在Pandas DataFrame矢量化操作中随机选择行

Question

I want to select a random row during a vector operation on a DataFrame. this is what my inpDF looks like:

    string1    string2
0   abc        dfe
1   ghi        jkl
2   mno        pqr
3   stu        vwx

I'm trying to find the function getRandomRow() here:

outDF['string1'] = inpDF['string1']
outDF['string2'] = inpDF.getRandomRow()['string2']

so that the outDF ends up looking (for example) like this:

    string1    string2
0   abc        jkl
1   ghi        pqr
2   mno        dfe
3   stu        pqr

EDIT 1:

I tried using the sample() function as suggested in this answer , but that just causes the same sample to get replicated accross all rows:

outDF['string1'] = inpDF['string1']
outDF['string2'] = inpDF.sample(n=1).iloc[0,:]['string2']

which gives:

    string1    string2
0   abc        pqr
1   ghi        pqr
2   mno        pqr
3   stu        pqr

EDIT 2:

For my particular use case, even picking the value from 'n' rows down would suffice. So, I tried doing this (I'm using inpDF.index based on what I read in this answer ):

numRows = len(inpDF)

outDF['string1'] = inpDF['string1']
outDF['string2'] = inpDF.iloc[(inpDF.index + 2)%numRows,:]['string2']

but it just ends up picking the value from the same row, and the outDF comes out to be this:

    string1    string2
0   abc        dfe
1   ghi        jkl
2   mno        pqr
3   stu        vwx

whereas I'm expecting it should be this:

    string1    string2
0   abc        pqr
1   ghi        vwx
2   mno        dfe
3   stu        jkl

Answer 1

try np.random.shuffle() :

np.random.shuffle(df.string2)
print(df)

  string1 string2
0     abc     pqr
1     ghi     vwx
2     mno     def
3     stu     jkl

If you don't want to shuffle inplace try:

df['string3']=np.random.permutation(df.string2)
print(df)

Answer 2

You use pandas.DataFrame.sample for this:

df['string2'] = df.string2.sample(len(df.string2)).to_list()

print(df)
  string1 string2
0     abc     vwx
1     ghi     jkl
2     mno     def
3     stu     pqr

Or

df['string2'] = df.string2.sample(len(df.string2)).values