TestRunner - Running Table-Level Tests

Execute data quality tests against tables in OpenMetadata using the TestRunner API

TestRunner - Running Table-Level Tests

The TestRunner class provides a fluent API for executing data quality tests against tables cataloged in OpenMetadata. It automatically fetches table metadata and service connections, allowing you to run tests with minimal configuration.

Table of contents

• Overview
• Basic Usage
• Complete Example
• Running Tests from OpenMetadata’s UI
• Customizing Test Metadata
• Configuring Row Count Computation
• Test Runner Configuration
• Understanding Test Results
• Integration with ETL Workflows
• Error Handling
• Best Practices
• Next Steps

⚠️ If you’re using OpenMetadata Cloud to run OpenMetadata, please refer to the section about External Secrets Managers

Overview

TestRunner enables you to:

• Execute tests defined in code against cataloged tables
• Run tests previously configured in the OpenMetadata UI
• Load test definitions from YAML workflow files
• Validate data at the table and column levels
• Get detailed test results for programmatic handling

Basic Usage

Creating a TestRunner

Create a runner for a specific table using its fully qualified name (FQN):

from metadata.sdk import configure
from metadata.sdk.data_quality import TestRunner

# Configure SDK first
configure(
    host="http://localhost:8585/api",
    jwt_token="your-jwt-token"
)

# Create runner for a table
runner = TestRunner.for_table("MySQL.ecommerce.public.customers")

The table FQN format is: {service}.{database}.{schema}.{table}

Adding Tests

Add test definitions to the runner:

from metadata.sdk.data_quality import (
    TableRowCountToBeBetween,
    ColumnValuesToBeNotNull,
    ColumnValuesToBeUnique
)

# Add a table-level test
runner.add_test(
    TableRowCountToBeBetween(min_count=1000, max_count=100000)
)

# Add column-level tests
runner.add_test(
    ColumnValuesToBeNotNull(column="email")
)

runner.add_test(
    ColumnValuesToBeUnique(column="customer_id")
)

Adding Multiple Tests

Use add_tests() to add several tests at once:

runner.add_tests(
    TableRowCountToBeBetween(min_count=1000),
    ColumnValuesToBeNotNull(column="email"),
    ColumnValuesToBeUnique(column="customer_id"),
    ColumnValuesToMatchRegex(
        column="email",
        regex=r"^[a-zA-Z0-9._%+-]+@[a-zA-Z0-9.-]+\.[a-zA-Z]{2,}$"
    )
)

Running Tests

Execute all configured tests:

results = runner.run()

# Process results
for result in results:
    test_case = result.testCase
    test_result = result.testCaseResult

    print(f"Test: {test_case.name.root}")
    print(f"Status: {test_result.testCaseStatus}")
    print(f"Result: {test_result.result}")

Complete Example

Here’s a complete example of testing a customer table:

from metadata.sdk import configure
from metadata.sdk.data_quality import (
    TestRunner,
    TableRowCountToBeBetween,
    ColumnValuesToBeNotNull,
    ColumnValuesToBeUnique,
    ColumnValuesToMatchRegex,
    ColumnValuesToBeBetween
)

# Configure SDK
configure(
    host="http://localhost:8585/api",
    jwt_token="your-jwt-token"
)

# Create runner
runner = TestRunner.for_table("MySQL.ecommerce.public.customers")

# Define data quality tests
runner.add_tests(
    # Table should have between 1000 and 1M customers
    TableRowCountToBeBetween(min_count=1000, max_count=1000000),

    # Required fields must not be null
    ColumnValuesToBeNotNull(column="customer_id"),
    ColumnValuesToBeNotNull(column="email"),
    ColumnValuesToBeNotNull(column="created_at"),

    # Customer ID must be unique
    ColumnValuesToBeUnique(column="customer_id"),

    # Email must match valid pattern
    ColumnValuesToMatchRegex(
        column="email",
        regex=r"^[a-zA-Z0-9._%+-]+@[a-zA-Z0-9.-]+\.[a-zA-Z]{2,}$"
    ),

    # Age should be reasonable
    ColumnValuesToBeBetween(column="age", min_value=18, max_value=120)
)

# Run tests
results = runner.run()

# Handle results
all_passed = all(
    result.testCaseResult.testCaseStatus == "Success"
    for result in results
)

if all_passed:
    print("✓ All data quality tests passed!")
else:
    print("✗ Some tests failed:")
    for result in results:
        if result.testCaseResult.testCaseStatus != "Success":
            print(f"  - {result.testCase.name.root}: {result.testCaseResult.result}")

Running Tests from OpenMetadata UI

Instead of defining tests in code, you can run tests that data stewards have configured in the OpenMetadata UI. This enables a collaborative workflow where:

• Data stewards define and maintain test criteria in the UI
• Engineers execute those tests automatically in pipelines

from metadata.sdk.data_quality import TestRunner

# Create runner
runner = TestRunner.for_table("BigQuery.analytics.customer_360")

# Run all tests defined in OpenMetadata for this table
results = runner.run()

# The runner automatically executes all tests configured in the UI
for result in results:
    print(f"{result.testCase.name.root}: {result.testCaseResult.testCaseStatus}")

This approach ensures:

• Test definitions stay synchronized with business requirements
• Engineers don’t need to modify code when test criteria change
• All stakeholders own data quality

