Pandas groupby 到新栏目

Question

I have a dataframe with columns code and images .

Column images is a string of urls joined by a comma: <URL>,<URL2>,...

Column code is NOT unique and I need to make it unique but store all images (from all variants) in a new column images_all .

For example:

code something images
1    x         url1,url2,url3
1    x         url1,url4

Result is: code something images_all 1 x url1,url2,url3,url4

I did

grouped = csv.groupby('code')
csv = csv.drop_duplicates(subset=['code'], keep='last')
csv['images_all'] = csv.apply(lambda r:  list(set(
    [image for image in grouped.get_group(r['code'])['images']]
)))

which raises:

KeyError: 'code'

But even if it didn't raise this, the problem is that images wouldn't be [url1,url2,url3,url4] . Instead, it would be ["url1,url2,url3","url1,url4"] .

Do you know how to fix it?

EDIT

I also want to keep other columns (they are the same for all rows with the same code, that's why I then just drop_duplicates and keep the last row)

Answer 1

Use GroupBy.transform with custom function for flatten splitted values, then converted to sets and last join unique values:

f = lambda x: ','.join(set([z for y in x for z in y.split(',')]))
df['images_all'] = df.groupby('code')['images'].transform(f)
print (df)
   code something          images           images_all
0     1         x  url1,url2,url3  url1,url3,url2,url4
1     1         x       url1,url4  url1,url3,url2,url4