rating/models/rating_mixin.py

183 lines
9.8 KiB
Python
Raw Permalink Normal View History

2024-05-03 15:15:12 +03:00
# -*- coding: utf-8 -*-
# Part of Odoo. See LICENSE file for full copyright and licensing details.
from odoo import api, fields, models
from odoo.addons.rating.models import rating_data
from odoo.osv import expression
from odoo.tools.float_utils import float_compare, float_round
class RatingMixin(models.AbstractModel):
"""This mixin adds rating statistics to mail.thread that already support ratings."""
_name = 'rating.mixin'
_description = "Rating Mixin"
_inherit = 'mail.thread'
rating_last_value = fields.Float('Rating Last Value', groups='base.group_user', compute='_compute_rating_last_value', compute_sudo=True, store=True)
rating_last_feedback = fields.Text('Rating Last Feedback', groups='base.group_user', related='rating_ids.feedback')
rating_last_image = fields.Binary('Rating Last Image', groups='base.group_user', related='rating_ids.rating_image')
rating_count = fields.Integer('Rating count', compute="_compute_rating_stats", compute_sudo=True)
rating_avg = fields.Float("Average Rating", groups='base.group_user',
compute='_compute_rating_stats', compute_sudo=True, search='_search_rating_avg')
rating_avg_text = fields.Selection(rating_data.RATING_TEXT, groups='base.group_user',
compute='_compute_rating_avg_text', compute_sudo=True)
rating_percentage_satisfaction = fields.Float("Rating Satisfaction", compute='_compute_rating_satisfaction', compute_sudo=True)
rating_last_text = fields.Selection(string="Rating Text", groups='base.group_user', related="rating_ids.rating_text")
@api.depends('rating_ids', 'rating_ids.rating', 'rating_ids.consumed')
def _compute_rating_last_value(self):
# Pure SQL instead of calling read_group to allow ordering array_agg
self.flush_model(['rating_ids'])
self.env['rating.rating'].flush_model(['consumed', 'rating'])
if not self.ids:
self.rating_last_value = 0
return
self.env.cr.execute("""
SELECT
array_agg(rating ORDER BY write_date DESC, id DESC) AS "ratings",
res_id as res_id
FROM "rating_rating"
WHERE
res_model = %s
AND res_id in %s
AND consumed = true
GROUP BY res_id""", [self._name, tuple(self.ids)])
read_group_raw = self.env.cr.dictfetchall()
rating_by_res_id = {e['res_id']: e['ratings'][0] for e in read_group_raw}
for record in self:
record.rating_last_value = rating_by_res_id.get(record.id, 0)
@api.depends('rating_ids.res_id', 'rating_ids.rating')
def _compute_rating_stats(self):
""" Compute avg and count in one query, as thoses fields will be used together most of the time. """
domain = expression.AND([self._rating_domain(), [('rating', '>=', rating_data.RATING_LIMIT_MIN)]])
read_group_res = self.env['rating.rating']._read_group(domain, ['res_id'], aggregates=['__count', 'rating:avg']) # force average on rating column
mapping = {res_id: {'rating_count': count, 'rating_avg': rating_avg} for res_id, count, rating_avg in read_group_res}
for record in self:
record.rating_count = mapping.get(record.id, {}).get('rating_count', 0)
record.rating_avg = mapping.get(record.id, {}).get('rating_avg', 0)
def _search_rating_avg(self, operator, value):
if operator not in rating_data.OPERATOR_MAPPING:
raise NotImplementedError('This operator %s is not supported in this search method.' % operator)
rating_read_group = self.env['rating.rating'].sudo()._read_group(
[('res_model', '=', self._name), ('consumed', '=', True), ('rating', '>=', rating_data.RATING_LIMIT_MIN)],
['res_id'], ['rating:avg'])
res_ids = [
res_id
for res_id, rating_avg in rating_read_group
if rating_data.OPERATOR_MAPPING[operator](float_compare(rating_avg, value, 2), 0)
]
return [('id', 'in', res_ids)]
@api.depends('rating_avg')
def _compute_rating_avg_text(self):
for record in self:
record.rating_avg_text = rating_data._rating_avg_to_text(record.rating_avg)
@api.depends('rating_ids.res_id', 'rating_ids.rating')
def _compute_rating_satisfaction(self):
""" Compute the rating satisfaction percentage, this is done separately from rating_count and rating_avg
since the query is different, to avoid computing if it is not necessary"""
domain = expression.AND([self._rating_domain(), [('rating', '>=', rating_data.RATING_LIMIT_MIN)]])
# See `_compute_rating_percentage_satisfaction` above
read_group_res = self.env['rating.rating']._read_group(domain, ['res_id', 'rating'], aggregates=['__count'])
default_grades = {'great': 0, 'okay': 0, 'bad': 0}
grades_per_record = {record_id: default_grades.copy() for record_id in self.ids}
for record_id, rating, count in read_group_res:
grade = rating_data._rating_to_grade(rating)
grades_per_record[record_id][grade] += count
for record in self:
grade_repartition = grades_per_record.get(record.id, default_grades)
grade_count = sum(grade_repartition.values())
record.rating_percentage_satisfaction = grade_repartition['great'] * 100 / grade_count if grade_count else -1
def write(self, values):
""" If the rated ressource name is modified, we should update the rating res_name too.
