Pandas + Groupby + Filter unique values

[JosepMaria]

1. First, convert the date column into a datetime() index:

df.date = pd.to_datetime(df.date)

2. Next use min() and list comprehension to create a new column containing your earliest (use max() for latest) type a dates per user as follows:

df['a_type_date'] = [min(df[(df.user==i)&(df.type=='a')]['date']).date() for i in df.user]