We must use the Databricks SDK for Python in this project. It is a toolkit for our project.
If something doesn't naturally belong to the WorkspaceClient, it must go through a "mixin" process before it can be used with the SDK.
Imagine the WorkspaceClient as the main control center and the "mixin" process as a way to adapt other things to work with it.
You can find an example of how mixins are used with StatementExecutionExt. There's a specific example of how to make something
work with the WorkspaceClient using StatementExecutionExt. This example can help you understand how mixins work in practice.
Favoring standard libraries over external dependencies, especially in specific contexts like Databricks, is a best practice in software development.
There are several reasons why this approach is encouraged:
- Standard libraries are typically well-vetted, thoroughly tested, and maintained by the official maintainers of the programming language or platform. This ensures a higher level of stability and reliability.
- External dependencies, especially lesser-known or unmaintained ones, can introduce bugs, security vulnerabilities, or compatibility issues that can be challenging to resolve. Adding external dependencies increases the complexity of your codebase.
- Each dependency may have its own set of dependencies, potentially leading to a complex web of dependencies that can be difficult to manage. This complexity can lead to maintenance challenges, increased risk, and longer build times.
- External dependencies can pose security risks. If a library or package has known security vulnerabilities and is widely used, it becomes an attractive target for attackers. Minimizing external dependencies reduces the potential attack surface and makes it easier to keep your code secure.
- Relying on standard libraries enhances code portability. It ensures your code can run on different platforms and environments without being tightly coupled to specific external dependencies. This is particularly important in settings like Databricks, where you may need to run your code on different clusters or setups.
- External dependencies may have their versioning schemes and compatibility issues. When using standard libraries, you have more control over versioning and can avoid conflicts between different dependencies in your project.
- Fewer external dependencies mean faster build and deployment times. Downloading, installing, and managing external packages can slow down these processes, especially in large-scale projects or distributed computing environments like Databricks.
- External dependencies can be abandoned or go unmaintained over time. This can lead to situations where your project relies on outdated or unsupported code. When you depend on standard libraries, you have confidence that the core functionality you rely on will continue to be maintained and improved.
While minimizing external dependencies is essential, exceptions can be made case-by-case. There are situations where external dependencies are justified, such as when a well-established and actively maintained library provides significant benefits, like time savings, performance improvements, or specialized functionality unavailable in standard libraries.
When adding a new entry to the known.json file, ensure that the entry is unique and follows the correct format. The known.json file
stores information about known (in)compatibilities with Unity Catalog. The file speeds up static code analysis and prevents false positives.
- load into virtual environment
- run
make knownto update the file - review the changes and fold the imported packages
- remove any false positives and add any missing entries
- commit the changes
When you introduce a change in the code, specifically a deeply technical one, please ensure that the change provides same or improved set of capabilities. PRs that remove existing functionality shall be properly discussed and justified.
When writing code, divide it into two main parts: Components for API Interaction and Components for Business Logic. API Interaction should only deal with talking to external systems through APIs. They are usually integration-tested, and mocks are simpler. Business Logic handles the actual logic of your application, like calculations, data processing, and decision-making.
Keep API components simple. In the components responsible for API interactions, try to keep things as straightforward as possible. Refrain from overloading them with complex logic; instead, focus on making API calls and handling the data from those calls.
Inject Business Logic. If you need to use business logic in your API-calling components, don't build it directly there. Instead, inject (or pass in) the business logic components into your API components. This way, you can keep your API components clean and flexible, while the business logic remains separate and reusable.
Test your Business Logic. It's essential to test your business logic to ensure it works correctly and thoroughly. When writing unit tests, avoid making actual API calls - unit tests are executed for every pull request, and take seconds to complete. For calling any external services, including Databricks Connect, Databricks Platform, or even Apache Spark, unit tests have to use "mocks" or fake versions of the APIs to simulate their behavior. This makes testing your code more manageable and catching any issues without relying on external systems. Focus on testing the edge cases of the logic, especially the scenarios where things may fail. See this example as a reference of an extensive unit test coverage suite and the clear difference between unit tests and integration tests.
