Problem with load_project() function

[DMaxJ]

I am running some simulations. To save time required for extracting the network every time I run a simulation I tried saving the project with
ws.save_project(proj)
which creates the desired .pnm file with the project name.

When I later try to open the same, I get some error. I could use the same functions in the earlier versions of openpnm. Now, I get error when I use
load_project('myproject')
the details of which I have provided below

ws.load_project('myproject')

Traceback (most recent call last):


 Input In [55] in <cell line: 1>
    ws.load_project('myproject')

  File c:\users\XYZ\documents\porenetwork\openpnm\utils\_workspace.py:223 in load_project
    proj = PNM.load_project(filename=filename)

  File c:\users\XYZ\documents\porenetwork\openpnm\io\_pnm.py:94 in load_project
    proj, obj = create_obj(root, name, proj)

  File c:\users\XYZ\documents\porenetwork\openpnm\io\_pnm.py:128 in create_obj
    obj.models.update(json.loads(root[name].attrs['models']))

  File c:\users\XYZ\documents\porenetwork\openpnm\core\_models.py:28 in update
    target = self._find_parent()

  File c:\users\XYZ\documents\porenetwork\openpnm\core\_models.py:41 in _find_parent
    raise Exception("No parent object found!")

Exception: No parent object found!

All my scripts and files are inside the same directory. There is no issue when I export files to VTK format. While saving an object, project, or workspace I get no error message. Could you please help solve this problem?

[jgostick]

The best work around for now, since we may not be able to look into this right away, is to save the dict coming out of the snow2 algorithm using python's pickle module. This dict is just a pure python object with no complicated dependencies and will always just work. Loading the dict into openpnm is not the slow part right?

[DMaxJ]

Thanks. I will check that. Loading the dict is definitely not the problem!!

[jgostick]

We have quite a few tests for the save/load functionality, so on the one hand I'm surprized that it's breaking, but on the other hand, it's actually pretty complicated under the hood.

[DMaxJ]

I actually tried going to the implemented functions under the hood, after after spending a couple of hours I gave up. I tried to read the file running the functions line by line (as implemented in the source code.
obj.models.update(json.loads(root[name].attrs['models']))
The saved .pnm file actually has 'models' but I could not figure out why the
_find_parent(self):
fails.

[jgostick]

FWIW, the "interconnectedness" of objects in OpenPNM is a HUGE pain, leading to problems like the on you're having...and we have decided to strip it ALL OUT in V3. This will delay the release of V3, but will be a major improvement in terms of maintainability and hopefully usability too.