bio-phys/bioen

BioEn - Bayesian Inference Of ENsembles

bioen's Issues

Path to BioEn analyze executable on Mac

On my laptop I had problems with DEER example Jupyter notebook as it did not find the BioEn executable. I installed the BioEn package with the --user flag.

bioen

-bash: bioen: command not found

The executable lives in /Users/lukas/.local/bin/bioen, which is was not on my path. A simple fix was simple to add this folder to the path. The question is whether there is a more general fix.

Creating `Observables` in a python session might crash the interpreter.

BioEn/bioen/analyze/observables/observables.py

Lines 265 to 268 in 1927780

	print('ERROR: The experimental data type {} is not implemented in BioEn yet. ' \
	'Please try \"generic\" (useful for distances, NOEs, CS, J-couplings, ' \
	'PREs etc.)'.format(experiment))
	sys.exit()

This is a hard exit of the python interpreter. When using BioEn as a library (e.g. in a notebook) it can crash the process instead of failing gracefully. To fail gracefully an exception should be raised.

To maintain well-formatted error messages in the CLI tool one can use try-except in the main function

try:
   obs = function_that_can_raise()
except Exception as e:
   print("Exception occurred")
   print("Error Message: ")
   print(e)
   if verbose:
       raise(e)  # to show stacktrace
   sys.exit(1)  # show tool exited with an error with return-code 1

Leaked file descriptor in Deer observables class

BioEn/bioen/analyze/observables/deer/deer.py

Lines 42 to 43 in 1927780

	pickle.dump([self.labels, self.moddepth, self.nrestraints, self.exp_tmp,
	self.exp_err_tmp, self.sim_tmp], open(self.out_pkl, 'wb'))

This line contains a leak. The call to open returns a file descriptor that is never closed again. This is OK when using only the bioen CLI tool. It will be a problem when using the library from python and analyzing a lot of deer data examples with pkl_out set because each process only is allowed a limited number of open file descriptors. The correct solution would be to use a context manager.