v1.8

This release introduces initial development verson for 3D filament identification for PPV (FilFinderPPV) and PPP (FilFinderPPP). There are additionally various updates and fixes to the 2D code and infrastructure.

What's Changed

• Fix an un-declared variable issue by @keflavich in #75
• syntax tweaks by @jdhenshaw in #81
• refactor: add an optional argument to disable the display of open fig… by @andizq in #83
• Parallelize and checkpoint by @e-koch in #86
• FilFinder3d by @samdf96 in #57
• Parallel checks and debugging for empty filament cases by @e-koch in #88
• Various infrastructure and tutorial updates by @e-koch in #91
• Bump actions/setup-python from 4 to 5 in /.github/workflows in the actions group by @dependabot in #92

New Contributors

• @andizq made their first contribution in #83
• @samdf96 made their first contribution in #57
• @dependabot made their first contribution in #92

Full Changelog: v1.7.2...v1.8

v1.7.2

Primarily added infrastructure improvements:

• adding testing support for python 3.7 to 3.9
• dropping support for python 3.6
• adding an action to build the distribution and wheels, and uploads to pypi

Fix for networkx >v2

As noted in #54, the v1.6 release broke with new changes to networkx. This release fixes that issue.

Major code re-write and fixes for critical bugs

Implements the new FilFinder2D and Filament2D classes for running the algorithm. The old fil_finder_2D is now deprecated. Also improved and expanded the suite of tests.

To see a full list of changes, see #41.

Support for networkx 2.0

Version 2 of networkx introduced some breaking changes. These are fixed in the new release and all version of networkx should work now.

Other changes:

• fixed an issue in reading in the angular pixel scale from a header when the PC keywords were set.
• fixed a small bug in the failure criteria for Gaussian profile fits

Fixes for numpy v1.12

Provides fixes for numpy 1.12, which requires integers for slicing arrays.

Fix for changes in skimage 0.12

threshold_adaptive in skimage was slightly altered, causing issues for FilFinder.

Other notable changes:

• radial_profile was updated to be more robust for removing variations due to overlap with different objects (auto_cut)
• Tools for finding the width along the main filament branch (profiles across each skeleton pixel) were introduced in fil_finder.width_profiles.

Update to work with Numpy 1.10

The radial profiling did not work with the newest numpy release. This patches that problem.

Clean-up, more output flexibility

Biggest updates are to the documentation and tutorial.

The keyword save_png has been added to save png plots and allow interactive plotting to be avoided for large maps.

The output tables have been fixed for individual skeleton branches. HDF5 tables are now supported and are encouraged for the branches (since the different number of branches in each skeleton makes it difficult to include in a single table).

Mostly minor bug fixes

There were some small mistakes treating the non-parametric width method properly, as well as the beamwidth conversion to physical units.

This release marks the version actually used in the accompanying paper.