Select rows based on a criteria

[zivoni]

Selecting rows with row number=1 over partition can be emulated with .groupby() followed by .first():

df['absdatediff'] = abs(df.DateDiff)
first_ones = df.reset_index().sort_values('absdatediff').groupby('ID_x').first().set_index('index')

reset_index() and set_index is used to preserve original index for joining with rows with df.CrCode_y == 9000

To get dataframe containing all "first_ones" rows and all rows with df.CrCode == 9000 without eventual duplicities pd.Index.union() can be used:

 df.loc[pd.Index.union(df[df.DrCode_y == 9000].index, first_ones.index)]