Version	Docs	Status
	docs
Master

Download GUI (latest is from 2019/5/24)

Also, check out the:

• Discussion forum (intended for questions about the latest release)
• Developer forum (intended for questions about the master branch)
• Docs

for more detailed information.

pyNastran is an interface library to the various Nastran file formats (BDF, OP2, OP4). Using the BDF interface, you can read/edit/write Nastran geometry without worrying about field formatting. Many checks are also performed to verify that your model is correct. Using the OP2 interface, you can read large result files quckly and efficiently. Additionally, you can also extract a subset of the result data and write OP2/F06 result files. For a more detailed list of features, see:

• Features

Using the pyNastran GUI, you can read in Nastran models and quickly view results for a model. While it's no FEMAP/Patran, it can replace many tasks that would otherwise require a commercial program.

I keep saying to myself there's not much to add, but Nastran is huge. Beyond HDF5 support in the BDF, I'm a huge fan of the new ability to keep track of which include file a card came from and write it as a separate file. It's limited in usefulness, but very handy in certain cases. There's a new (still preliminary) superelement capability. I'm far from a superelement expert, but it's probably useful.

The OP2 reader now supports SORT2 along with much improved random results reading. If you're using 60+ GB OP2s, you probably have had issues with RAM usage in the past. With the new ability to dump the OP2 directly to HDF5, this should not be an as much of an issue. It's not 100% implemented, so let me know if you need it for another result.

Regarding the GUI, there are also some new features. Groups work a bit better, but aren't quite perfect. Logging has been dramatically sped up so the GUI loads faster and you can load Nastran models even faster if you disable additional results (e.g., element quality).

Finally, Python 2.7 is end of life. Numpy, scipy, and matplotlib have all dropped Python 2.7 support. It's time for pyNastran to as well. The OP2 reader is 30% faster in Python 3.6+ than Python 2.7, so it's not all bad!

Programmatics:

• This is the last Python 2.7 release.
• Dropping Python 3.5 support
• Supports Python 2.7, 3.6-3.7
• GUI is compatible with PyQt4/PyQt5 as well as PySide/PySide2
• improved testing of old versions of packages

BDF:

• 373 cards supported (up from 343)

• added abiltity to write models to separate include files

  >>> model = BDF()
  >>> model.read_bdf(bdf_filename, save_file_structure=True)
  
  out_filenames = {
      bdf_filename : bdf_filename_new,
      include_filename : include_filename_new,
  }
  >>> model.write_bdfs(out_filenames, relative_dirname=None, is_windows=None)
  >>> ifile = model.grids[1].ifile

• HDF5 import/export

  >>> model = read_bdf(bdf_filename)
  >>> model.export_hdf5_filename(hdf5_filename)
  >>> model_new = OP2()
  >>> model_new.load_hdf5_filename(hdf5_filename)

• preliminary superelement support

  >>> model.read_bdf(bdf_filename)
  >>> model.superelement_models[1].nodes

OP2:

• reorganization of random op2 results into op2.results.psd (or ato, no, crm, rms) to aide in finding data

• reorganization of op2 class to reduce number of functions in the object. This affects any custom table reading.

• improved optimzation response reading

• limited SORT2 support

• fixed CD transformation bug for BOUGV1 and BOPHIG1 tables

• Improved HDF5 export/import support (e.g., matrices, random results)

• Can optionally save directly to HDF5 instead of numpy (limited).

• Loading OP2s to an HDF5 file to decrease memory usage

  >>> op2_model = OP2()
  >>> op2_model.load_as_h5 = True
  >>> op2_model.read_op2(op2_filename)

OP2Geom:

• HDF5 support
• reading EQEXIN/S, GPT, GPDT, CSTM/S tables (recovery of nodes & coordinate with OP2Geom)
• fixed theta/mcid reading for CTRIA3/CQUAD4
• fixed CQUAD8 bug

GUI:

• sped up HTML logging
• much improved groups menu
• options for Nastran in preferences menu to speed up loading/limit memory usage
• pyNastran BDF pickle reading
• pyNastran OP2 HDF5 reading (not MSC's format)
• visualization when pickling nodes/elements
• min/max labels
• highlight menu
• Patran-style colors
• custom force vectors
• AVL support

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.

This result has been superseeded by 1.2.1. See release notes for details.

Version	Docs	Status
	docs

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
• Supports Python 2.7, 3.5-3.6
• 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

  >>> op2_model = read_op2(op2_filename)
  >>> op2_model.export_hdf5_filename(hdf5_filename)
  >>> op2_model_new = OP2()
  >>> op2_model_new.load_hdf5_filename(hdf5_filename, combine=True)

• 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.

See v1.0.0 for information regarding enhancements.

Version	Docs
v0.8.0

See v0.8.0 for information regarding enhancements.

See v0.7.2 for information regarding enhancements.

Version 0.6 improves BDF reading. The reader is more robust and also requires proper BDF field formatting (e.g. a integer field can't be a float). Additionally, cards also have a comment() method.

Marcin Gąsiorek participated in the latest pyNastran under the European Space Agency's (ESA) "Summer of Code In Space" SOCIS program. The program provides a stipend to students to work on open-source projects. He did a great job of simplifying code and creating nicer documentation.