v0.4.4

What's Changed

• Fix typing issues with the DataCatalog. by @tobiasraabe in #510
• [automated] Update plugin list by @github-actions in #511
• Improve the documentation. by @tobiasraabe in #509

Full Changelog: v0.4.3...v0.4.4

v0.4.3

What's Changed

This release contains a lot of smaller improvements and bug fixes. Here is a short list.

• #484 raises an error message when a PathNode was used with a directory instead a file.
• #496 makes pytask even lazier. When a preceding task is executed and produces the same outputs, the following task will no longer be executed.
• Objects in task modules that overwrite __getattr__ should not cause any problems anymore (#507 was fixed in #508). Same applies to import Task in task modules.

Complete list of changes

• Simplify the teardown of a task. by @tobiasraabe in #483
• Correctly unconfigure pytask. by @tobiasraabe in #485
• Raise informative error when path nodes point to directories. by @tobiasraabe in #484
• Add default names to PPathNodes. by @tobiasraabe in #486
• Modernize TopologicalSorter. by @tobiasraabe in #458
• Raise error for invalid value in return annotation. by @tobiasraabe in #488
• Refactor and better test products. by @tobiasraabe in #489
• Refactor and better test parsing of dependencies. by @tobiasraabe in #490
• Addition to #489. by @tobiasraabe in #491
• Make pytask even lazier. by @tobiasraabe in #496
• Bump sigstore/gh-action-sigstore-python from 1.2.3 to 2.1.0 by @dependabot in #495
• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci in #494
• Remove unnecessary code from the collection of tasks. by @tobiasraabe in #497
• Fix errors when using Task and TaskWithoutPath in task modules. by @tobiasraabe in #498
• Allow tasks to depend on other tasks. by @tobiasraabe in #493
• Move test dependencies to pyproject.toml by @tobiasraabe in #500
• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci in #499
• Remove MetaNode. by @tobiasraabe in #501
• [automated] Update plugin list by @github-actions in #505
• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci in #504
• Catch objects pretending to be PTask. by @tobiasraabe in #508

Full Changelog: v0.4.2...v0.4.3

v0.4.2

Highlights

This release contains a new feature and some improvements for users.

• 🚀 The new feature is the pytask.DataCatalog that allows users to manage dependencies and products in projects more easily. Read the tutorial to get started. 🚀
• File changes are now detected by hashes instead of modification timestamps. It should prevent accidental executions when working with cloud storage providers like Dropbox or OneDrive and in many other situations. To save runtime, pytask uses a cache for the hashes when the modification timestamp has not changed.
• Nodes now have signatures that separate how nodes are named and displayed from how nodes are identified internally. If you have written a custom node, please update it according to the how-to guide.
• All of pytask's internal files are now stored in a .pytask folder in your project. The file .pytask.sqlite3 is moved to this location as well. Add .pytask to your .gitignore to prevent accidentally committing the folder.

What's Changed

• Simplify building the plugin manager. by @tobiasraabe in #449
• Rename graph.py to dag_command.py and improve collect_command.py. by @tobiasraabe in #451
• Remove more .svgs and replace them with animations. by @tobiasraabe in #454
• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci in #452
• [automated] Update plugin list by @github-actions in #453
• Add more explanation when PNode.load() fails during execution. by @tobiasraabe in #455
• Refer to source code on Github in API docs. by @tobiasraabe in #456
• Refactor code for format_node_name. by @tobiasraabe in #457
• Add hook to sort __all__. by @tobiasraabe in #459
• Simplify removing internal tracebacks from exceptions with cause. by @tobiasraabe in #460
• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci in #461
• Fix import error for pluggy<1.3. by @tobiasraabe in #462
• Raise error when function is defined outside the loop body. by @tobiasraabe in #463
• Improve pins. by @tobiasraabe in #464
• Test that internal tracebacks are removed by reports. by @tobiasraabe in #465
• Add is_product to PNode.load(). by @tobiasraabe in #472
• Add a data catalog. by @tobiasraabe in #419
• Hash files instead of relying on modification timestamps. by @tobiasraabe in #469
• Move .pytask.sqlite3 to .pytask/. by @tobiasraabe in #470
• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci in #471
• Update PyPI action. by @tobiasraabe in #477
• Add node signatures. by @tobiasraabe in #473
• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci in #476
• Add snapshot tests. by @tobiasraabe in #475
• Switch from black to ruff-format. by @tobiasraabe in #478
• Rework reports and tracebacks. by @tobiasraabe in #474
• Give skips higher precendence than ancestor failed as outcome. by @tobiasraabe in #479
• Remove checks for missing root nodes. by @tobiasraabe in #480
• Improve coverage. by @tobiasraabe in #481
• Fix handling of names and signatures of PythonNodes. by @tobiasraabe in #482

Full Changelog: v0.4.1...v0.4.2

v0.4.1

What's Changed

Of course, it's a mandatory bug fix release after a bigger release.

