🚀 DeepGraph Release Notes

This release introduces a new decorator, output_names, designed for use with user-defined connector and selector functions. If you've ever encountered an OSError: could not get source code using either create_edges or create_edges_ft, this release addresses that issue. The output_names decorator allows you to explicitly define the names of the output variables returned by these functions, avoiding the need to read source code.

The public API remains unchanged. This release introduces new functionality without affecting existing behavior.

Added

• deepgraph.output_names decorator: A new decorator to explicitly define the output variable names of connector or selector functions. This is necessary when the source code of the function is not available at runtime. The decorator attaches the specified output names to the decorated function via the _output_names attribute.

  Example:

  import deepgraph as dg
  
  @dg.output_names("dx", "dt", "v")
  def velocity(x_s, x_t, time_s, time_t):
      dx = x_t - x_s
      dt = time_t - time_s
      v = dx/dt
      return dx, dt, v

Fixed

• Fixed OSError: could not get source code by introducing the output_names decorator for explicitly defining output variable names in user-defined functions.

🔗 Links

• Documentation
• Installation Guide
• GitHub Repository

📝 Upgrade Notes

• For installing, refer to the installation instructions.

If you encounter any issues, please open an issue or join the discussion.

🚀 DeepGraph Release Notes

This release adds support for new, valid return formats of user-defined connector/selector functions (see create_edges). Specifically, explicit tuples and multi-line return statements are now possible. This enhancement increases flexibility and improves input validation.

The public API remains unchanged.

Added

• Introduced a robust AST-based implementation for extracting output variables from user-defined functions.

  • This improves reliability and parsing accuracy across supported Python versions (3.9+).

  • The extraction now strictly enforces that return statements contain only variable names.

  • It is now also possible to return output variables as:

    Example: explicit tuple

    def velocity(dt, x_s, x_t):
        dx = x_t - x_s
        v = dx / dt
        return (v, dx)

    Example: multi-line tuple

    def velocity(dt, x_s, x_t):
        dx = x_t - x_s
        v = dx / dt
        return (
            v,
            dx
        )

• Added CHANGELOG.md file.

Changed

• Replaced previous string-based return parsing using inspect.getsourcelines() with a more accurate approach using inspect.getsource() and the ast module.
• Output validation errors will now be raised if the return statement contains expressions, function calls, or anything other than variable names.
• Updated input argument parsing logic to use inspect.signature() instead of inspect.getfullargspec(), improving forward compatibility.

Removed

• The conda recipe to build a conda package locally. Conda packages are build by conda-forge using this recipe.

Fixed

• Resolved "UnclosedFileWarning: Closing remaining open file" warnings in tests involving hdf tables.
• Homogenized license information across project files
• pyproject.toml had wrong license information (BSD-2-Clause), changed it to the correct one (BSD-3-Clause).

🔗 Links

• Documentation
• Installation Guide
• GitHub Repository

📝 Upgrade Notes

• For installing with optional dependencies, refer to the installation instructions.

If you encounter any issues, please open an issue or join the discussion.

🚀 DeepGraph Release Notes

We’re excited to announce a new major release of DeepGraph! This update brings several improvements, bug fixes, and modernization efforts to the package, its documentation and distribution.

✨ What’s Changed

• Resolved pandas warnings and dtype issues
  Fixed compatibility warnings and a bug related to pandas data types for smoother user experience.

• Switched to uv for package management
  Adopted the uv tool for faster and more reliable package management.

• Modernized project structure
  Restructured the project and migrated metadata to pyproject.toml for PEP 621 compliance and improved maintainability.

• Updated GitHub Actions workflow
  The GitHub Workflow to build, test and publish both source distributions and wheels has been updated.

• Optional dependencies support
  Added a way to install optional dependencies. See the documentation for details.

• Documentation improvements
  Updated the README and enhanced overall documentation for clarity and completeness.

• ReadTheDocs build fix
  Fixed issues with documentation builds on ReadTheDocs.

• BREAKING CHANGE
  Removal of support for end-of-life Python versions (Python < 3.9).

• Support for NumPy 2.0

• Support for Python 3.13

🔗 Links

• Documentation
• Installation Guide
• GitHub Repository

📝 Upgrade Notes

• For installing with optional dependencies, refer to the installation instructions.

If you encounter any issues, please open an issue or join the discussion.

Happy graphing!

for changes, see https://github.com/deepgraph/deepgraph/commits/master

for changes, see https://github.com/deepgraph/deepgraph/commits/master.

for changes, see https://github.com/deepgraph/deepgraph/commits/master.

for changes, see https://github.com/deepgraph/deepgraph/commits/master.

Substantial internal speed up of DeepGraph.create_edges(_ft) by cythonizing bottlenecks.

For further changes, see: https://github.com/deepgraph/deepgraph/commits/master.

Substantial internal speed up of DeepGraph.create_edges().

For further changes, see: https://github.com/deepgraph/deepgraph/commits/master.

for changes, see https://github.com/deepgraph/deepgraph/commits/master.