264a9af7bf
Changelog: fixed
2707 lines
132 KiB
Python
2707 lines
132 KiB
Python
# -*- coding: utf-8 -*-
|
|
# Part of Odoo. See LICENSE file for full copyright and licensing details.
|
|
|
|
import logging
|
|
import pytz
|
|
import threading
|
|
from ast import literal_eval
|
|
from collections import OrderedDict, defaultdict
|
|
from datetime import date, datetime, timedelta
|
|
from markupsafe import Markup
|
|
from psycopg2 import sql
|
|
|
|
from odoo import api, fields, models, tools, SUPERUSER_ID
|
|
from odoo.addons.mail.tools import mail_validation
|
|
from odoo.addons.phone_validation.tools import phone_validation
|
|
from odoo.exceptions import UserError, AccessError
|
|
from odoo.osv import expression
|
|
from odoo.tools.translate import _
|
|
from odoo.tools import date_utils, email_split, is_html_empty, groupby, parse_contact_from_email
|
|
from odoo.tools.misc import get_lang
|
|
|
|
from . import crm_stage
|
|
|
|
_logger = logging.getLogger(__name__)
|
|
_schema = logging.getLogger('odoo.schema')
|
|
|
|
|
|
CRM_LEAD_FIELDS_TO_MERGE = [
|
|
# UTM mixin
|
|
'campaign_id',
|
|
'medium_id',
|
|
'source_id',
|
|
# Mail mixin
|
|
'email_cc',
|
|
# description
|
|
'name',
|
|
'user_id',
|
|
'color',
|
|
'company_id',
|
|
'lang_id',
|
|
'team_id',
|
|
'referred',
|
|
# pipeline
|
|
'stage_id',
|
|
# revenues
|
|
'expected_revenue',
|
|
'recurring_plan',
|
|
'recurring_revenue',
|
|
# dates
|
|
'create_date',
|
|
'date_automation_last',
|
|
'date_deadline',
|
|
# partner / contact
|
|
'partner_id',
|
|
'title',
|
|
'partner_name',
|
|
'contact_name',
|
|
'email_from',
|
|
'function',
|
|
'mobile',
|
|
'phone',
|
|
'website',
|
|
]
|
|
|
|
# Subset of partner fields: sync any of those
|
|
PARTNER_FIELDS_TO_SYNC = [
|
|
'lang',
|
|
'mobile',
|
|
'title',
|
|
'function',
|
|
'website',
|
|
]
|
|
|
|
# Subset of partner fields: sync all or none to avoid mixed addresses
|
|
PARTNER_ADDRESS_FIELDS_TO_SYNC = [
|
|
'street',
|
|
'street2',
|
|
'city',
|
|
'zip',
|
|
'state_id',
|
|
'country_id',
|
|
]
|
|
|
|
# Those values have been determined based on benchmark to minimise
|
|
# computation time, number of transaction and transaction time.
|
|
PLS_COMPUTE_BATCH_STEP = 50000 # odoo.models.PREFETCH_MAX = 1000 but larger cluster can speed up global computation
|
|
PLS_UPDATE_BATCH_STEP = 5000
|
|
|
|
|
|
class Lead(models.Model):
|
|
_name = "crm.lead"
|
|
_description = "Lead/Opportunity"
|
|
_order = "priority desc, id desc"
|
|
_inherit = ['mail.thread.cc',
|
|
'mail.thread.blacklist',
|
|
'mail.thread.phone',
|
|
'mail.activity.mixin',
|
|
'utm.mixin',
|
|
'format.address.mixin',
|
|
'mail.tracking.duration.mixin',
|
|
]
|
|
_primary_email = 'email_from'
|
|
_check_company_auto = True
|
|
_track_duration_field = 'stage_id'
|
|
|
|
# Description
|
|
name = fields.Char(
|
|
'Opportunity', index='trigram', required=True,
|
|
compute='_compute_name', readonly=False, store=True)
|
|
user_id = fields.Many2one(
|
|
'res.users', string='Salesperson', default=lambda self: self.env.user,
|
|
domain="[('share', '=', False)]",
|
|
check_company=True, index=True, tracking=True)
|
|
user_company_ids = fields.Many2many(
|
|
'res.company', compute='_compute_user_company_ids',
|
|
help='UX: Limit to lead company or all if no company')
|
|
team_id = fields.Many2one(
|
|
'crm.team', string='Sales Team', check_company=True, index=True, tracking=True,
|
|
compute='_compute_team_id', ondelete="set null", readonly=False, store=True, precompute=True)
|
|
lead_properties = fields.Properties(
|
|
'Properties', definition='team_id.lead_properties_definition',
|
|
copy=True)
|
|
company_id = fields.Many2one(
|
|
'res.company', string='Company', index=True,
|
|
compute='_compute_company_id', readonly=False, store=True)
|
|
referred = fields.Char('Referred By')
|
|
description = fields.Html('Notes')
|
|
active = fields.Boolean('Active', default=True, tracking=True)
|
|
type = fields.Selection([
|
|
('lead', 'Lead'), ('opportunity', 'Opportunity')], required=True, tracking=15, index=True,
|
|
default=lambda self: 'lead' if self.env['res.users'].has_group('crm.group_use_lead') else 'opportunity')
|
|
# Pipeline management
|
|
priority = fields.Selection(
|
|
crm_stage.AVAILABLE_PRIORITIES, string='Priority', index=True,
|
|
default=crm_stage.AVAILABLE_PRIORITIES[0][0])
|
|
stage_id = fields.Many2one(
|
|
'crm.stage', string='Stage', index=True, tracking=True,
|
|
compute='_compute_stage_id', readonly=False, store=True,
|
|
copy=False, group_expand='_read_group_stage_ids', ondelete='restrict',
