative thresholds, but not change-over-time thresholds. See Checks with fixed thresholds for more detail.

checks for dim_reseller:
# a check with a fixed threshold
  - invalid_count(email_address) = 0:
      valid format: email
# a check with a relative threshold
  - invalid_percent(english_education) < 3%:
      valid max length: 100

Specify valid or invalid values

Use a nested configuration key:value pair to provide your own definition of a valid or invalid value. There are several configuration keys that you can use to define what qualifies as valid; the examples below illustrate the use of just a few config keys. See a complete List of configuration keys below.

A check that uses a validity metric has six mutable parts:

The example below defines two checks. The first check applies to the column house_owner_flag. The valid values configuration key specifies that if a row in that column contains anything other than the two valid values in the list, Soda registers them as invalid. The check fails if Soda discovers any values that are not 0 or 1.

• Values in a list must be enclosed in square brackets.
• Known issue: Do not wrap numeric values in single quotes if you are scanning data in a BigQuery data source.

The second check uses a regular expression to define what qualifies as an invalid value in the last_name column so that any values that match the pattern defined by the regex qualify as invalid.

checks for dim_customer:
  - invalid_count(house_owner_flag) = 0:
      valid values: [0, 1]
  - invalid_count(last_name) = 0:
      invalid regex: (?:XX)

First check:

Second check:

The invalid values configuration key specifies that if a row in that column contains the invalid values in the list, Soda registers them as invalid. In the example below, the check fails if Soda discovers any values that are Antonio.

Values in a list must be enclosed in square brackets.

checks for dim_customer:
  - invalid_count(first_name) = 0:
      invalid values: [Antonio]

Specify valid format

If the data type of the column you are checking is TEXT (such as character, character varying, or string) then you can use the valid format configuration key. This config key uses built-in values that test the data in the column for specific formats, such as email address format, date format, or uuid format. See List of valid formats below.

The check below validates that all values in the email_address column conform to an email address format.

checks for dim_customer:
  - invalid_percent(email_address) = 0:
      valid format: email

Troubleshoot valid format and values

Problem: You are using a valid format to test the format of values in a column and the CLI returns the following error message when you run a scan.

  | HINT:  No operator matches the given name and argument types. You might need to add explicit type casts.

Error occurred while executing scan.
  | unsupported operand type(s) for *: 'Undefined' and 'int'

Solution: The error indicates that the data type of the column is not TEXT. Adjust your check to use a different configuration key, instead.

Problem: Using an invalid_count check, the list of valid_values includes a value with a single quote, such as Tuesday's orders. During scanning, he check results in and error because it does not recognize the special character.

Solution: When using single-quoted strings, any single quote ' inside its contents must be doubled to escape it. For example, Tuesday''s orders.

Failed row samples

Checks with validity metrics automatically collect samples of any failed rows to display Soda Cloud. The default number of failed row samples that Soda collects and displays is 100.

If you wish to limit or broaden the sample size, you can use the samples limit configuration in a check with a validity metric. You can add this configuration to your checks YAML file for Soda Library, or when writing checks as part of an agreement in Soda Cloud.

checks for dim_customer:
  - invalid_percent(email_address) < 50:
      samples limit: 2

For security, you can add a configuration to your data source connection details to prevent Soda from collecting failed rows samples from specific columns that contain sensitive data. Refer to Disable failed rows sampling for specific columns.

Alternatively, you can set the samples limit to 0 to prevent Soda from collecting and sending failed rows samples for an individual check, as in the following example.

checks for dim_customer:
  - invalid_percent(email_address) < 50:
      samples limit: 0

You can also use a samples columns configuration to a check to specify the columns for which Soda must implicitly collect failed row sample values, as in the following example. Soda only collects this check’s failed row samples for the columns you specify in the list.

Note that the comma-separated list of samples columns does not support wildcard characters (%).

checks for dim_employee:
  - invalid_count(gender) = 0:
      valid values: ["M", "Q"]
      samples columns: [employee_key, first_name]

To review the failed rows in Soda Cloud, navigate to the Checks dashboard, then click the row for a check for validity values. Examine failed rows in the Failed Rows tab; see Examine failed row samples for further details.

Optional check configurations

Example with check name

checks for dim_customer:
  - invalid_count(first_name) = 0 :
      valid min length: 2
      name: First name has 2 or more characters

Example with alert configuration

  - invalid_count(house_owner_flag):
      valid values: [0, 1]
      warn: when between 1 and 5
      fail: when > 6  

Example with in-check filter

checks for dim_customer:
  - invalid_percent(marital_status) = 0:
      valid max length: 1
      filter: total_children = 0

Example with quotes

checks for dim_customer:
  - invalid_count("number_cars_owned") = 0:
      valid min: 1

Example with for each

for each dataset T:
  datasets:
    - dim_customer
    - dim_customer_%
  checks:
    - invalid_count(email_address) = 0:
        valid format: email

Example with dataset filter

filter CUSTOMERS [daily]:
  where: TIMESTAMP '{ts_start}' <= "ts" AND "ts" < TIMESTAMP '${ts_end}'

checks for CUSTOMERS [daily]:
  - invalid_count(email_address) = 0:
      valid format: email

List of validity metrics

List of configuration keys

The column configuration key:value pair defines what SodaCL ought to consider as valid values.

List of valid formats

• Though table below lists valid formats, the same apply for invalid formats.
• Valid formats apply only to columns using data type TEXT, not DATE or NUMBER.
• The Soda Library package for MS SQL Server has limited support for valid formats. See the separate list below of formats supported for MS SQL Server.

Formats supported with Soda for MS SQL Server

List of comparison symbols and phrases

 = 
 < 
 >
 <=
 >=
 !=
 <> 
 between 
 not between 

Go further

• Use validity metrics in checks with alert configurations to establish warn and fail zones
• Use validity metrics in checks to define ranges of acceptable thresholds using boundary thresholds.
• Need help? Join the Soda community on Slack.
• Reference tips and best practices for SodaCL.
  

Was this documentation helpful?

What could we do to improve this page?

• Suggest a docs change in GitHub.
• Share feedback in the Soda community on Slack.

Documentation always applies to the latest version of Soda products
Last modified on 24-Jul-24

Sign up for Soda