无法使用drop_duplicates从DataFrame中删除重复项

Question

So I am using DataFrame from Pandas, python.

The dataframe, I will be referring to was created by the following way:

search = DataFrame([[262,'ny', '20'],[515,'paris','19'],[669,'ldn','10'], [669,'ldn', 10],[669,'ldn',5]],columns = ['subscriber_id','location','radius' ])

title = DataFrame([[262,'director'],[515,'artist'],[669,'scientist']],columns = ['subscriber_id','title' ])

Both the title and search DataFrames are then merged.

mergedTable = merge(title, search, on='subscriber_id', how= 'outer')

This forms the dataframe:

   subscriber_id      title location radius
0            262   director       ny     20
1            515     artist    paris     19
2            669  scientist      ldn     10
3            669  scientist      ldn     10
4            669  scientist      ldn      5

As we can see it has been merged correctly, so we now have data for a subscriber in multiple rows dependent on their searches.

Now I do not want to get rid of the subscribers having multiple rows with different values, but I do want to get rid of duplicate rows.

This is the desired final result:

   subscriber_id      title location radius
0            262   director       ny     20
1            515     artist    paris     19
2            669  scientist      ldn     10
4            669  scientist      ldn      5

The row 3, a duplicate of row 2, is removed.

I have been researching this and it seems that drop_duplicates() should work, ie

mergedTable.drop_duplicates()

But this doesn't work, rows are not removed. Any tips/solutions available?

Answer 1

Your radius is of dtype object due to some strings within: [669,'ldn','10'] . And '10' != 10 . Converting to integer will do the trick:

>>> mergedTable.radius = mergedTable.radius.astype(int)
>>> mergedTable.drop_duplicates()
   subscriber_id      title location  radius
0            262   director       ny      20
1            515     artist    paris      19
2            669  scientist      ldn      10
4            669  scientist      ldn       5