Reording a multiindex by only specifiying the order on level 0

[jankatins]

starting PR is here: #6647

By reading the docs on Dataframe.reindex(), I would have guessed (not sure what "broadcast" is...) that this works:

arrays = [['bar', 'bar', 'baz', 'baz', 'foo', 'foo', 'qux', 'qux'],
['one', 'two', 'one', 'two', 'one', 'two', 'one', 'two']]
tuples = zip(*arrays)
index = MultiIndex.from_tuples(tuples, names=['first', 'second'])
df = DataFrame({"a":list("abcdefgh"), "b":list("abcdefgh")}, index=index)
order = ["baz", "bar", "foo", "qux"]
df.reindex(order, level=0) # still returns the same ordering

It works that way on a normal index (without the level argument):

df2 = DataFrame({"a":list("ABCD")}, index=['bar', 'baz', 'foo', 'qux'])
df2.reindex(order) # new ordering

It also works when I pass a list of tuples:

_new_index = []
for _a in order:
    for _b in df.index.levels[1]:
        _new_index.append((_a, _b))
df.reindex(_new_index) # new ordering

[jtratner]

Are you missing a line where you convert tuples to MultiIndex there?

[cpcloud]

also this

In [5]: df.reindex(order)
Out[5]:
         a    b
b a z  NaN  NaN
    r  NaN  NaN
f o o  NaN  NaN
q u x  NaN  NaN

seems wrong.

[jankatins]

@jtratner Sorry, yes, that line went missing. Change it above and now it (not..) worked in a new ipython session

@cpcloud it's df2, not df! (but yes, I also encountered this behavior :-) seems that the the string is taken as an iterable)

[jreback]

@JanSchulz turns out this was not implemented kind of trivial once I figured out what the problem was.

[cpcloud]

xref: http://stackoverflow.com/questions/11194610/how-can-i-reorder-multi-indexed-dataframe-columns-at-a-specific-level