df.reorder_levels() ... drops name attribute of DataFrame

[sanguineturtle]

Using Pandas 0.11.0
A DataFrame with MultiIndex:

df = df.reorder_levels() drops the dataframe name attribute.

thus, df.name after the reorder_levels() has no attribute 'name'

[cpcloud]

@sanguineturtle DataFrame does not have a name attribute by default. As you probably know, though, Python objects have a __dict__ attribute which is the instance's namespace and that allows one to assign arbitrary attributes at runtime to an instance (or a class). The reason name is getting dropped from this particular DataFrame instance is just that, you've assigned an attribute to an instance of a class and thus when reorder_levels creates a new instance that new instance correspondingly has a new namespace. You can assign a class variable if you want. So, this is not really an issue so much as a feature (or bug if you're feeling pessimistic) of the Python object model/runtime.

[sanguineturtle]

I see. Agreed - this is more along the lines of a feature request. The reason I like to have the DataFrame name attribute propagate is that I am using pandas to track 'labelled' matrices. I have a function that takes a DataFrame converts it to a matrix for computation and the returns it as a DataFrame. Therefore, when I do relevant data manipulations (such as reorder_levels()), it would be helpful for the name label to persist as a label of the overall matrix data. (i.e. country-product level export data).

The main benefit is that the matrix data can then be sorted and matched with other relevant data sources along their relevant axis (axis=0 and axis=1). Is there a Features Register?

[cpcloud]

A more general version of what you want is in the works, see #2485. And the name attribute is slated for 0.12, see #447.

[jreback]

closing this in favor of the above referenced issues