See https://mypy.readthedocs.io/en/stable/cheat_sheet_py3.html for more details
We don't use if typing.TYPE_CHECKING: in our codebase, because PyCharm and mypy can handle it without it and Refactoring -> Move may produce incorrect results.
- Add
assert ... is not Noneif it's a body of a method. Example:
# error: Argument 1 to "delete" of "DashboardWidgetsAPI" has incompatible type "str | None"; expected "str"
self._ws.dashboard_widgets.delete(widget.id)
after
assert widget.id is not None
self._ws.dashboard_widgets.delete(widget.id)
- Add
... | Noneif it's in the dataclass. Example:cloud: str = None->cloud: str | None = None
Add .as_posix() to convert Path to str
Add a valid default value for the dictionary return.
Example:
def viz_type(self) -> str:
return self.viz.get("type", None)after:
Example:
def viz_type(self) -> str:
return self.viz.get("type", "UNKNOWN")Integration tests must accompany all new code additions. Integration tests help us validate that various parts of our application work correctly when they interact with each other or external systems. This practice ensures that our software functions as a cohesive whole. Integration tests run every night and take approximately 15 minutes for the entire test suite to complete.
We encourage using predefined test infrastructure provided through environment variables for integration tests. These fixtures are set up in advance to simulate specific scenarios, making it easier to test different use cases. These predefined fixtures enhance test consistency and reliability and point to the real infrastructure used by integration testing. See Unified Authentication Documentation for the latest reference of environment variables related to authentication.
CLOUD_ENV: This environment variable specifies the cloud environment where Databricks is hosted. The values typically indicate the cloud provider being used, such as "aws" for Amazon Web Services and "azure" for Microsoft Azure.DATABRICKS_ACCOUNT_ID: This variable stores the unique identifier for your Databricks account.DATABRICKS_HOST: This variable contains the URL of your Databricks workspace. It is the web address you use to access your Databricks environment and typically looks like "https://dbc-....cloud.databricks.com."TEST_DEFAULT_CLUSTER_ID: This variable holds the identifier for the default cluster used in testing. The value resembles a unique cluster ID, like "0824-163015-tdtagl1h."TEST_DEFAULT_WAREHOUSE_DATASOURCE_ID: This environment variable stores the identifier for the default warehouse data source used in testing. The value is a unique identifier for the data source, such as "3c0fef12-ff6c-...".TEST_DEFAULT_WAREHOUSE_ID: This variable contains the identifier for the default warehouse used in testing. The value resembles a unique warehouse ID, like "49134b80d2...".TEST_INSTANCE_POOL_ID: This environment variable stores the identifier for the instance pool used in testing. You must utilise existing instance pools as much as possible for cluster startup time and cost reduction. The value is a unique instance pool ID, like "0824-113319-...".TEST_LEGACY_TABLE_ACL_CLUSTER_ID: This variable holds the identifier for the cluster used in testing legacy table access control. The value is a unique cluster ID, like "0824-161440-...".TEST_USER_ISOLATION_CLUSTER_ID: This environment variable contains the identifier for the cluster used in testing user isolation. The value is a unique cluster ID, like "0825-164947-...".