Using the product annotation, Annotated[..., Product] did not work with multiple products.

• Fix setting the name of PythonNode. by @tobiasraabe in #443
• Move content of setup.cfg to pyproject.toml. by @tobiasraabe in #444
• [automated] Update plugin list by @github-actions in #445
• Fix when multiple product annotations are used. by @tobiasraabe in #448
• Fix PythonNode when used as return. by @tobiasraabe in #446
• Simplify the tree_map code for generating the DAG. by @tobiasraabe in #447

Full Changelog: v0.4.0...v0.4.1

v0.4.0

News

pytask became three years old in July, which is a suitable event to rethink pytask's design and blow dust off of some of its oldest components.

Here are the highlights of v0.4.0 🚀 ⭐

Highlights

New interfaces for products.

Every argument can be declared as a product with the new' Product' annotation. The path can be passed as a default value.

from pathlib import Path

from pytask import Product
from typing_extensions import Annotated


def task_hello_earth(path: Annotated[Path, Product] = Path("hello_earth.txt")):
    path.write_text("Hello, earth!")

More explanation can be found at https://tinyurl.com/yrezszr4.

It is also possible to use the return of the task function as a product, which allows wrapping any third-party function as a task function. Read more about it here: https://tinyurl.com/pytask-return.

from pathlib import Path

from pytask import Product
from typing_extensions import Annotated


def task_hello_earth() -> Annotated[str, Path("hello_earth.txt")]:
    return "Hello, earth!"

Every task argument is a dependency

In older pytask versions, only paths were treated as task dependencies. That meant when you passed other arguments to the task, and they changed, it did not trigger a rerun of the task.

Now, every argument to a task can be a dependency, and you can hash them if they should trigger a rerun. It is explained in https://tinyurl.com/pytask-hash.

from pathlib import Path
from typing import Annotated

from pytask import Product
from pytask import PythonNode


def task_example(
    text: Annotated[str, PythonNode(value="Hello, World", hash=True)],
    path: Annotated[Path, Product] = Path("file.txt"),
) -> None:
    path.write_text(text)

A new functional interface

The functional interface for pytask has been reworked and accepts a list of task functions. You can use it within your terminal or a Jupyter notebook. Read this guide to learn more about it: https://tinyurl.com/pytask-functional.

from pathlib import Path
from typing import Annotated

from pytask import build


def create_text() -> Annotated[str, Path("hello_earth.txt")]:
    return "Hello, earth!"


session = build(tasks=[create_text])

Custom Nodes through Protocols

In the newest version, nodes (dependencies and products) and tasks follow protocols. It allows for customizations like PickleNodes that store any Python object as a pickle file and inject the object into the task when used as a dependency. It is explained in more detail in this guide: https://tinyurl.com/pytask-custom-nodes.

Other notable changes

• Python 3.12 is supported, and support for Python 3.7 is dropped.
• @pytask.mark.depends_on and @pytask.mark.produces are deprecated. There are better options to define dependencies and products explained in https://tinyurl.com/yrezszr4.
• @pytask.mark.task is also deprecated and replaced by from pytask import task and @task.

