Filter a data-frame and add a new column according to the given condition

Today's edition of Over Engineered with Numpy

Though admittedly very little obvious Numpy

i, rows = pd.factorize([*zip(df.ID, df.col1.replace('None'))])
k, cols = pd.factorize(df.groupby(i).cumcount())

dleft = pd.DataFrame(dict(zip(['ID', 'col1'], zip(*rows))))
drigt = pd.DataFrame(index=dleft.index, columns=np.arange(len(cols)) + 2).add_prefix('col')
drigt.values[i, k] = df.col2.values

dleft.join(drigt)

   ID        col1        col2        col3
0   1  Abc street  2017-07-27  2017-08-17
1   1  Def street  2018-07-15  2018-08-13
2   2  fbg street  2018-01-07  2018-08-12
3   2  trf street  2019-01-15         NaN

Try:

filters = df['col1'].isna()
s = df.loc[filters, 'col2'].copy()
df = df[~filters]
df['col3'] = s.values

---

Edit: as you mentioned, the filter you want is 'None', not None, then:

filters = df['col1'].eq('None')

I am using cumcount with merge

df1=df.loc[df.col1.ne('None'),:].copy()
df2=df.loc[df.col1.eq('None'),:].copy()
df1['Key']=df1.groupby('ID').cumcount()
df2['Key']=df2.groupby('ID').cumcount()
df1.merge(df2.drop('col1',1),on=['ID','Key'],how='left')
Out[816]: 
   ID       col1      col2_x  Key      col2_y
0   1  Abcstreet  2017-07-27    0  2017-08-17
1   1  Defstreet  2018-07-15    1  2018-08-13
2   2  fbgstreet  2018-01-07    0  2018-08-12
3   2  trfstreet  2019-01-15    1         NaN

Using ffill + pivot_table. This assumes that None follows the proper value, which it appears to from your data.

---

u = df.assign(col1=df.col1.replace('None'))
g = ['ID', 'col1']
idx = u.groupby(g).cumcount()

(u.assign(idx=idx)
    .pivot_table(index=g, columns='idx', values='col2', aggfunc='first')
    .reset_index())  

idx   ID        col1           0           1
0      1  Abc street  2017-07-27  2017-08-17
1      1  Def street  2018-07-15  2018-08-13
2      2  fbg street  2018-01-07  2018-08-12
3      2  trf street  2019-01-15         NaN