With the final results looking like (Note, for Azure Cloud, no DATABRICKS_TOKEN value):
export DATABRICKS_ACCOUNT_ID=999f9f99-89ba-4dd2-9999-a3301d0f21c0
export DATABRICKS_HOST=https://adb-8590162618999999.14.azuredatabricks.net
export DATABRICKS_CLUSTER_ID=8888-999999-6w1gh5v6
export TEST_DEFAULT_WAREHOUSE_ID=a73e69a521a9c91c
export TEST_DEFAULT_CLUSTER_ID=8888-999999-6w1gh5v6
export TEST_LEGACY_TABLE_ACL_CLUSTER_ID=8888-999999-9fafnezi
export TEST_INSTANCE_POOL_ID=7777-999999-save46-pool-cghrk21sThe test workspace must have test user accounts, matching displayName pattern test-uesr-*
To create test users:
databricks users create --active --display-name "test-user-1" --user-name "[email protected]"
databricks users create --active --display-name "test-user-2" --user-name "[email protected]"Before running integration tests on Azure Cloud, you must log in (and clear any TOKEN authentication):
az login
unset DATABRICKS_TOKENUse the following command to run the integration tests:
make integrationWe'd like to encourage you to leverage the extensive set of pytest fixtures. These fixtures follow a consistent naming pattern, starting with "make_". These functions can be called multiple times to create and clean up objects as needed for your tests. Reusing these fixtures helps maintain clean and consistent test setups across the codebase. In cases where your tests require unique fixture setups, keeping the wall clock time of fixture initialization under one second is crucial. Fast fixture initialization ensures that tests run quickly, reducing development cycle times and allowing for more immediate feedback during development.
from databricks.sdk.service.workspace import AclPermission
from databricks.labs.ucx.mixins.fixtures import * # noqa: F403
def test_secret_scope_acl(make_secret_scope, make_secret_scope_acl, make_group):
scope_name = make_secret_scope()
make_secret_scope_acl(scope=scope_name, principal=make_group().display_name, permission=AclPermission.WRITE)If the fixture requires no argument and special cleanup, you can simplify the fixture from
@pytest.fixture
def make_thing(...):
def inner():
...
return x
return innerto:
@pytest.fixture
def thing(...):
...
return xEach integration test must be debuggable within the free IntelliJ IDEA (Community Edition) with the Python plugin (Community Edition). If it works within IntelliJ CE, then it would work in PyCharm. Debugging capabilities are essential for troubleshooting and diagnosing issues during development. Please make sure that your test setup allows for easy debugging by following best practices.
Adhering to these guidelines ensures that our integration tests are robust, efficient, and easily maintainable. This, in turn, contributes to the overall reliability and quality of our software.
Currently, VSCode IDE is not supported, as it does not offer interactive debugging single integration tests. However, it's possible that this limitation may be addressed in the future.
You can add @retried decorator to deal with flaky tests:
from datetime import timedelta
from databricks.sdk.errors import NotFound
from databricks.sdk.retries import retried
@retried(on=[NotFound], timeout=timedelta(minutes=5))
def test_something(ws):
...This section provides a step-by-step guide to set up and start working on the project. These steps will help you set up your project environment and dependencies for efficient development.
To begin, install Hatch, which is our build tool.
On MacOSX, this is achieved using the following:
brew install hatchThe clone the github repo, and cd into it.
To begin, run make dev to create the default environment and install development dependencies.
make devTo use different python version, specify it in HATCH_PYTHON variable:
HATCH_PYTHON="$(which python3.10)" make clean dev testConfigure your IDE to use .venv/bin/python from the virtual environment when developing the project:
Verify installation with
make testBefore every commit, apply the consistent styleguide and formatting of the code, as we want our codebase to look consistent. Consult the Google Python Style Guide if you have any doubts. Make sure to run the tests.
make fmt testTo get an overview of the linter message codes, run the following command:
$ python tests/integration/source_code/message_codes.py
cannot-autofix-table-reference
catalog-api-in-shared-clusters
changed-result-format-in-uc
direct-filesystem-access
direct-filesystem-access-in-sql-query
default-format-changed-in-dbr8
dependency-not-found
jvm-access-in-shared-clusters
legacy-context-in-shared-clusters
not-supported
notebook-run-cannot-compute-value
python-udf-in-shared-clusters
rdd-in-shared-clusters
spark-logging-in-shared-clusters
sql-parse-error
sys-path-cannot-compute-value
table-migrated-to-uc
to-json-in-shared-clusters
unsupported-magic-lineHere are the example steps to submit your first contribution:
- Make a Fork from the repo
git clonegit checkout main(orgcmif you're using oh-my-zsh).git pull(orglif you're using oh-my-zsh).git checkout -b FEATURENAME(orgcb FEATURENAMEif you're using oh-my-zsh).- ... do the work
make fmt- ... fix if any
make test- ... fix if any
git commit -a. Make sure to enter a meaningful commit message title.git push origin FEATURENAME- Go to GitHub UI and create PR. Alternatively,
gh pr create(if you have GitHub CLI installed). Use a meaningful pull request title because it'll appear in the release notes. UseResolves #NUMBERin pull request description to automatically link it to an existing issue. - Announce PR for the review.
