99 lines
4.2 KiB
Python
99 lines
4.2 KiB
Python
#-*- coding:utf-8 -*-
|
|
# Part of Odoo, Flectra. See LICENSE file for full copyright and licensing details.
|
|
|
|
from flectra import api, models
|
|
|
|
class PayslipDetailsReport(models.AbstractModel):
|
|
_name = 'report.hr_payroll.report_payslipdetails'
|
|
|
|
def get_details_by_rule_category(self, payslip_lines):
|
|
PayslipLine = self.env['hr.payslip.line']
|
|
RuleCateg = self.env['hr.salary.rule.category']
|
|
|
|
def get_recursive_parent(current_rule_category, rule_categories=None):
|
|
if rule_categories:
|
|
rule_categories = current_rule_category | rule_categories
|
|
else:
|
|
rule_categories = current_rule_category
|
|
|
|
if current_rule_category.parent_id:
|
|
return get_recursive_parent(current_rule_category.parent_id, rule_categories)
|
|
else:
|
|
return rule_categories
|
|
|
|
res = {}
|
|
result = {}
|
|
|
|
if payslip_lines:
|
|
self.env.cr.execute("""
|
|
SELECT pl.id, pl.category_id, pl.slip_id FROM hr_payslip_line as pl
|
|
LEFT JOIN hr_salary_rule_category AS rc on (pl.category_id = rc.id)
|
|
WHERE pl.id in %s
|
|
GROUP BY rc.parent_id, pl.sequence, pl.id, pl.category_id
|
|
ORDER BY pl.sequence, rc.parent_id""",
|
|
(tuple(payslip_lines.ids),))
|
|
for x in self.env.cr.fetchall():
|
|
result.setdefault(x[2], {})
|
|
result[x[2]].setdefault(x[1], [])
|
|
result[x[2]][x[1]].append(x[0])
|
|
for payslip_id, lines_dict in result.items():
|
|
res.setdefault(payslip_id, [])
|
|
for rule_categ_id, line_ids in lines_dict.items():
|
|
rule_categories = RuleCateg.browse(rule_categ_id)
|
|
lines = PayslipLine.browse(line_ids)
|
|
level = 0
|
|
for parent in get_recursive_parent(rule_categories):
|
|
res[payslip_id].append({
|
|
'rule_category': parent.name,
|
|
'name': parent.name,
|
|
'code': parent.code,
|
|
'level': level,
|
|
'total': sum(lines.mapped('total')),
|
|
})
|
|
level += 1
|
|
for line in lines:
|
|
res[payslip_id].append({
|
|
'rule_category': line.name,
|
|
'name': line.name,
|
|
'code': line.code,
|
|
'total': line.total,
|
|
'level': level
|
|
})
|
|
return res
|
|
|
|
def get_lines_by_contribution_register(self, payslip_lines):
|
|
result = {}
|
|
res = {}
|
|
for line in payslip_lines.filtered('register_id'):
|
|
result.setdefault(line.slip_id.id, {})
|
|
result[line.slip_id.id].setdefault(line.register_id, line)
|
|
result[line.slip_id.id][line.register_id] |= line
|
|
for payslip_id, lines_dict in result.items():
|
|
res.setdefault(payslip_id, [])
|
|
for register, lines in lines_dict.items():
|
|
res[payslip_id].append({
|
|
'register_name': register.name,
|
|
'total': sum(lines.mapped('total')),
|
|
})
|
|
for line in lines:
|
|
res[payslip_id].append({
|
|
'name': line.name,
|
|
'code': line.code,
|
|
'quantity': line.quantity,
|
|
'amount': line.amount,
|
|
'total': line.total,
|
|
})
|
|
return res
|
|
|
|
@api.model
|
|
def get_report_values(self, docids, data=None):
|
|
payslips = self.env['hr.payslip'].browse(docids)
|
|
return {
|
|
'doc_ids': docids,
|
|
'doc_model': 'hr.payslip',
|
|
'docs': payslips,
|
|
'data': data,
|
|
'get_details_by_rule_category': self.get_details_by_rule_category(payslips.mapped('details_by_salary_rule_category').filtered(lambda r: r.appears_on_payslip)),
|
|
'get_lines_by_contribution_register': self.get_lines_by_contribution_register(payslips.mapped('line_ids').filtered(lambda r: r.appears_on_payslip)),
|
|
}
|