What's Changed

• Remove Python 3.7 support and add a new action for mamba. by @tobiasraabe in #323
• Replace pony with sqlalchemy>=1.4.36. by @tobiasraabe in #387
• Remove @pytask.mark.parametrize. by @tobiasraabe in #391
• Parse dependencies from all args if depends_on is not used. by @tobiasraabe in #384
• Add products with typing.Annotation. by @tobiasraabe in #394
• Refactor pybaum to _pytask.tree_util. by @tobiasraabe in #395
• Replace pybaum with optree and add paths to PythonNode names. by @tobiasraabe in #396
• Add support for NamedTuple and attrs classes in @pytask.mark.task(kwargs=...). by @tobiasraabe in #397
• Deprecate decorators for depends_on and produces. by @tobiasraabe in #398
• Use protocols instead of ABCs. by @tobiasraabe in #402
• Allow tasks to return products. by @tobiasraabe in #404
• Tracking changes in v0.4.0. by @tobiasraabe in #400
• Bump peter-evans/create-pull-request from 5.0.1 to 5.0.2 by @dependabot in #390
• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci in #388
• Allow to use prefix trees as nodes to parse function returns. by @tobiasraabe in #406
• Remove .value from Node protocol. by @tobiasraabe in #408
• Make .from_annot an optional feature of nodes. by @tobiasraabe in #409
• Allow to pass functions to PythonNode(hash=...). by @tobiasraabe in #410
• Add protocols for tasks. by @tobiasraabe in #412
• Remove scripts to generate .svgs. by @tobiasraabe in #413
• Allow more ruff rules. by @tobiasraabe in #414
• A new functional interface. by @tobiasraabe in #411
• Deprecate @pytask.mark.task in favor of @pytask.task. by @tobiasraabe in #417
• Simplify and fix code in dag.py. by @tobiasraabe in #418
• Convert DeprecationWarning to FutureWarning for deprecated decorators. by @tobiasraabe in #420
• Remove deprecation warning for produces. by @tobiasraabe in #421
• Document new interface. by @tobiasraabe in #392
• Fix import_path. by @tobiasraabe in #424
• Publish pytask.tree_util. by @tobiasraabe in #426
• Fix type annotations of task.depends_on and task.produces. by @tobiasraabe in #427
• Document functional interface. by @tobiasraabe in #423
• Update example in README.md. by @tobiasraabe in #428
• Add better error message when node.state() throws error during DAG validation. by @tobiasraabe in #429
• Update parts of the documentation. by @tobiasraabe in #430
• Enable colors in WSL. by @tobiasraabe in #431
• Fix type checking for pytask.mark.x. by @tobiasraabe in #432
• Fix ids of PythonNodes. by @tobiasraabe in #433
• Add support for Python 3.12. by @tobiasraabe in #434
• Fix detection of task functions. by @tobiasraabe in #437
• Clarify some types. by @tobiasraabe in #438
• Refine typing. by @tobiasraabe in #440

Full Changelog: v0.3.2...v0.4.0

v0.4.0rc4

The last pre-release.

v0.4.0rc3

A couple of new fixes. Most notably a fix for the ids of PythonNodes that should prevent rebuilds.

v0.4.0rc2

Another release candidate that fixes the installation via conda and adds full support for pytask-parallel.

v0.4.0rc1

This is the first release candidate for the v0.4.* release series.

The final release still requires some changes. For example, the documentation needs to be extended. But, the essential parts are already there, and it is time to collect some final feedback! Let me know what you think and what needs to be improved. You can comment in the discussion for this release #422.

To install the pre-release, use

$ pip install pytask --pre
$ conda install -c "conda-forge/label/pytask_rc" pytask

Now, let's take a look at the changes.

What's Changed

New

• Dependencies and products of tasks have new interfaces that are explained in this tutorial.
• You can also now declare products by allowing task functions to return. Follow this guide.
• If you have inputs to task functions that should be hashed to detect any changes, follow this guide.
• Before, only pathlib.Paths received special treatment as dependencies or products to task functions. Now, it is possible to define your own nodes that simplify, for example, loading pickle files as this guide explains. But many more extensions are possible, like defining data in an S3 bucket as a dependency or product.
• The functional interface has been reworked and now accepts tasks directly, allowing you to execute pytask on the command line or in Jupyter notebooks. The documentation must still be written, but here is your starting point.

Removals

• Python 3.7 is no longer supported.
• @pytask.mark.parametrize is removed. Follow this tutorial instead.

Deprecations

• @pytask.mark.depends_on, @pytask.mark.produces are deprecated and will be removed in v0.5.0.
• @pytask.mark.task is deprecated. Use @pytask.task instead.
• Paths defined as strings are deprecated and should be replaced with proper pathlib.Path objects.

