Overall, we are trying to compare the data from D0 to the data for D1 and see what differences in those numbers are not accounted for by the transactions that took place over the course of the day. To do this, I've converted the incoming space-separated recon.in data into different pandas dataframes. After processing the transaction data to turn it into a summary of changes to either the stocks or the total amount of cash, we can merge the data into one dataframe and quickly check for the discrepancies that we can export and report in our recon.out file.

1. Process the recon.in data to create 3 dataframes for D0-POS, D1-TRN, and D1-POS
2. Convert the D1-TRN data into a dataframe containing the stock name and the net change to the shares or the cash value
3. Merge the D0-POS, D1-POS, and the converted D1-TRN dataframes into one dataframe.
4. Calculate the recon.out using data in the dataframe; select the stock names and reconciliation values (that are not 0) and convert them into space-separated data in a txt file.

• Install pandas as pd
• Write a function get_line_number that takes the parameters string and data and from them finds the line number of that string.
• Use the get_line_number function to find the value of skiprows and the cutoff of nrows for each invocation of pd.read_csv so that we can process the 3 different dataframes (this is necessary because the D1-TRN data has a different shape than the D0-POS and D1-POS data so we cannot process it all at once)
• Use pd.read_csv with the aforementioned skiprows and nrows with delimiter=' ' to convert the space-separated data into different pandas dataframes.
• Assign standardized column names ('stock', 'shares_0' / 'stock', 'action', 'shares', 'value' / 'stock', 'shares_1') to make the data easier to manipulate and merge.

• Set up a dictionary called cash_actions for which the keys are the strings of the different actions (ex: 'BUY') and the values are -1 or 1 depending on whether that action increases or decreases the amount of cash.
• Declare an empty dictionary called change_obj and initialize a variable called cash_change to 0.
• Use a for loop to iterate through the D1-TRN dataframe values
  • If the stock name is not already in change_obj, add it as a key with the value of -1 * cash_actions[*action*] * *shares*
  • Otherwise, if it is in change_obj, add that value to the current value stored in the dictionary
  • Increment cash_change by cash_actions[*action*] * *value*
• Add a 'Cash' key to cash_obj with the value of cash_change
• Create a data from from cash_obj with the columns 'stock' and 'change'

• Using an outer join with pd.merge, merge the dataframes for D0-POS, D1-POS, and the new dataframe from the D1-TRN on the 'stock' column.
• Use the .fillna(0) to fill in any missing values in the merged dataframe.

• Add a new column to the dataframe called recon_out. Calculate it by subtracting the values in shares_0 and change from shares_1 to find any discrepancies in the transactions.
• Create a new dataframe that contains the stock name and the recon_out value where the recon_out value is not zero (i.e. there is a discrepancy).
• Use .to_csv with a sep=' ' to convert the dataframe to a space-separated txt file called recon_out.txt. Set the index and the header to false.