The __format_version__ in _version.py has not been increased eventhough the format got changed!
This is an issue in the develop branch not in master!
In my fork I made an update to many test cases since I changed the format, so I saved many networks in as files, they contain the current format_version (2.14.0). After merging the current version of develop I got some tests that suddenly failed eventhough my code should not mess with them. So I did a little diging and found that the expected and actual results differ in net.res_switch_est DataFrame. This is because the expected result only contains the old columns while the actual result contains the updated columns.
This is because the expected results are loaded form file using the pandapower.from_json function and since then format version is the same as the current format verison in _version.py the conversion to the newest format is not done. So the network is returned as loaded from file.
The actual results however are a product of a conversion from a different network type. So they are the output of a converter that creates a new pandapowerNet. These then contain all new columns.
If new columns are added __format_version__ should be incremented at least in the bugfix number. But I would expect that this constitutes at least a minor release as a new format version most likely breaks backwards compatibility. On a bugfix version I would expect I go backwards and forwards without issue. But this is not the case if the format version changes! A 2.13.1 Network should sucessfully load on 2.13.0 but this will not work if new columns are added. So this change should be reflected by an increase of the format verison to at least 2.15.0 in my opinion.
The breaking commit is 516f8af as it changed the format without changeing the format version.
Issue Description
The
__format_version__
in_version.py
has not been increased eventhough the format got changed!This is an issue in the develop branch not in master!
In my fork I made an update to many test cases since I changed the format, so I saved many networks in as files, they contain the current format_version (2.14.0). After merging the current version of develop I got some tests that suddenly failed eventhough my code should not mess with them. So I did a little diging and found that the expected and actual results differ in
net.res_switch_est
DataFrame. This is because the expected result only contains the old columns while the actual result contains the updated columns.This is because the expected results are loaded form file using the
pandapower.from_json
function and since then format version is the same as the current format verison in_version.py
the conversion to the newest format is not done. So the network is returned as loaded from file. The actual results however are a product of a conversion from a different network type. So they are the output of a converter that creates a new pandapowerNet. These then contain all new columns.If new columns are added
__format_version__
should be incremented at least in the bugfix number. But I would expect that this constitutes at least a minor release as a new format version most likely breaks backwards compatibility. On a bugfix version I would expect I go backwards and forwards without issue. But this is not the case if the format version changes! A 2.13.1 Network should sucessfully load on 2.13.0 but this will not work if new columns are added. So this change should be reflected by an increase of the format verison to at least 2.15.0 in my opinion.The breaking commit is 516f8af as it changed the format without changeing the format version.