* Automatically enable workspace filesystem if the feature is disabled
([#42](#42)).

This PR onboards this repo for a new experimental github action

Retry for 15 seconds if both installation are not detected

[![lines of
code](https://tokei.rs/b1/github/databrickslabs/blueprint)]([https://codecov.io/github/databrickslabs/blueprint](https://github.com/databrickslabs/blueprint))

This release introduces several improvements and additions to the
databricks.labs.blueprint module. It includes an automated upgrade
framework with a new `upgrades.py` file, providing mechanisms for
managing and applying upgrades to the product. The `ProductInfo` class
now includes a `checkout_root` method, a `as_semver` property, a
`wheels` property, and a `_infer_version_file` class method for better
handling of versioning and wheel building processes. A new exception,
`SingleSourceVersionError`, has been added to indicate that the version
file is not found in the given project. Additionally, test code
organization has been improved with new directories and files for
fixtures and unit tests, along with new test cases and functions for the
upgrades functionality. The `test_wheels.py` file has also been updated
with new functions to check the version of the Databricks SDK and handle
cases where the version marker is missing or does not contain the
`__version__` variable. These changes enhance the functionality,
maintainability, and adaptability of the system.

)

In the rare circumstances when you cannot use
[@DataClass](#loading-dataclass-configuration) or you get `SerdeError`
that you cannot explain, you can implement `from_dict(cls, raw: dict) ->
'T'` and `as_dict(self) -> dict` methods on the class:

```python
from databricks.sdk import WorkspaceClient
from databricks.labs.blueprint.installation import Installation

class SomePolicy:
    def __init__(self, a, b):
        self._a = a
        self._b = b

    def as_dict(self) -> dict:
        return {"a": self._a, "b": self._b}

    @classmethod
    def from_dict(cls, raw: dict):
        return cls(raw.get("a"), raw.get("b"))

    def __eq__(self, o):
        assert isinstance(o, SomePolicy)
        return self._a == o._a and self._b == o._b

policy = SomePolicy(1, 2)
installation = Installation.current(WorkspaceClient(), "blueprint")
installation.save(policy, filename="backups/policy-123.json")
load = installation.load(SomePolicy, filename="backups/policy-123.json")

assert load == policy
```

* Added automated upgrade framework
([#50](#50)). This
update introduces an automated upgrade framework for managing and
applying upgrades to the product, with a new `upgrades.py` file that
includes a `ProductInfo` class having methods for version handling,
wheel building, and exception handling. The test code organization has
been improved, and new test cases, functions, and a directory structure
for fixtures and unit tests have been added for the upgrades
functionality. The `test_wheels.py` file now checks the version of the
Databricks SDK and handles cases where the version marker is missing or
does not contain the `__version__` variable. Additionally, a new
`Application State Migrations` section has been added to the README,
explaining the process of seamless upgrades from version X to version Z
through version Y, addressing the need for configuration or database
state migrations as the application evolves. Users can apply these
upgrades by following an idiomatic usage pattern involving several
classes and functions. Furthermore, improvements have been made to the
`_trim_leading_whitespace` function in the `commands.py` file of the
`databricks.labs.blueprint` module, ensuring accurate and consistent
removal of leading whitespace for each line in the command string,
leading to better overall functionality and maintainability.
* Added brute-forcing `SerdeError` with `as_dict()` and `from_dict()`
([#58](#58)). This
commit introduces a brute-forcing approach for handling `SerdeError`
using `as_dict()` and `from_dict()` methods in an open-source library.
The new `SomePolicy` class demonstrates the usage of these methods for
manual serialization and deserialization of custom classes. The
`as_dict()` method returns a dictionary representation of the class
instance, and the `from_dict()` method, decorated with `@classmethod`,
creates a new instance from the provided dictionary. Additionally, the
GitHub Actions workflow for acceptance tests has been updated to include
the `ready_for_review` event type, ensuring that tests run not only for
opened and synchronized pull requests but also when marked as "ready for
review." These changes provide developers with more control over the
deserialization process and facilitate debugging in cases where default
deserialization fails, but should be used judiciously to avoid brittle
code.
* Fixed nightly integration tests run as service principals
([#52](#52)). In this
release, we have enhanced the compatibility of our codebase with service
principals, particularly in the context of nightly integration tests.
The `Installation` class in the `databricks.labs.blueprint.installation`
module has been refactored, deprecating the `current` method and
introducing two new methods: `assume_global` and `assume_user_home`.
These methods enable users to install and manage `blueprint` as either a
global or user-specific installation. Additionally, the `existing`
method has been updated to work with the new `Installation` methods. In
the test suite, the `test_installation.py` file has been updated to
correctly detect global and user-specific installations when running as
a service principal. These changes improve the testability and
functionality of our software, ensuring seamless operation with service
principals during nightly integration tests.
* Made `test_existing_installations_are_detected` more resilient
([#51](#51)). In this
release, we have added a new test function
`test_existing_installations_are_detected` that checks if existing
installations are correctly detected and retries the test for up to 15
seconds if they are not. This improves the reliability of the test by
making it more resilient to potential intermittent failures. We have
also added an import from `databricks.sdk.retries` named `retried` which
is used to retry the test function in case of an `AssertionError`.
Additionally, the test function `test_existing` has been renamed to
`test_existing_installations_are_detected` and the `xfail` marker has
been removed. We have also renamed the test function `test_dataclass` to
`test_loading_dataclass_from_installation` for better clarity. This
change will help ensure that the library is correctly detecting existing
installations and improve the overall quality of the codebase.