#!/usr/bin/env python

from __future__ import print_function
import argparse
from collections import OrderedDict
import csv
from datetime import datetime, timedelta
from decimal import Decimal

parser = argparse.ArgumentParser()
parser.add_argument('incoming', help='CSV export of transactions')
parser.add_argument('outgoing', help='CSV export of internal transactions')
args = parser.parse_args()

# A place to gather the incoming transaction data
data_in = OrderedDict()

with open(args.incoming) as f:
	incoming = csv.reader(f)
	# Skip CSV header row
	next(incoming)
	for _,_,timestamp,_,_,_,_,value_in,value_out,_,_,_,_,_,_ in incoming:
		# Sanity check, just in case
		assert not float(value_out)
		# Round the timestamp down to the last 10-minute mark
		timestamp = datetime.utcfromtimestamp(int(timestamp))
		rounded_minute = timestamp.minute // 10 * 10
		timestamp = timestamp.replace(minute=rounded_minute, second=0, microsecond=0)
		# First parsed row, store initial record 
		if not data_in:
			first_timestamp = timestamp
			data_in[timestamp] = Decimal(value_in)
			last_timestamp = timestamp
			continue
		# We've already had a transaction for this 10-minute period, add this one to it
		if timestamp in data_in:
			data_in[timestamp] += Decimal(value_in)
			last_timestamp = timestamp
			continue
		# Fill in 10-minute periods without transactions with the total from
		# the previous period (probably unnecessary, but avoids gaps)
		while last_timestamp + timedelta(minutes=10) < timestamp:
			data_in[last_timestamp + timedelta(minutes=10)] = data_in[last_timestamp]
			last_timestamp += timedelta(minutes=10)
		# And now add the new 10-minute period with the cumulative total
		data_in[timestamp] = data_in[last_timestamp] + Decimal(value_in)
		last_timestamp = timestamp

# This is where we'll gather the outgoing transaction data
data_out = OrderedDict()
# Start it off with 0 at the time of the first incoming transaction
data_out[first_timestamp] = Decimal('0.0')
# The reason I keep adding these silly *_timestamp variables is because I can't
# do data_in.keys()[0] or data_in.keys()[-1], and I don't want to keep casting
# them into lists.
last_out_timestamp = first_timestamp

with open(args.outgoing) as f:
	outgoing = csv.reader(f)
	# Skip CSV header row
	next(outgoing)
	for _,_,timestamp,_,_,_,_,value_in,value_out,_,_,_,_,_ in outgoing:
		# Sanity check, just in case
		assert not float(value_in)
		# Round the timestamp down to the last 10-minute mark
		timestamp = datetime.utcfromtimestamp(int(timestamp))
		rounded_minute = timestamp.minute // 10 * 10
		timestamp = timestamp.replace(minute=rounded_minute, second=0, microsecond=0)
		# We've already had a transaction for this 10-minute period, add this one to it
		if timestamp in data_out:
			data_out[timestamp] += Decimal(value_out)
			last_out_timestamp = timestamp
			continue
		# Fill in 10-minute periods without transactions as we did for incoming
		while last_out_timestamp + timedelta(minutes=10) < timestamp:
			data_out[last_out_timestamp + timedelta(minutes=10)] = data_out[last_out_timestamp]
			last_out_timestamp += timedelta(minutes=10)
		# And now add the new 10-minute period with the cumulative total
		data_out[timestamp] = data_out[last_out_timestamp] + Decimal(value_in)
		last_out_timestamp = timestamp

# If the outgoing transactions end more than 10 minutes before
# the incoming ones, fill in the empty periods at the end
while last_out_timestamp < last_timestamp:
	data_out[last_out_timestamp + timedelta(minutes=10)] = data_out[last_out_timestamp]
	last_out_timestamp += timedelta(minutes=10)
	
# Output our data in CSV format
for k, v in data_in.items():
	print('%s,%s,%s' % (k, v, data_out[k]))