Customizing Test Metadata

You can customize test names, display names, and descriptions:

from metadata.sdk.data_quality import ColumnValuesToBeNotNull

# Use fluent API for customization
test = ColumnValuesToBeNotNull(column="email") \
    .with_name("email_not_null_check") \
    .with_display_name("Email Not Null Validation") \
    .with_description("Ensures all customer records have valid email addresses")

runner.add_test(test)

Or pass values directly to the constructor:

test = ColumnValuesToBeNotNull(
    column="email",
    name="email_not_null_check",
    display_name="Email Not Null Validation",
    description="Ensures all customer records have valid email addresses"
)

Configuring Row Count Computation

Some tests support computing the number and percentage of rows that passed or failed:

from metadata.sdk.data_quality import ColumnValuesToBeBetween

test = ColumnValuesToBeBetween(
    column="age",
    min_value=18,
    max_value=120
).with_compute_row_count(True)

runner.add_test(test)

This provides detailed metrics about test failures, useful for:

• Identifying the scope of data quality issues
• Prioritizing remediation efforts
• Tracking data quality trends over time

Test Runner Configuration

Customize the test runner behavior using the setup() method:

from metadata.generated.schema.metadataIngestion.workflow import LogLevels

runner = TestRunner.for_table("MySQL.ecommerce.public.orders")

# Configure runner behavior
runner.setup(
    force_test_update=True,        # Update existing tests
    log_level=LogLevels.DEBUG,     # Set logging level
    raise_on_error=False,          # Don't raise exceptions on errors
    success_threshold=90,          # Pass if 90% of tests succeed
    enable_streamable_logs=True    # Enable streamable log output
)

# Add and run tests
runner.add_test(TableRowCountToBeBetween(min_count=100))
results = runner.run()

Configuration Parameters

Parameter	Type	Default	Description
force_test_update	bool	False	Force update even if tests already exist
log_level	LogLevels	INFO	Logging level (DEBUG, INFO, WARN, ERROR)
raise_on_error	bool	False	Raise exceptions if test data already exists
success_threshold	int	90	Percentage threshold for overall success
enable_streamable_logs	bool	False	Enable streamable log output

Understanding Test Results

Test results contain detailed information about test execution:

results = runner.run()

for result in results:
    # Test case information
    test_case = result.testCase
    print(f"Test Name: {test_case.name.root}")
    print(f"Test Type: {test_case.testDefinition.name}")

    # Test result information
    test_result = result.testCaseResult
    print(f"Status: {test_result.testCaseStatus}")  # Success, Failed, Aborted
    print(f"Result: {test_result.result}")           # Detailed result message
    print(f"Timestamp: {test_result.timestamp}")

    # Row-level statistics (if enabled)
    if test_result.passedRows is not None:
        print(f"Passed Rows: {test_result.passedRows}")
        print(f"Failed Rows: {test_result.failedRows}")
        print(f"Passed %: {test_result.passedRowsPercentage}")

Test Status Values

• Success: Test passed all validation criteria
• Failed: Test did not meet validation criteria
• Aborted: Test execution was interrupted or could not complete

Integration with ETL Workflows

Integrate TestRunner into your ETL pipelines:

import pandas as pd
from sqlalchemy import create_engine
from metadata.sdk import configure
from metadata.sdk.data_quality import TestRunner, TableRowCountToBeBetween

# Configure SDK
configure(host="http://localhost:8585/api", jwt_token="token")

# ETL: Extract
source_df = pd.read_csv("data/customers.csv")

# ETL: Transform
transformed_df = transform_data(source_df)

# ETL: Load
engine = create_engine("postgresql://user:pass@localhost/db")
transformed_df.to_sql("customers", engine, if_exists="replace")

# Validate: Run data quality tests
runner = TestRunner.for_table("Postgres.analytics.public.customers")
runner.add_test(
    TableRowCountToBeBetween(min_count=len(transformed_df))
)
results = runner.run()

# Check results
if all(r.testCaseResult.testCaseStatus == "Success" for r in results):
    print("✓ Data quality validation passed")
else:
    print("✗ Data quality validation failed")
    # Optionally rollback or alert
    rollback_transaction()

Error Handling

Handle potential errors gracefully:

from metadata.sdk.data_quality import TestRunner

try:
    runner = TestRunner.for_table("MySQL.ecommerce.public.customers")
    runner.add_test(TableRowCountToBeBetween(min_count=100))
    results = runner.run()

except ValueError as e:
    print(f"Configuration error: {e}")
    # Table not found or service not configured

except Exception as e:
    print(f"Test execution failed: {e}")
    # Network issues, authentication problems, etc.

Best Practices

1. Use descriptive test names: Make test failures easy to understand

   test = TableRowCountToBeBetween(min_count=1000) \
       .with_name("daily_customer_count_check") \
       .with_description("Validates daily customer ingestion meets minimum threshold")
   

2. Leverage UI-defined tests: Let data stewards define test criteria

   # Just run - no need to define tests in code
   runner = TestRunner.for_table("Table.FQN")
   results = runner.run()
   

3. Handle results programmatically: Don’t just print - take action

   if any(r.testCaseResult.testCaseStatus == "Failed" for r in results):
       rollback_transaction()
   

4. Use appropriate thresholds: Set realistic min/max values based on data patterns

5. Combine table and column tests: Ensure both structural and content quality

If your organization uses an external secrets manager (AWS, Azure, GCP), see External Secrets Managers before using the TestRunner API.

Next Steps

• Learn about DataFrame Validation for validating transformations
• Review the Test Definitions Reference for all available tests
• Explore Advanced Usage including YAML workflows