Pandas bug + creating a list of dataframes from paystubs

gistfile1.txt

@login_required
@json_view
def api_generate_941_report(request):
	""" Generate 941 report """
	params = {}
	params['success'] = True
	params['is_valid'] = True

	company_id = request.REQUEST.get("company_id")
	if company_id and request.user.is_superuser:
		company = MCompany.objects.get(id=company_id)
	else:
		company = request.user.get_company()

	try:
		start = parse(request.REQUEST.get("start"))
		end = parse(request.REQUEST.get("end"))
	except:
		return dict(success=False,error="Must specify start and end dates")


	paystubs = MPaystub.objects.filter( Q(cid=company.id) & 
										Q(chck_dt__lte=end) &
										Q(chck_dt__gte=start) &
										Q(is_submitted=True))

	if not paystubs.count():
		return {
			'success' : True,
			'is_valid' : False
		}

	# aggregate paystub dictionaries for earnings, general deductions + garnishments
	employee_earnings_results = pandas.concat( [pandas.DataFrame( stub.employee_earnings ) for stub in paystubs ] )
	employee_deductions_results = pandas.concat( [pandas.DataFrame( stub.employee_deductions, index=[0] ) for stub in paystubs if stub.employee_deductions ] )
	employee_garnishments_results = pandas.concat( [pandas.DataFrame( stub.employee_garnishments, index=[0] ) for stub in paystubs ] )
	employee_taxes_results = pandas.concat( [pandas.DataFrame( stub.employee_taxes ) for stub in paystubs ] )

	re_fit_reg 			= "00-000-0000-FIT-000-Res-Reg"
	re_fit_sup 			= "00-000-0000-FIT-000-Res-Sup"
	gross_compensation = round(employee_earnings_results['ste_gross']['ctd'].sum(),2)
	federal_withholding = round(employee_taxes_results[re_fit_reg]['ctd'].sum() + employee_taxes_results[re_fit_sup]['ctd'].sum() )

you need to format the data frame so all the pay stub attributes are columns, and the different paystubs are rows, and then you can do something like this

df = pd.Dataframe(paystubs formatted as rows)
rows = df.start_date >= and df.end_date <= ....... more filters
results = df[rows].groupby(df[some attribute]).sum()
return results.to_json(orient='records') <--- returns json for display

or

return results.to_csv("path to csv", sep=",") <--- dumps to csv

The point is, the concatenation of paystubs should be done in mongo, and then returned and loaded into the DataFrame as a list of python dictionaries.