Can't load restart - pickling problems?

[rwest]

I am getting this when I try to restart the methylformate example:

  File "/nics/d/home/rwest/Code/RMG-Py/rmgpy/rmg/main.py", line 291, in initialize
    self.loadRestartFile(os.path.join(self.outputDirectory,'restart.pkl'))
  File "/nics/d/home/rwest/Code/RMG-Py/rmgpy/rmg/main.py", line 589, in loadRestartFile
    self.reactionModel = cPickle.load(f)
  File "quantity.py", line 275, in rmgpy.quantity.ArrayQuantity.__init__ (build/pyrex/rmgpy/quantity.c:4503)
rmgpy.quantity.QuantityError: (QuantityError('Dimension 0 has 7 elements for the given value, but 3 elements for the given uncertainty.',), <type 'rmgpy.quantity.ArrayQuantity'>, (array([  750.        ,   770.        ,  3400.        ,  2100.        ,
         647.28325342,   647.28380352,  3592.16637415]), 'cm^-1', array([ 0.,  0.,  0.]), '+|-'))

I'm not sure what the offending object is, but it's something with units of cm^-1. I'm guessing vibrational frequencies (first four integer ones from Franklin's groups, last 3 floats from fitting to the Cp). Not sure why it only has 3 uncertainties (that are all zero!).

Do we have a unit test on pickling and unpickling of frequencies?

[rwest]

I think this particular problem is addressed by d8b38a0
(When the extra frequencies were being estimated, the frequencies' value_si array was extended but its uncertainty array was not).
But so far nothing has been done to check for other errors of this type (eg. a check on set for value or value_si)

[rwest]

The above commit does indeed solve the loading from restart problem, but now immediately after loading we fail to save the HTML output:

Loading previous restart file...
Saving current model to HTML file...

Stack trace:

  File "/nics/d/home/rwest/Code/RMG-Py/rmgpy/rmg/main.py", line 361, in execute
    self.saveEverything()
  File "/nics/d/home/rwest/Code/RMG-Py/rmgpy/rmg/main.py", line 517, in saveEverything
    self.saveOutputHTML()
  File "/nics/d/home/rwest/Code/RMG-Py/rmgpy/rmg/main.py", line 658, in saveOutputHTML
    saveOutputHTML(os.path.join(self.outputDirectory, 'output.html'), self.reactionModel)
  File "/nics/d/home/rwest/Code/RMG-Py/rmgpy/rmg/output.py", line 113, in saveOutputHTML
    family = rxn.getSource().label
AttributeError: 'rmgpy.reaction.Reaction' object has no attribute 'getSource'

Seems we can't save reactions that came from a restart? Is the source being pickled correctly?

[jwallen]

The source depends on the type of reaction being saved. I think in this case it's rmgpy.rmg.pdep.PDepReaction that is the issue; it is missing a specialized __reduce__() method, so it's using the one from the base class, which doesn't know how to save the source.

[rwest]

Thanks. Fixed in 60ed251.

..although a unit test would be nice.