If the rated ressource parent is changed we should update the parent_res_id too"""
result = super(RatingMixin, self).write(values)
for record in self:
if record._rec_name in values: # set the res_name of ratings to be recomputed
res_name_field = self.env['rating.rating']._fields['res_name']
self.env.add_to_compute(res_name_field, record.rating_ids)
if record._rating_get_parent_field_name() in values:
record.rating_ids.sudo().write({'parent_res_id': record[record._rating_get_parent_field_name()].id})
return result
def _rating_get_parent_field_name(self):
"""Return the parent relation field name. Should return a Many2One"""
return None
def _rating_domain(self):
""" Returns a normalized domain on rating.rating to select the records to
include in count, avg, ... computation of current model.
"""
return ['&', '&', ('res_model', '=', self._name), ('res_id', 'in', self.ids), ('consumed', '=', True)]
def _rating_get_repartition(self, add_stats=False, domain=None):
""" get the repatition of rating grade for the given res_ids.
:param add_stats : flag to add stat to the result
:type add_stats : boolean
:param domain : optional extra domain of the rating to include/exclude in repartition
:return dictionnary
if not add_stats, the dict is like
- key is the rating value (integer)
- value is the number of object (res_model, res_id) having the value
otherwise, key is the value of the information (string) : either stat name (avg, total, ...) or 'repartition'
containing the same dict if add_stats was False.
"""
base_domain = expression.AND([self._rating_domain(), [('rating', '>=', 1)]])
if domain:
base_domain += domain
rg_data = self.env['rating.rating']._read_group(base_domain, ['rating'], ['__count'])
# init dict with all possible rate value, except 0 (no value for the rating)
values = dict.fromkeys(range(1, 6), 0)
for rating, count in rg_data:
rating_val_round = float_round(rating, precision_digits=1)
values[rating_val_round] = values.get(rating_val_round, 0) + count
# add other stats
if add_stats:
rating_number = sum(values.values())
return {
'repartition': values,
'avg': sum(float(key * values[key]) for key in values) / rating_number if rating_number > 0 else 0,
'total': sum(count for __, count in rg_data),
}
return values
def rating_get_grades(self, domain=None):
""" get the repatition of rating grade for the given res_ids.
:param domain : optional domain of the rating to include/exclude in grades computation
:return dictionnary where the key is the grade (great, okay, bad), and the value, the number of object (res_model, res_id) having the grade
the grade are compute as 0-30% : Bad
31-69%: Okay
70-100%: Great
"""
data = self._rating_get_repartition(domain=domain)
res = dict.fromkeys(['great', 'okay', 'bad'], 0)
for key in data:
grade = rating_data._rating_to_grade(key)
res[grade] += data[key]
return res
def rating_get_stats(self, domain=None):
""" get the statistics of the rating repatition
:param domain : optional domain of the rating to include/exclude in statistic computation
:return dictionnary where
- key is the name of the information (stat name)
- value is statistic value : 'percent' contains the repartition in percentage, 'avg' is the average rate
and 'total' is the number of rating
"""
data = self._rating_get_repartition(domain=domain, add_stats=True)
result = {
'avg': data['avg'],
'total': data['total'],
'percent': dict.fromkeys(range(1, 6), 0),
}
for rate in data['repartition']:
result['percent'][rate] = (data['repartition'][rate] * 100) / data['total'] if data['total'] > 0 else 0
return result