Difference between df.reindex() and df.loc[]

Question

Say I want to compute the relative complement df2 - df1 between two MultiIndex dataframes. Assuming that they have the same indexing schema, based on what I saw in this answer from Andy Hayden, I could do the following:

diff_indices = df2.index - df1.index

And then either:

df2.reindex(diff_indices, inplace=True)

or
df2 = df2.loc[diff_indices]

What would be the difference between 1. and 2. above? What is the difference between df.reindex and df.loc?

perhaps `reindex` creates a new dataframe, `ix` returns a view — Paul H, Apr 09 '14 at 18:29
@PaulH that is not correct, NEITHER return views in general. — Jeff, Apr 09 '14 at 19:26
@Jeff you are, of course, absolutely right. what do `loc` and `iloc` do? — Paul H, Apr 09 '14 at 20:44

score 9 · Accepted Answer · edited Aug 24 '18 at 02:08

9

Both approaches return a new series/dataframe, and basically do the same thing.

The reason for the seeming redundancy is that, while using loc is syntacticly limiting (you can only pass a single argument to __getitem__), reindex is a method, which supports taking various optional parameters. (docs)

edited Aug 24 '18 at 02:08

David Nehme

21,379
8
78
117

answered Apr 09 '14 at 18:35

shx2

61,779
13
130
153

1

This is true for ``getting``, but you cannot set with ``reindex`` – Jeff Apr 09 '14 at 19:27
@jeff correct. I believe OP is asking about getting. – shx2 Apr 09 '14 at 19:28
Thanks - What happens if I I want to update `df2` with the output of these commands? I am getting different results when using `reindex` with `inplace=True` vs using `ix` (I updated the OP) – Amelio Vazquez-Reina Apr 17 '14 at 15:33
@user815423426 it would help if you explained *what* different results you're getting – shx2 Apr 17 '14 at 17:05
Thanks I will try to put together an example, but the first command one does **not** seem to change the Dataframe, while the second one does. – Amelio Vazquez-Reina Apr 17 '14 at 17:18

Difference between df.reindex() and df.loc[]

1 Answers1