The first is a fix for a minor problem that I have ancountered in rerun_fw in combination with the recover_launch option. The recover_launch dictionary is coming from the checkpoint keyword in the Launch history. This can contain the serialized version of an MSONable object. This gets deserialized while loading, but when doing the
it is not serialized and the update of a dictionary containing this object fails. I added a recursive_dict that solves the problem and should be harmless for all the other cases.
I just have the doubt if this should instead be done inside get_recovery. The API of that method is not really explicit about what should be its output.
The second point is the addition of a -wf option for the set_priority command in lpad. I sometimes want to set a priority for a whole workflow and this seemed a functionality that could be useful for the generic user.
Coverage decreased (-0.08%) to 60.676% when pulling 6508bc66a4204892511a505105b9fb7acc688376 on gpetretto:bugfix into 34fba137c5d384c5239d5d8cf53d182441f923bd on materialsproject:master.
This PR contains two modifications.
The first is a fix for a minor problem that I have ancountered in
rerun_fw
in combination with therecover_launch
option. Therecover_launch
dictionary is coming from thecheckpoint
keyword in the Launch history. This can contain the serialized version of an MSONable object. This gets deserialized while loading, but when doing theit is not serialized and the update of a dictionary containing this object fails. I added a
recursive_dict
that solves the problem and should be harmless for all the other cases.I just have the doubt if this should instead be done inside
get_recovery
. The API of that method is not really explicit about what should be its output.The second point is the addition of a
-wf
option for theset_priority
command inlpad
. I sometimes want to set a priority for a whole workflow and this seemed a functionality that could be useful for the generic user.