If you encounter any package dependency errors after git pull, run make clean
Sometimes, when dependencies are updated via dependabot for example, the environment may report the following error:
...
ERROR: Cannot install databricks-labs-ucx[test]==0.0.3 and databricks-sdk~=0.8.0 because these package versions have conflicting dependencies.
ERROR: ResolutionImpossible: for help visit https://pip.pypa.io/en/latest/topics/dependency-resolution/#dealing-with-dependency-conflictsThe easiest fix is to remove the environment and have the re-run recreate it:
$ hatch env show
Standalone
┏━━━━━━━━━━━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━┓
┃ Name ┃ Type ┃ Dependencies ┃ Scripts ┃
┡━━━━━━━━━━━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━┩
│ default │ virtual │ │ │
├─────────────┼─────────┼────────────────────────────────┼─────────────────┤
│ unit │ virtual │ databricks-labs-ucx[test] │ test │
│ │ │ delta-spark<3.0.0,>=2.4.0 │ test-cov-report │
│ │ │ pyspark<=3.5.0,>=3.4.0 │ │
├─────────────┼─────────┼────────────────────────────────┼─────────────────┤
│ integration │ virtual │ databricks-labs-ucx[dbconnect] │ test │
│ │ │ databricks-labs-ucx[test] │ │
│ │ │ delta-spark<3.0.0,>=2.4.0 │ │
├─────────────┼─────────┼────────────────────────────────┼─────────────────┤
│ lint │ virtual │ black>=23.1.0 │ fmt │
│ │ │ isort>=2.5.0 │ verify │
│ │ │ ruff>=0.0.243 │ │
└─────────────┴─────────┴────────────────────────────────┴─────────────────┘
$ make clean dev test
========================================================================================== test session starts ===========================================================================================
platform darwin -- Python 3.11.4, pytest-7.4.1, pluggy-1.3.0 -- /Users/lars.george/Library/Application Support/hatch/env/virtual/databricks-labs-ucx/H6b8Oom-/unit/bin/python
cachedir: .pytest_cache
rootdir: /Users/lars.george/projects/work/databricks/ucx
configfile: pyproject.toml
plugins: cov-4.1.0, mock-3.11.1
collected 103 items
tests/unit/test_config.py::test_initialization PASSED
tests/unit/test_config.py::test_reader PASSED
...
tests/unit/test_tables.py::test_uc_sql[table1-CREATE VIEW IF NOT EXISTS new_catalog.db.view AS SELECT * FROM table;] PASSED
tests/unit/test_tables.py::test_uc_sql[table2-CREATE TABLE IF NOT EXISTS new_catalog.db.external_table LIKE catalog.db.external_table COPY LOCATION;ALTER TABLE catalog.db.external_table SET TBLPROPERTIES ('upgraded_to' = 'new_catalog.db.external_table');] PASSED
---------- coverage: platform darwin, python 3.11.4-final-0 ----------
Coverage HTML written to dir htmlcov
========================================================================================== 103 passed in 12.61s ==========================================================================================
$Sometimes, when multiple Python versions are installed in the workstation used for UCX development, one might encounter the following:
$ make dev
ERROR: Environment `default` is incompatible: cannot locate Python: 3.10The easiest fix is to reinstall Python 3.10. If required remove the installed hatch package manager and reinstall it:
$ deactivate
$ brew install [email protected]
$ rm -rf venv
$ pip3 uninstall hatch
$ python3.10 -m pip install hatch
$ make dev
$ make testNote: Before performing a clean installation, deactivate the virtual environment and follow the commands given above.
Note: The initial hatch env show is just to list the environments managed by Hatch and is not needed.