Add maintenance function to update TYPE in HDF5 files #1243
Conversation
Pull out the code that extracts class name from TYPE field
pmrv
added
integration
PyironTable save analysis functions as pickled objects. Afterwards we moved them to different modules so unpickling breaks. This method now imports all the updated modules, reloads and rewrites the tables back, so that on the next load they can be unpickled from the new locations.
|
Reminder to myself: forgot to add new test file. |
There was a problem hiding this comment.
Nice. I have some questions/suggestions, but they're "accept with minor revision not requiring re-review" 😜 I also assume you'll get the integration tests passing, but I see an open PR over on pyiron_atomistics on this topic, so I am confident you're already on top of this 👍
One thing on my wishlist is that we actually have an old project that we test the updates on? Maybe we have this and I'm just not aware? If such a thing is indeed missing, I can accept that it's outside scope for this PR.
pyiron_base/project/maintenance.py
Outdated
| def add_module_conversion(old: str, new: str): | ||
| """ | ||
| Add a new module conversion. | ||
| After this call any object that was previously defined in module `old` and is now defined in `new`, will be |
There was a problem hiding this comment.
This is not strictly true, or? It looks to me like you need to invoke add_module_conversion then still do a maintenance cycle.
There was a problem hiding this comment.
Correct, I had just copied this over from #1234, my bad.
| job.run() | ||
|
|
||
| # manipulate type to fake an old module | ||
| old_path_job = "job.module" |
There was a problem hiding this comment.
Beautiful. Even my stretch goal for this PR is satisfied 😂 🚀
|
From my side this is done and I'll merge it tonight and bump the version if there's no other comments. @jan-janssen Is it realistic for you update the PR on contrib until Monday? |
|
Compatibility tests are failing because of some mamba issues. |
Problem as discussed in #1234.
This adds
pyiron_base.project.maintenance.add_module_conversionthat takes to module paths and saves them for updates. Upon calling the corresponding maintenance function HDF5 files (jobs and project.data) are scanned and the TYPE field is updated, substituting the old for the new module path. When an object is loaded and the import fails while its module path is known to have been updated a readable error message is raised directing users to use the maintenance function.In the future whenever classes are moved between packages and modules a corresponding call to
add_module_conversionhas to be inserted into the__init__of the package where those classes were previously defined.