ipet
ipet copied to clipboard
GregorCH
Use sortlevel argument in evaluation
Have you read the documentation of sortlevel?
sortlevel : level on which to base column sorting, '0' for group level, '1' for column level
The column headers consist of >= 1 levels, depending on attributes such as indexsplit.
The sortlevel functionality should allow to specify the level of the column keys, by which the resulting long table should be sorted. Implicitly, we currently use a sortlevel of 1, I guess.
You are right, group level is too specific and has been generalized in the mean time (because multiple levels of group keys are possible).
Thank you. So it should just be a pandas call to some kind of sortlevel func in a few lines, right? Does this apply to both aggregated and long table?
Yes, there is a pandas function, (sort_axis, I guess). We are free to do whatever we want. I would suggest that it applies to both data frames, although the meaning would be a bit different for the aggregated table, because the column keys are used for the rows, as well.
There should be, IMHO, the possibilty not to sort, maybe by assigning None to the sortlevel (Note that this should be even the default behavior).
Why did you decide to push this directly instead of creating a merge request?
Because it was a simple thing (i thought). And in this project simple things seem to almost never be done via a pull request.