How to create a new Checkpoint

This guide will help you create a new Checkpoint, which allows you to couple an Expectation Suite with a data set to validate.

Note: As of Great Expectations version 0.13.7, we refer to “class-based Checkpoints” as these are now fully implemented as their own Python class. This is mainly a change “behind the scenes”. You can continue to use your existing legacy Checkpoint worfklows if you’re working with concepts from the stable API (e.g. Datasources). If you’re using concepts from the experimental API, please refer to the guides for class-based Checkpoints.


Docs for Legacy Checkpoints (<=0.13.7)

  1. First, run the CLI command below.

great_expectations checkpoint new my_checkpoint my_suite
  1. Next, you will be prompted to select a data asset you want to couple with the Expectation Suite. Note: The CLI currently only supports Datasources that are configured using the stable API. If you have set up a Datasource using the experimental API, please see the docs in the respective tab.

  2. You will then see a message that indicates the Checkpoint has been added to your project.

A checkpoint named `my_checkpoint` was added to your project!
- To edit this checkpoint edit the checkpoint file: /home/ubuntu/my_project/great_expectations/checkpoints/my_checkpoint.yml
- To run this checkpoint run `great_expectations checkpoint run my_checkpoint`

Docs for Class-Based Checkpoints (>=0.13.8)

  1. Instantiate a DataContext

The current version of class-based Checkpoints is not yet supported by the CLI. In order to configure and run Checkpoints, you will need to write some Python code. We suggest using a Jupyter notebook for convenience.

Create a new Jupyter Notebook and instantiate a DataContext by running the following lines:

import great_expectations as ge
context = ge.get_context()
  1. Create or copy a yaml config

You can create your own, or copy an example. For this example, we’ll demonstrate using a basic Checkpoint configuration with the SimpleCheckpoint class, which takes care of some defaults. Replace all names such as my_datasource with the respective DataSource, DataConnector, DataAsset, and Expectation Suite names you have configured in your great_expectations.yml.

config = """
name: my_checkpoint
config_version: 1
class_name: SimpleCheckpoint
  - batch_request:
      datasource_name: my_datasource
      data_connector_name: my_data_connector
      data_asset_name: MyDataAsset
        index: -1
    expectation_suite_name: my_suite

This is the minimum required to configure a Checkpoint that will run the Expectation Suite my_suite against the data asset MyDataAsset. See How to configure a new Checkpoint using test_yaml_config for advanced configuration options.

  1. Run context.test_yaml_config.


When executed, test_yaml_config will instantiate the component and run through a self_check procedure to verify that the component works as expected.

In the case of a Checkpoint, this means

  1. validating the yaml configuration,

  2. verifying that the Checkpoint class with the given configuration, if valid, can be instantiated, and

  3. printing warnings in case certain parts of the configuration, while valid, may be incomplete and need to be better specified for a successful Checkpoint operation.

The output will look something like this:

Attempting to instantiate class from config...
Instantiating as a SimpleCheckpoint, since class_name is SimpleCheckpoint
Successfully instantiated SimpleCheckpoint

Checkpoint class name: SimpleCheckpoint

If something about your configuration wasn’t set up correctly, test_yaml_config will raise an error.

  1. Check your stored Checkpoint config.

If the Store Backend of your Checkpoint Store is on the local filesystem, you can navigate to the checkpoints store directory that is configured in great_expectations.yml and find the configuration files corresponding to the Checkpoints you created.

  1. (Optional:) Test running the new Checkpoint.

Note that when test_yaml_config runs successfully, it saves the specified Checkpoint configuration to the Store backend configured for the Checkpoint Configuration store of your Data Context. This means that you can also test context.run_checkpoint, right within your notebook:

checkpoint_run_result = context.run_checkpoint(

Before running a Checkpoint, make sure that all classes and Expectation Suites referred to in the configuration exist.

When run_checkpoint returns, the checkpoint_run_result can then be checked for the value of the success field (all validations passed) and other information associated with running the specified actions.

For more advanced configurations of Checkpoints, please see How to configure a new Checkpoint using test_yaml_config.