problem with pickle

[PolyachenkoYA]

The problem: .names doubles after interacting with pickle, cPickle and dill.

How to reproduce (test.txt):

import gromacs.formats as gmx
import os
import pickle

def load_xvg(filepath, failsafe=2):
    xvg_file = gmx.XVG()
    xvg_file.read(filepath)
    return xvg_file

def test_fnc(xvg_file):
    print('names: ', xvg_file.names)
    print('shape: ', xvg_file.array.shape)
    print('names after: ', xvg_file.names)    
    print('shape after: ', xvg_file.array.shape)
    
xvg_filepath = 'test.txt'
pickle_filepath = 'test.pkl'

xvg_file = load_xvg(xvg_filepath)
print('names out: ', xvg_file.names)
print('shape out: ', xvg_file.array.shape)
test_fnc(xvg_file)

pickle.dump(xvg_file, open(pickle_filepath, 'wb'))
xvg_file = pickle.load(open(pickle_filepath, 'rb'))
print('\n2 out: ', xvg_file.names)
test_fnc(xvg_file)

os.remove(pickle_filepath)

gives

names out:  ['Temperature']
shape out:  (2, 2)
names:  ['Temperature']
shape:  (2, 2)
names after:  ['Temperature']
shape after:  (2, 2)

2 out:  ['Temperature']
names:  ['Temperature']
shape:  (2, 2)
names after:  ['Temperature', 'Temperature']
shape after:  (2, 2)

So, something wrong happens on the stage of either dumping or loading. It's also the case for cPickle and dill libs.

[orbeckst]

I think the following happens after unpickling

1. x.names shows the content that it had when pickled
2. x.shape triggers x.parse() (because by default savedata=False) so that the array is re-read from disk; on line

   GromacsWrapper/gromacs/fileformats/xvg.py

   Lines 496 to 498 in 76a80ad

   	if line.startswith("@ s") and "subtitle" not in line:
   	name = line.split("legend ")[-1].replace('"','').strip()
   	self.names.append(name)

   the column names are appended to x.names. Now the column name appears twice.

If you try with

def load_xvg(filepath, failsafe=2):
    xvg_file = gmx.XVG(savedata=True)
    xvg_file.read(filepath)
    return xvg_file

then does the same problem appear? I would think not because in this case, the array itself would be included in the pickle file (which might make it big) and the re-reading from disk won't be triggered.

[orbeckst]

@PolyachenkoYA I am sorry that I haven't replied to this issue sooner. Please feel free to ping me with @orbeckst when you open another issue or PR: create the issue, then add a comment with "can you please have a look @orbeckst " ... and then I get at least a notification. Otherwise it's too difficult to keep track of every open source project that I have code in. Thank you!! ❤️

[jandom]

There hasn't been any activity here – can we close this one out @orbeckst?

[orbeckst]

I don't really know if that's something that needs to be changed or just documented.