2

How can I remove contiguous/consecutive/adjacent duplicates in a DataFrame?

I'm manipulating data in CSV format, am sorting by date, then by an identifying number. The identifying number can appear on different days, but I only want to delete the daily duplicates. drop_duplicates leaves one unique instance, but then deletes that identifier on all other days. I've tried this, but get the error:

localhost:~/Desktop/Public$ python3 test.py 
Traceback (most recent call last):
  File "test.py", line 31, in <module>
    df2.loc[df2.shift(1) != df2]
  File "/usr/lib/python3/dist-packages/pandas/core/indexing.py", line 1028, in __getitem__
    return self._getitem_axis(key, axis=0)
  File "/usr/lib/python3/dist-packages/pandas/core/indexing.py", line 1148, in _getitem_axis
    raise ValueError('Cannot index with multidimensional key')
ValueError: Cannot index with multidimensional key

Edited original post to add:

I tried index_reset() to remove any multiindex. Here's a sample of the dataset:

,DATE,REC,NAME
0,07/02/2009,682566,"Schmoe, Joe"
1,07/02/2009,244828,"Doe, Joe"
2,07/11/2009,325640,"Black, Joe"
3,07/11/2009,544440,"Dirt, Joe"
4,07/11/2009,544440,"Dirt, Joe"
5,07/16/2009,200560,"White, Joe"
6,07/16/2009,685370,"Purple, Joe"
7,07/16/2009,685370,"Purple, Joe"
8,07/16/2009,635400,"Red, Joe"
9,07/16/2009,348562,"Blue, Joe
Community
  • 1
  • 1
user3478193
  • 57
  • 1
  • 8
  • 1
    I would think `df2.drop_duplicates(['id_no','date'])` would work for you (or whatever your identifying number is called). – Karl D. May 20 '14 at 03:28
  • Would be very helpful to show an example of the dataset, and, are you using `MultiIndex`? – CT Zhu May 20 '14 at 03:42
  • Well, I feel like an idiot... drop_duplicates(['REC', 'DATE']) works... However, the original problem is still perplexing me, and I'm sure the same issue is going to come up again. Data sample to come. – user3478193 May 20 '14 at 04:03

1 Answers1

7

The way you index with .loc will only work if df2 is a Series not a DataFrame. You're essentially trying to index with a dataframe of boleens and .loc doesn't know what to do (it tries to use it as a multiindex):

>>> df

        DATE     REC         NAME
0 2009-07-02  682566  Schmoe, Joe
1 2009-07-02  244828     Doe, Joe
2 2009-07-11  325640   Black, Joe
3 2009-07-11  544440    Dirt, Joe
4 2009-07-11  544440    Dirt, Joe
5 2009-07-16  200560   White, Joe
6 2009-07-16  685370  Purple, Joe
7 2009-07-16  685370  Purple, Joe
8 2009-07-16  635400     Red, Joe
9 2009-07-16  348562    Blue, Joe

>>> df.shift() != df

    DATE    REC   NAME
0   True   True   True
1  False   True   True
2   True   True   True
3  False   True   True
4  False  False  False
5   True   True   True
6  False   True   True
7  False  False  False
8  False   True   True
9  False   True   True

Instead, you want to do something like the following:

>>> df.loc[df.DATE.shift() != df.DATE]

        DATE     REC         NAME
0 2009-07-02  682566  Schmoe, Joe
2 2009-07-11  325640   Black, Joe
5 2009-07-16  200560   White, Joe

.loc works here because we just create a boleen Series for index:

>>> df.DATE.shift() != df.DATE

0     True
1    False
2     True
3    False
4    False
5     True
6    False
7    False
8    False
9    False

Of course, that's not the data you want. To be equivalent to df.drop_duplicates(['REC','DATE']), you want the following:

>>>  df.loc[(df.DATE != df.DATE.shift(1)) | (df.REC != df.REC.shift(1))]

        DATE     REC         NAME
0 2009-07-02  682566  Schmoe, Joe
1 2009-07-02  244828     Doe, Joe
2 2009-07-11  325640   Black, Joe
3 2009-07-11  544440    Dirt, Joe
5 2009-07-16  200560   White, Joe
6 2009-07-16  685370  Purple, Joe
8 2009-07-16  635400     Red, Joe
9 2009-07-16  348562    Blue, Joe

Comparison with drop_duplicates:

>>> df.drop_duplicates(['REC','DATE'])

        DATE     REC         NAME
0 2009-07-02  682566  Schmoe, Joe
1 2009-07-02  244828     Doe, Joe
2 2009-07-11  325640   Black, Joe
3 2009-07-11  544440    Dirt, Joe
5 2009-07-16  200560   White, Joe
6 2009-07-16  685370  Purple, Joe
8 2009-07-16  635400     Red, Joe
9 2009-07-16  348562    Blue, Joe
Karl D.
  • 13,332
  • 5
  • 56
  • 38