It's been roughly a year and ~100 tickets closed since the last version, so it's probably time for another release! I want to thank everybody who helped with this release, especially Michael Redmond. He is working on h5Nastran, which which ties in with pyNastran. It's not quite ready yet, but it offers the possibility of major speedups for large models.

Probably the most best thing is the documentation actually builds again. It was too big and took too long, so readthedocs failed. I'm pleased to announce that the continuing problems of up-to-date documentation will hopefully be a thing of the past. M4 Engineering has offered to host the documentation on http://www.pynastran.m4-engineering.com.
There's still some work to do regarding hosting documentation for older versions and the master, but that's hopefully coming soon. Outside of that, it's the same open-source project and will still be on Github.

Regarding features, the focus has again been on robustness and testing. There has been a 10% increase in the testing coverage (the same as v0.8 to v1.0). There are a few changes (mainly in the BDF) though. The GUI now also supports PyQt4, PyQt5, and Pyside with the same API, so it's a bit easier to install from source as simplifying licensing issues as PyQt is GPL.

Programmatics

• Dropping Python 3.4 support
• dropping VTK 5/6 suppoprt for the GUI

BDF:

• 343 cards supported (up from 312)
• cross-referencing is now more straightforward to new users (much of v1.0 works using the _ref option)
  • *_ref attributes are cross-referenced
    • element.nodes is not cross-referenced
    • element.nodes_ref is cross-referenced
• pickling to reload your deck ~5x faster
• decreased time required for Case Control Deck with large SETs and many load cases
• improved optimization checks

OP2:

• HDF5 export/import support
• pandas support for matrices
• couple more results vectorized (e.g., complex strain energy, DMIG strain energy, some forces)
• grid_point_stressses supported (disabled since v0.7)
• fixed sparse matrices being stored as dense matrices

GUI:

• preliminary support for PySide
• can now mix and match fringe/displacement/vector results (e.g., max principal stress shown on a displaced model)
• improved animation menu
  • in gui animation
  • more animation profiles
• bar profile visualzation
• nominal geometry (useful for deflection plots)
• improved optimization support
• improved picking display
• better PSHELL/PCOMP distinction

Known issues:

• Transient Pandas Dataframes will fail for newer versions of numpy/pandas. If anyone knows how to use a MultiIndex, this is probably pretty easy to fix.