conditional cumulative sum based on column comparison in pandas dataframe

Question

I'm relatively new to pandas, and I'm sure there is an easy solution, but I could not figure it out on my own. I have a dataframe of transactions that looks like this:

OrderId Size    Price           Side    TimeSecO    TimeUSecO   TimeSecOT   TimeUSecOT AmountBuy AmountSell
10      100     41.44000000     BUY     1403200077  47720       1403200100  640070
11      100     41.43000000     BUY     1403200077  47979       1403200112  43383
12      100     41.45000000     SELL    1403200077  48311       1403200090  61100
14      100     41.45000000     BUY     1403200092  253793      1403200092  374767
17      100     41.44000000     SELL    1403200103  24382       1403200125  929563
20      100     41.43000000     SELL    1403200116  208057      1403200116  226762
31      100     41.46000000     SELL    1403200214  874124      1403200259  751002
37      100     41.46000000     BUY     1403200278  494827      1403200300  729545
42      100     41.45000000     BUY     1403200335  601039      1403200361  925384
42      100     41.45000000     BUY     1403200335  601039      1403200361  925415
45      500     15.54000000     SELL    1403200365  997248      1403200741  26216
49      100     41.45000000     SELL    1403200375  419253      1403200402  959968
53      100     42.61000000     SELL    1403200377  403525      1403200377  403680
54      100     42.61000000     BUY     1403200377  501636      1403200377  501770

I want to calculate rolling cumulative sums for each OrderId and put them into 2 new columns with respect to Side column, CumAmountBuy and CumAmountSell, where TimeSecO > TimeSecOT.

For example, for the above dataframe the correct cumulative sums for OrderId 10, OrderId 11 and OrderId 12 would be CumAmountBuy = 0 and CumAmountSell = 0, because there are no records in the dataframe where 1403200077 > TimeUSecOT.

For OrderId 14, CumAmountBuy = 0, and CumAmountSell = 100, as OrderId 12 has already happened at this point, and it was a Side=SELL, and it fulfilled the requirement of TimeSecO > TimeSecOT (1403200092 > 1403200090).

Answer 1

I can think of a dirty trick but when the dataframe gets huge, I don't think it's efficient.

In [42]: df['flag'] = df.TimeSecO.map(lambda sec: (sec > df.TimeSecOT).values)

In [43]: df['CumAmountBuy'] = df.flag.map(lambda f: np.dot(f,df['Size']*(df['Side']=='BUY')))

In [44]: df['CumAmountSell'] = df.flag.map(lambda f: np.dot(f,df['Size']*(df['Side']=='SELL')))

In [45]: df[['CumAmountBuy','CumAmountSell']]
Out[45]: 
         CumAmountBuy  CumAmountSell
OrderId                             
10                  0              0
11                  0              0
12                  0              0
14                  0            100
17                200            100
20                300            100
31                300            300
37                300            400
42                400            400
42                400            400
45                600            400
49                600            400
53                600            400
54                600            400