BUG: SparseDataFrame does not allow single value data

[BrenBarn]

You can create a normal DataFrame with, e.g., pandas.DataFrame(0, index=[1, 2, 3], columns=["A", "B", "C"]). However, this fails with a SparseDataFrame:

>>> x = pandas.SparseDataFrame(0, index=[1, 2, 3], columns=["A", "B", "C"])
Traceback (most recent call last):
  File "<pyshell#97>", line 1, in <module>
    x = pandas.SparseDataFrame(0, index=[1, 2, 3], columns=["A", "B", "C"])
  File "c:\users\brenbarn\documents\python\extensions\pandas\pandas\sparse\frame.py", line 123, in __init__
    NDFrame.__init__(self, mgr)
UnboundLocalError: local variable 'mgr' referenced before assignment

It looks like the code in SparseDataFrame.__init__ is not handling the case where the data is a single value. It has a bunch of if statements to handle different kinds of data, but falls through without creating mgr if none of them are met.

This makes it difficult to create an "empty" SparseDataFrame whose default fill value is something other than nan, e.g., pandas.SparseDataFrame(0, index=[1, 2, 3], columns=["A", "B", "C"], default_fill_value=0). In fact, I would suggest that if a default_fill_value is provided but no data is provided, the SparseDataFrame should be filled with the fill value (not nan), but perhaps that warrants a separate issue.

[jreback]

would of course welcome PR / API ideas on this (and other sparse issues).

[Motorrat]

# "real world" data that can be used to demonstrate this behavior
#http://archive.ics.uci.edu/ml/datasets/URL+Reputation
# http://scikit-learn.org/stable/modules/generated/sklearn.datasets.load_svmlight_file.html
import pandas as pd
from sklearn.datasets import load_svmlight_file
X,y = load_svmlight_file('Day0.svm')
dataframe=pd.SparseDataFrame(X)

Traceback (most recent call last):
File "", line 1, in
File "C:\Anaconda\lib\site-packages\pandas\sparse\frame.py", line 122, in init
NDFrame.init(self, mgr)
UnboundLocalError: local variable 'mgr' referenced before assignment

[jreback]

looks like this is not tested, so pull-requests are welcome.

this is a nice work-around actually.

In [28]: pd.SparseDataFrame( index=[1, 2, 3], columns=["A", "B", "C"]).fillna(0)
Out[28]: 
     A    B    C
1  0.0  0.0  0.0
2  0.0  0.0  0.0
3  0.0  0.0  0.0