Map duplicate rows in Pandas dataframe by vlalue of a column?-CodePudding

In a dataframe with 2 columns [id][string], I need to know which lines are duplicates of which lines based on the value of the column [string]. My dataframe has thousands of rows but only 2 columns.

Sample of the input dataframe:

id,string
0,"A B C D"
1,"D B C D E Z"
2,"A B C D"
3,"Z Z Z Z Z Z Z Z Z Z Z Z"
4,"D B C D E Z"
5,"A B C D"

In this sample, rows 0, 2, 5 are duplicates of each other. Also rows 1 and 4 are duplicates of each other. (id is unique)

I want the following output:

[["0","2","5"]],["1","4"]]

CodePudding user response：

I'd use groupby and a listcomp.

>>> df
   id                   string
0   0                  A B C D
1   1              D B C D E Z
2   2                  A B C D
3   3  Z Z Z Z Z Z Z Z Z Z Z Z
4   4              D B C D E Z
5   5                  A B C D
>>>
>>> [l for l in df.groupby('string')['id'].apply(list) if len(l) > 1]
[[0, 2, 5], [1, 4]]

If you really want strings in the result, use

>>> [[str(x) for x in l] for l in df.groupby('string')['id'].apply(list) if len(l) > 1]
[['0', '2', '5'], ['1', '4']]

CodePudding user response：

You can filter by length of lists after aggregate list per string in boolean indexing with Series.str.len:

s = df.assign(id = df['id'].astype(str)).groupby('string')['id'].apply(list)
out = s[s.str.len().gt(1)].tolist()

If already id are strings:

s =  df.groupby('string')['id'].apply(list)
out = s[s.str.len().gt(1)].tolist()

CodePudding user response：

another option is using duplicated:

>>> df[df.duplicated('string',False)].groupby('string')['id'].apply(lambda x: x.astype(str).tolist()).tolist()

# [['0', '2', '5'], ['1', '4']]

if there's no need to change 'id' type:

>>> df[df.duplicated('string',False)].groupby('string')['id'].apply(list).tolist()

# [[0, 2, 5], [1, 4]]