You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
fixed bugs in PBUSHT, PDAMPT, PELST, TABRND1, PTUBE, EIGC
imporved SET/SPOINT writing
improved messages when cross-referencing
readBDF is now gone, replaced by read_bdf (was previously deprecated)
large field & double precision format writing now supported
write_bdf method now has a is_double=True/False option
fixed bug with SETx writing. Nastran limits a SETx card to one THRU
per card, which is inconsistent with other cards that use THRU.
fixed GRID Position bug when using referenced cylindrical/spherical
coordinate systems
fixed CONM2 offset bug for coordinate systems
added CAERO1/2 sub-element access methods for use in GUI
additional xref options
improved error handling
OP2
rewrite of the OP2 allows for increased robustness and speed
Most fatal errors caught
BDF input errors not caught
added grid point weight table (see beam_modes demo)
OP2 no longer defines the filename; use read_op2.
readOP2 is now gone, replaced by read_op2 (was previously deprecated)
renamed OP2 class names to be more descriptive
(e.g. DisplacementObject -> RealDisplacement)
OP2 vectorized support. It's slower, but it takes up much less
memory (~5x). Note: the F06 writing cannot be vectorized, so if
you do vectorize your code, it may be much faster. The API and data
members are still in development and should be expected to change.
superelement support
F06
removed; use the OP2
GUI
added much improved qt GUI
results sidebar
controllable legend
vectorized support for reduced memory usage
color coded logging
scripting capability
removed wx GUI
Converters / Additional GUI Options
pyNastran's code base makes it easy to develop other useful tools
that make use of common code. As such, additional formats are supported
in terms of readers/writers/converters/viewing, but are not a main focus.
These include: