You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
 
 
 
 

773 lines
31 KiB

# -*- encoding: utf-8 -*-
##############################################################################
#
# mis_builder module for OpenERP, Management Information System Builder
# Copyright (C) 2014 ACSONE SA/NV (<http://acsone.eu>)
#
# This file is a part of mis_builder
#
# mis_builder is free software: you can redistribute it and/or modify
# it under the terms of the GNU Affero General Public License v3 or later
# as published by the Free Software Foundation, either version 3 of the
# License, or (at your option) any later version.
#
# mis_builder is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU Affero General Public License v3 or later for more details.
#
# You should have received a copy of the GNU Affero General Public License
# v3 or later along with this program.
# If not, see <http://www.gnu.org/licenses/>.
#
##############################################################################
from datetime import datetime, timedelta
from dateutil import parser
import traceback
import re
import pytz
from openerp.osv import orm, fields
from openerp.tools.safe_eval import safe_eval
from openerp.tools.translate import _
from openerp import tools
from collections import OrderedDict
class AutoStruct(object):
def __init__(self, **kwargs):
for k, v in kwargs.items():
setattr(self, k, v)
def _get_selection_label(selection, value):
for v, l in selection:
if v == value:
return l
return ''
def _utc_midnight(d, tz_name, add_day=0):
d = datetime.strptime(d, tools.DEFAULT_SERVER_DATE_FORMAT)
if add_day:
d = d + timedelta(days=add_day)
utc_tz = pytz.timezone('UTC')
context_tz = pytz.timezone(tz_name)
local_timestamp = context_tz.localize(d, is_dst=False)
return datetime.strftime(local_timestamp.astimezone(utc_tz),
tools.DEFAULT_SERVER_DATETIME_FORMAT)
def _python_var(var_str):
return re.sub(r'\W|^(?=\d)', '_', var_str).lower()
def _python_bal_var(account_code):
return 'bal_' + re.sub(r'\W', '_', account_code)
def _get_bal_vars_in_expr(expr):
return re.findall(expr, r'\bbal_\w+')
def _get_vars_in_expr(expr, varnames=None):
if not varnames:
return []
varnames_re = r'\b' + r'\b|\b'.join(varnames) + r'\b'
return re.findall(expr, varnames_re)
class mis_report_kpi(orm.Model):
""" A KPI is an element of a MIS report.
In addition to a name and description, it has an expression
to compute it based on queries defined in the MIS report.
It also has various informations defining how to render it
(numeric or percentage or a string, a suffix, divider) and
how to render comparison of two values of the KPI.
KPI are ordered inside the MIS report, as some KPI expressions
can depend on other KPI that need to be computed before.
"""
_name = 'mis.report.kpi'
_columns = {
'name': fields.char(size=32, required=True,
string='Name'),
'description': fields.char(required=True,
string='Description',
translate=True),
'expression': fields.char(required=True,
string='Expression'),
'css_style': fields.char(string='CSS style expression'),
'type': fields.selection([('num', _('Numeric')),
('pct', _('Percentage')),
('str', _('String'))],
required=True,
string='Type'),
'divider': fields.selection([('1e-6', _('µ')),
('1e-3', _('m')),
('1', _('1')),
('1e3', _('k')),
('1e6', _('M'))],
string='Factor'),
'dp': fields.integer(string='Rounding'),
'suffix': fields.char(size=16, string='Suffix'),
'compare_method': fields.selection([('diff', _('Difference')),
('pct', _('Percentage')),
('none', _('None'))],
required=True,
string='Comparison Method'),
'sequence': fields.integer(string='Sequence'),
'report_id': fields.many2one('mis.report', string='Report'),
}
_defaults = {
'type': 'num',
'divider': '1',
'dp': 0,
'compare_method': 'pct',
'sequence': 100,
}
_order = 'sequence'
def _check_name(self, cr, uid, ids, context=None):
# TODO: kpi name cannot start with bal
# TODO: factor out the name check function (DRY)
for record_name in self.read(cr, uid, ids, ['name']):
if not re.match("[_A-Za-z][_a-zA-Z0-9]*$", record_name['name']):
return False
return True
_constraints = [
(_check_name, 'The name must be a valid python identifier', ['name']),
]
def onchange_name(self, cr, uid, ids, name, context=None):
# TODO: factor out the name check function (DRY)
# check it is a valid python identifier
res = {}
if name and not re.match("[_A-Za-z][_a-zA-Z0-9]*$", name):
res['warning'] = {
'title': 'Invalid name',
'message': 'The name must be a valid python identifier'}
return res
def onchange_description(self, cr, uid, ids, description, name,
context=None):
# construct name from description
res = {}
if description and not name:
res = {'value': {'name': _python_var(description)}}
return res
def onchange_type(self, cr, uid, ids, kpi_type, context=None):
res = {}
if kpi_type == 'pct':
res['value'] = {'compare_method': 'diff'}
elif kpi_type == 'str':
res['value'] = {'compare_method': 'none',
'divider': '',
'dp': 0}
return res
def _render(self, cr, uid, kpi, value, context=None):
""" render a KPI value as a unicode string, ready for display """
if kpi.type == 'num':
return self._render_num(cr, uid, value, kpi.divider, kpi.dp,
kpi.suffix, context=context)
elif kpi.type == 'pct':
return self._render_num(cr, uid, value, 0.01, kpi.dp, '%',
context=context)
else:
return unicode(value)
def _render_comparison(self, cr, uid, kpi, value, base_value,
average_value, average_base_value, context=None):
""" render the comparison of two KPI values, ready for display """
if value is None or base_value is None:
return ''
if kpi.type == 'pct':
return self._render_num(cr, uid, value - base_value, 0.01, kpi.dp,
_('pp'), sign='+', context=context)
elif kpi.type == 'num':
if average_value:
value = value / float(average_value)
if average_base_value:
base_value = base_value / float(average_base_value)
if kpi.compare_method == 'diff':
return self._render_num(cr, uid, value - base_value,
kpi.divider,
kpi.dp, kpi.suffix, sign='+',
context=context)
elif kpi.compare_method == 'pct' and base_value != 0:
return self._render_num(cr, uid, value / base_value - 1,
0.01, kpi.dp, '%',
sign='+', context=context)
return ''
def _render_num(self, cr, uid, value, divider, dp, suffix, sign='-',
context=None):
divider_label = _get_selection_label(
self._columns['divider'].selection, divider)
if divider_label == '1':
divider_label = ''
# format number following user language
lang = self.pool['res.users'].read(
cr, uid, uid, ['lang'], context=context)['lang']
language_id = self.pool['res.lang'].search(
cr, uid, [('code', '=', lang)], context=context)
value = round(value / float(divider or 1), dp) or 0
return '%s %s%s' % (self.pool['res.lang'].format(cr, uid, language_id,
'%%%s.%df' % (
sign, dp),
value,
grouping=True,
context=context),
divider_label, suffix or '')
class mis_report_query(orm.Model):
""" A query to fetch data for a MIS report.
A query works on a model and has a domain and list of fields to fetch.
At runtime, the domain is expanded with a "and" on the date/datetime field.
"""
_name = 'mis.report.query'
def _get_field_names(self, cr, uid, ids, name, args, context=None):
res = {}
for query in self.browse(cr, uid, ids, context=context):
field_names = []
for field in query.field_ids:
field_names.append(field.name)
res[query.id] = ', '.join(field_names)
return res
def onchange_field_ids(self, cr, uid, ids, field_ids, context=None):
# compute field_names
field_names = []
for field in self.pool.get('ir.model.fields').read(cr, uid,
field_ids[0][2],
['name'],
context=context):
field_names.append(field['name'])
return {'value': {'field_names': ', '.join(field_names)}}
_columns = {
'name': fields.char(size=32, required=True,
string='Name'),
'model_id': fields.many2one('ir.model', required=True,
string='Model'),
'field_ids': fields.many2many('ir.model.fields', required=True,
string='Fields to fetch'),
'field_names': fields.function(_get_field_names, type='char',
string='Fetched fields name',
store={'mis.report.query':
(lambda self, cr, uid, ids, c={}:
ids, ['field_ids'], 20), }),
'date_field': fields.many2one('ir.model.fields', required=True,
string='Date field',
domain=[('ttype', 'in', ('date',
'datetime'))]),
'domain': fields.char(string='Domain'),
'report_id': fields.many2one('mis.report', string='Report'),
}
_order = 'name'
def _check_name(self, cr, uid, ids, context=None):
# TODO: factor out the name check function (DRY)
# TODO: query name must start with bal
for record_name in self.read(cr, uid, ids, ['name']):
if not re.match("[_A-Za-z][_a-zA-Z0-9]*$", record_name['name']):
return False
return True
_constraints = [
(_check_name, 'The name must be a valid python identifier', ['name']),
]
class mis_report(orm.Model):
""" A MIS report template (without period information)
The MIS report holds:
* an implicit query fetching allow the account balances;
for each account, the balance is stored in a variable named
bal_{code} where {code} is the account code
* a list of explicit queries; the result of each query is
stored in a variable with same name as a query, containing as list
of data structures populated with attributes for each fields to fetch
* a list of KPI to be evaluated based on the variables resulting
from the balance and queries
"""
_name = 'mis.report'
_columns = {
'name': fields.char(size=32, required=True,
string='Name', translate=True),
'description': fields.char(required=False,
string='Description', translate=True),
'query_ids': fields.one2many('mis.report.query', 'report_id',
string='Queries'),
'kpi_ids': fields.one2many('mis.report.kpi', 'report_id',
string='KPI\'s'),
}
# TODO: kpi name cannot be start with query name
def create(self, cr, uid, vals, context=None):
# TODO: explain this
if 'kpi_ids' in vals:
mis_report_kpi_obj = self.pool.get('mis.report.kpi')
for idx, line in enumerate(vals['kpi_ids']):
if line[0] == 0:
line[2]['sequence'] = idx + 1
else:
mis_report_kpi_obj.write(
cr, uid, [line[1]], {'sequence': idx + 1},
context=context)
return super(mis_report, self).create(cr, uid, vals, context=context)
def write(self, cr, uid, ids, vals, context=None):
# TODO: explain this
res = super(mis_report, self).write(
cr, uid, ids, vals, context=context)
mis_report_kpi_obj = self.pool.get('mis.report.kpi')
for report in self.browse(cr, uid, ids, context):
for idx, kpi in enumerate(report.kpi_ids):
mis_report_kpi_obj.write(
cr, uid, [kpi.id], {'sequence': idx + 1}, context=context)
return res
class mis_report_instance_period(orm.Model):
""" A MIS report instance has the logic to compute
a report template for a give date period.
Periods have a duration (day, week, fiscal period) and
are defined as an offset relative to a pivot date.
"""
def _get_dates(self, cr, uid, ids, field_names, arg, context=None):
if isinstance(ids, (int, long)):
ids = [ids]
res = {}
for c in self.browse(cr, uid, ids, context=context):
d = parser.parse(c.report_instance_id.pivot_date)
if c.type == 'd':
date_from = d + timedelta(days=c.offset)
date_to = date_from + timedelta(days=c.duration - 1)
date_from = date_from.strftime(
tools.DEFAULT_SERVER_DATE_FORMAT)
date_to = date_to.strftime(tools.DEFAULT_SERVER_DATE_FORMAT)
period_ids = None
elif c.type == 'w':
date_from = d - timedelta(d.weekday())
date_from = date_from + timedelta(days=c.offset * 7)
date_to = date_from + timedelta(days=(7 * c.duration) - 1)
date_from = date_from.strftime(
tools.DEFAULT_SERVER_DATE_FORMAT)
date_to = date_to.strftime(tools.DEFAULT_SERVER_DATE_FORMAT)
period_ids = None
elif c.type == 'fp':
period_obj = self.pool['account.period']
all_period_ids = period_obj.search(
cr, uid,
[('special', '=', False),
'|', (
'company_id', '=',
False),
('company_id', '=',
c.company_id.id)],
order='date_start',
context=context)
current_period_ids = period_obj.search(
cr, uid,
[('special', '=', False),
('date_start', '<=', d),
('date_stop',
'>=', d),
'|',
('company_id',
'=', False),
('company_id', '=', c.company_id.id)],
context=context)
if not current_period_ids:
raise orm.except_orm(_("Error!"),
_("No current fiscal period for %s")
% d)
p = all_period_ids.index(current_period_ids[0]) + c.offset
if p < 0 or p >= len(all_period_ids):
raise orm.except_orm(_("Error!"),
_("No such fiscal period for %s "
"with offset %d") % (d, c.offset))
period_ids = all_period_ids[p:p + c.duration]
periods = period_obj.browse(cr, uid, period_ids,
context=context)
date_from = periods[0].date_start
date_to = periods[-1].date_stop
else:
raise orm.except_orm(_("Error!"),
_("Unimplemented period type %s") %
(c.type,))
res[c.id] = {
'date_from': date_from,
'date_to': date_to,
'period_from': period_ids and period_ids[0],
'period_to': period_ids and period_ids[-1],
}
return res
_name = 'mis.report.instance.period'
_columns = {
'name': fields.char(size=32, required=True,
string='Description', translate=True),
'type': fields.selection([('d', _('Day')),
('w', _('Week')),
('fp', _('Fiscal Period')),
# ('fy', _('Fiscal Year'))
],
required=True,
string='Period type'),
'offset': fields.integer(string='Offset',
help='Offset from current period'),
'duration': fields.integer(string='Duration',
help='Number of periods'),
'date_from': fields.function(_get_dates,
type='date',
multi="dates",
string="From"),
'date_to': fields.function(_get_dates,
type='date',
multi="dates",
string="To"),
'period_from': fields.function(_get_dates,
type='many2one', obj='account.period',
multi="dates", string="From period"),
'period_to': fields.function(_get_dates,
type='many2one', obj='account.period',
multi="dates", string="To period"),
'sequence': fields.integer(string='Sequence'),
'report_instance_id': fields.many2one('mis.report.instance',
string='Report Instance'),
'comparison_column_ids': fields.many2many(
'mis.report.instance.period',
'mis_report_instance_period_rel',
'period_id',
'compare_period_id',
string='Compare with'),
'company_id': fields.many2one('res.company', 'Company', required=True),
'normalize_factor': fields.integer(
string='Factor',
help='Factor to use to normalize the period (used in comparison'),
}
_defaults = {
'offset': -1,
'duration': 1,
'sequence': 100,
'normalize_factor': 1,
'company_id': lambda s, cr, uid, c:
s.pool.get('res.company')._company_default_get(
cr, uid,
'mis.report.instance.period',
context=c)
}
_order = 'sequence'
_sql_constraints = [
('duration', 'CHECK (duration>0)',
'Wrong duration, it must be positive!'),
('normalize_factor', 'CHECK (normalize_factor>0)',
'Wrong normalize factor, it must be positive!'),
('name_unique', 'unique(name, report_instance_id)',
'Period name should be unique by report'),
]
def _fetch_balances(self, cr, uid, c, context=None):
""" fetch the general account balances for the given period
returns a dictionary {bal_<account.code>: account.balance}
"""
if context is None:
context = {}
account_obj = self.pool['account.account']
search_ctx = dict(context)
if c.period_from:
search_ctx.update({'period_from': c.period_from.id,
'period_to': c.period_to.id})
else:
search_ctx.update({'date_from': c.date_from,
'date_to': c.date_to})
# TODO: initial balance?
# TODO: use child of company_id?
account_ids = account_obj.search(
cr, uid,
['|', ('company_id', '=', False),
('company_id', '=', c.company_id.id)],
context=context)
account_datas = account_obj.read(
cr, uid, account_ids, ['code', 'balance'], context=search_ctx)
balances = {}
for account_data in account_datas:
key = _python_bal_var(account_data['code'])
assert key not in balances
balances[key] = account_data['balance']
return balances
def _fetch_queries(self, cr, uid, c, context):
res = {}
report = c.report_instance_id.report_id
for query in report.query_ids:
obj = self.pool[query.model_id.model]
domain = query.domain and safe_eval(query.domain) or []
if query.date_field.ttype == 'date':
domain.extend([(query.date_field.name, '>=', c.date_from),
(query.date_field.name, '<=', c.date_to)])
else:
datetime_from = _utc_midnight(
c.date_from, context.get('tz', 'UTC'))
datetime_to = _utc_midnight(
c.date_to, context.get('tz', 'UTC'), add_day=1)
domain.extend([(query.date_field.name, '>=', datetime_from),
(query.date_field.name, '<', datetime_to)])
if obj._columns.get('company_id', False):
domain.extend(['|', ('company_id', '=', False),
('company_id', '=', c.company_id.id)])
field_names = [field.name for field in query.field_ids]
obj_ids = obj.search(cr, uid, domain, context=context)
obj_datas = obj.read(
cr, uid, obj_ids, field_names, context=context)
res[query.name] = [AutoStruct(**d) for d in obj_datas]
return res
def _get_bal_vars_in_report(self, report):
res = set()
for kpi in report.kpi_ids:
for bal_var in _get_bal_vars_in_expr(kpi.expression):
res.add(bal_var)
return res
def _compute(self, cr, uid, c, context=None):
if context is None:
context = {}
kpi_obj = self.pool['mis.report.kpi']
res = {}
localdict = {
'registry': self.pool,
'sum': sum,
'min': min,
'max': max,
'len': len,
'avg': lambda l: sum(l) / float(len(l)),
}
localdict.update(self._fetch_balances(cr, uid, c, context=context))
localdict.update(self._fetch_queries(cr, uid, c, context=context))
for kpi in c.report_instance_id.report_id.kpi_ids:
try:
kpi_val = safe_eval(kpi.expression, localdict)
except ZeroDivisionError:
kpi_val = None
kpi_val_rendered = '#DIV/0'
kpi_val_comment = traceback.format_exc()
except:
kpi_val = None
kpi_val_rendered = '#ERR'
kpi_val_comment = traceback.format_exc()
else:
kpi_val_rendered = kpi_obj._render(
cr, uid, kpi, kpi_val, context=context)
kpi_val_comment = None
localdict[kpi.name] = kpi_val
try:
kpi_style = None
if kpi.css_style:
kpi_style = safe_eval(kpi.css_style, localdict)
except:
kpi_style = None
res[kpi.name] = {
'val': kpi_val,
'val_r': kpi_val_rendered,
'val_c': kpi_val_comment,
'style': kpi_style,
'suffix': kpi.suffix,
'dp': kpi.dp,
'is_percentage': kpi.type == 'pct',
}
return res
class mis_report_instance(orm.Model):
""" The MIS report instance combines compute and
display a MIS report template for a set of periods """
def _get_pivot_date(self, cr, uid, ids, field_name, arg, context=None):
res = {}
for r in self.browse(cr, uid, ids, context=context):
if r.date:
res[r.id] = r.date
else:
res[r.id] = fields.date.context_today(self, cr, uid,
context=context)
return res
_name = 'mis.report.instance'
_columns = {
'name': fields.char(size=32, required=True,
string='Name', translate=True),
'description': fields.char(required=False,
string='Description', translate=True),
'date': fields.date(string='Base date',
help='Report base date '
'(leave empty to use current date)'),
'pivot_date': fields.function(_get_pivot_date,
type='date',
string="Pivot date"),
'report_id': fields.many2one('mis.report',
required=True,
string='Report'),
'period_ids': fields.one2many('mis.report.instance.period',
'report_instance_id',
required=True,
string='Periods'),
'target_move': fields.selection([('posted', 'All Posted Entries'),
('all', 'All Entries'),
], 'Target Moves', required=True),
}
_defaults = {
'target_move': 'posted',
}
def create(self, cr, uid, vals, context=None):
if 'period_ids' in vals:
mis_report_instance_period_obj = self.pool.get(
'mis.report.instance.period')
for idx, line in enumerate(vals['period_ids']):
if line[0] == 0:
line[2]['sequence'] = idx + 1
else:
mis_report_instance_period_obj.write(
cr, uid, [line[1]], {'sequence': idx + 1},
context=context)
return super(mis_report_instance, self).create(cr, uid, vals,
context=context)
def write(self, cr, uid, ids, vals, context=None):
res = super(mis_report_instance, self).write(
cr, uid, ids, vals, context=context)
mis_report_instance_period_obj = self.pool.get(
'mis.report.instance.period')
for instance in self.browse(cr, uid, ids, context):
for idx, period in enumerate(instance.period_ids):
mis_report_instance_period_obj.write(
cr, uid, [period.id], {'sequence': idx + 1},
context=context)
return res
def _format_date(self, cr, uid, date, context=None):
# format date following user language
lang = self.pool['res.users'].read(
cr, uid, uid, ['lang'], context=context)['lang']
language_id = self.pool['res.lang'].search(
cr, uid, [('code', '=', lang)], context=context)[0]
tformat = self.pool['res.lang'].read(
cr, uid, language_id, ['date_format'])['date_format']
return datetime.strftime(datetime.strptime(
date,
tools.DEFAULT_SERVER_DATE_FORMAT),
tformat)
def compute(self, cr, uid, _ids, context=None):
assert isinstance(_ids, (int, long))
if context is None:
context = {}
r = self.browse(cr, uid, _ids, context=context)
context['state'] = r.target_move
content = OrderedDict()
# empty line name for header
header = OrderedDict()
header[''] = {'kpi_name': '', 'cols': []}
# initialize lines with kpi
for kpi in r.report_id.kpi_ids:
content[kpi.name] = {'kpi_name': kpi.description, 'cols': []}
report_instance_period_obj = self.pool.get(
'mis.report.instance.period')
kpi_obj = self.pool.get('mis.report.kpi')
period_values = {}
for period in r.period_ids:
# add the column header
header['']['cols'].append(dict(
name=period.name,
date=(period.duration > 1 or period.type == 'w') and
_('from %s to %s' %
(period.period_from and period.period_from.name
or self._format_date(cr, uid, period.date_from,
context=context),
period.period_to and period.period_to.name
or self._format_date(cr, uid, period.date_to,
context=context)))
or period.period_from and period.period_from.name or
period.date_from))
# compute kpi values
values = report_instance_period_obj._compute(
cr, uid, period, context=context)
period_values[period.name] = values
for key in values:
content[key]['cols'].append(values[key])
# add comparison column
for period in r.period_ids:
for compare_col in period.comparison_column_ids:
# add the column header
header['']['cols'].append(
dict(name='%s - %s' % (period.name, compare_col.name),
date=''))
column1_values = period_values[period.name]
column2_values = period_values[compare_col.name]
for kpi in r.report_id.kpi_ids:
content[kpi.name]['cols'].append(
{'val_r': kpi_obj._render_comparison(
cr,
uid,
kpi,
column1_values[kpi.name]['val'],
column2_values[kpi.name]['val'],
period.normalize_factor,
compare_col.normalize_factor,
context=context)})
return {'header': header,
'content': content}