Full list of changes

• Remove Python 3.7 support and add a new action for mamba. by @tobiasraabe in #323
• Replace pony with sqlalchemy>=1.4.36. by @tobiasraabe in #387
• Remove @pytask.mark.parametrize. by @tobiasraabe in #391
• Parse dependencies from all args if depends_on is not used. by @tobiasraabe in #384
• Add products with typing.Annotation. by @tobiasraabe in #394
• Refactor pybaum to _pytask.tree_util. by @tobiasraabe in #395
• Replace pybaum with optree and add paths to PythonNode names. by @tobiasraabe in #396
• Add support for NamedTuple and attrs classes in @pytask.mark.task(kwargs=...). by @tobiasraabe in #397
• Deprecate decorators for depends_on and produces. by @tobiasraabe in #398
• Use protocols instead of ABCs. by @tobiasraabe in #402
• Allow tasks to return products. by @tobiasraabe in #404
• Tracking changes in v0.4.0. by @tobiasraabe in #400
• Bump peter-evans/create-pull-request from 5.0.1 to 5.0.2 by @dependabot in #390
• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci in #388
• Allow to use prefix trees as nodes to parse function returns. by @tobiasraabe in #406
• Remove .value from Node protocol. by @tobiasraabe in #408
• Make .from_annot an optional feature of nodes. by @tobiasraabe in #409
• Allow to pass functions to PythonNode(hash=...). by @tobiasraabe in #410
• Add protocols for tasks. by @tobiasraabe in #412
• Remove scripts to generate .svgs. by @tobiasraabe in #413
• Allow more ruff rules. by @tobiasraabe in #414
• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci in #407
• A new functional interface. by @tobiasraabe in #411
• Deprecate @pytask.mark.task in favor of @pytask.task. by @tobiasraabe in #417
• Simplify and fix code in dag.py. by @tobiasraabe in #418
• Convert DeprecationWarning to FutureWarning for deprecated decorators. by @tobiasraabe in #420
• Remove deprecation warning for produces. by @tobiasraabe in #421
• Document new interface. by @tobiasraabe in #392
• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci in #415

Full Changelog: v0.3.2...v0.4.0rc1

v0.3.2

Highlights

This release contains the following highlights:

• Previously, if you accidentally hit the save button on an unchanged task file, the task would be rerun by pytask, although nothing had changed. Now, pytask wouldn't rerun the task because it also compares the hashes of task files, not only the modification timestamp.
• If you want to enforce rerunning tasks, there is now a --force flag. Take the function name/id of the task and run pytask -k <task id> --force, and the task + its necessary tasks will be executed. Or delete a product from the task you want to rerun.
• The import mechanism for task modules has been reworked, and errors resolved. Thanks to @NickCrews!

Additionally, the @pytask.mark.parametrize decorator is deprecated and will be removed in pytask v0.4. If you use the decorator, you will have two options:

1. (Recommended) Upgrade your code to the new approach for repeating tasks described in this tutorial.
2. Or, pin pytask to pytask<0.4 and silence the deprecation warning by setting silence_parametrize_deprecation = true in your pyproject.toml under [tool.pytask.ini_options].

What's Changed

• Update version numbers in animations. by @tobiasraabe in #345
• Add dependabot for GitHub actions. by @tobiasraabe in #348
• Publish db. by @tobiasraabe in #352
• Refactor nodes. by @tobiasraabe in #355
• Add -f/--force to force executing tasks. by @tobiasraabe in #354
• Add hashing to avoid re-executing tasks when modification times changed. by @tobiasraabe in #357
• Update update_plugin_list.py. by @tobiasraabe in #364
• Rework panel with sphinx-design. by @tobiasraabe in #365
• Add light and dark logos for the documentation. by @tobiasraabe in #366
• Fix the panel on the index page of the documentation. by @tobiasraabe in #367
• Fix error introduced in #364. by @tobiasraabe in #369
• Revert change turning Node.state() into a hook. by @tobiasraabe in #370
• Rename Node back to MetaNode. by @tobiasraabe in #371
• Clearer documentation for pytask dag -o. by @tobiasraabe in #376
• Conditionally skip tests on MacOS. by @tobiasraabe in #378
• Deprecate @pytask.mark.parametrize. by @tobiasraabe in #381
• Fix the import mechanism for task modules. by @NickCrews in #373
• Update changes. by @tobiasraabe in #383

New Contributors

• @dependabot made their first contribution in #349
• @NickCrews made their first contribution in #373

Full Changelog: v0.3.1...v0.3.2