|
|
domain="['|', ('team_id', '=', False), ('team_id', '=', team_id)]")
|
|
kanban_state = fields.Selection([
|
|
('grey', 'No next activity planned'),
|
|
('red', 'Next activity late'),
|
|
('green', 'Next activity is planned')], string='Kanban State',
|
|
compute='_compute_kanban_state')
|
|
tag_ids = fields.Many2many(
|
|
'crm.tag', 'crm_tag_rel', 'lead_id', 'tag_id', string='Tags',
|
|
help="Classify and analyze your lead/opportunity categories like: Training, Service")
|
|
color = fields.Integer('Color Index', default=0)
|
|
# Revenues
|
|
expected_revenue = fields.Monetary('Expected Revenue', currency_field='company_currency', tracking=True)
|
|
prorated_revenue = fields.Monetary('Prorated Revenue', currency_field='company_currency', store=True, compute="_compute_prorated_revenue")
|
|
recurring_revenue = fields.Monetary('Recurring Revenues', currency_field='company_currency', tracking=True)
|
|
recurring_plan = fields.Many2one('crm.recurring.plan', string="Recurring Plan")
|
|
recurring_revenue_monthly = fields.Monetary('Expected MRR', currency_field='company_currency', store=True,
|
|
compute="_compute_recurring_revenue_monthly")
|
|
recurring_revenue_monthly_prorated = fields.Monetary('Prorated MRR', currency_field='company_currency', store=True,
|
|
compute="_compute_recurring_revenue_monthly_prorated")
|
|
recurring_revenue_prorated = fields.Monetary('Prorated Recurring Revenues', currency_field='company_currency',
|
|
compute="_compute_recurring_revenue_prorated", store=True)
|
|
company_currency = fields.Many2one("res.currency", string='Currency', compute="_compute_company_currency", compute_sudo=True)
|
|
# Dates
|
|
date_closed = fields.Datetime('Closed Date', readonly=True, copy=False)
|
|
date_automation_last = fields.Datetime('Last Action', readonly=True)
|
|
date_open = fields.Datetime(
|
|
'Assignment Date', compute='_compute_date_open', readonly=True, store=True)
|
|
day_open = fields.Float('Days to Assign', compute='_compute_day_open', store=True)
|
|
day_close = fields.Float('Days to Close', compute='_compute_day_close', store=True)
|
|
date_last_stage_update = fields.Datetime(
|
|
'Last Stage Update', compute='_compute_date_last_stage_update', index=True, readonly=True, store=True)
|
|
date_conversion = fields.Datetime('Conversion Date', readonly=True)
|
|
date_deadline = fields.Date('Expected Closing', help="Estimate of the date on which the opportunity will be won.")
|
|
# Customer / contact
|
|
partner_id = fields.Many2one(
|
|
'res.partner', string='Customer', check_company=True, index=True, tracking=10,
|
|
help="Linked partner (optional). Usually created when converting the lead. You can find a partner by its Name, TIN, Email or Internal Reference.")
|
|
partner_is_blacklisted = fields.Boolean('Partner is blacklisted', related='partner_id.is_blacklisted', readonly=True)
|
|
contact_name = fields.Char(
|
|
'Contact Name', index='trigram', tracking=30,
|
|
compute='_compute_contact_name', readonly=False, store=True)
|
|
partner_name = fields.Char(
|
|
'Company Name', index='trigram', tracking=20,
|
|
compute='_compute_partner_name', readonly=False, store=True,
|
|
help='The name of the future partner company that will be created while converting the lead into opportunity')
|
|
function = fields.Char('Job Position', compute='_compute_function', readonly=False, store=True)
|
|
title = fields.Many2one('res.partner.title', string='Title', compute='_compute_title', readonly=False, store=True)
|
|
email_from = fields.Char(
|
|
'Email', tracking=40, index='trigram',
|
|
compute='_compute_email_from', inverse='_inverse_email_from', readonly=False, store=True)
|
|
email_normalized = fields.Char(index='trigram') # inherited via mail.thread.blacklist
|
|
phone = fields.Char(
|
|
'Phone', tracking=50,
|
|
compute='_compute_phone', inverse='_inverse_phone', readonly=False, store=True)
|
|
mobile = fields.Char('Mobile', compute='_compute_mobile', readonly=False, store=True)
|
|
phone_sanitized = fields.Char(index='btree_not_null') # inherited via mail.thread.phone
|
|
phone_state = fields.Selection([
|
|
('correct', 'Correct'),
|
|
('incorrect', 'Incorrect')], string='Phone Quality', compute="_compute_phone_state", store=True)
|
|
email_state = fields.Selection([
|
|
('correct', 'Correct'),
|
|
('incorrect', 'Incorrect')], string='Email Quality', compute="_compute_email_state", store=True)
|
|
website = fields.Char('Website', help="Website of the contact", compute="_compute_website", readonly=False, store=True)
|
|
lang_id = fields.Many2one(
|
|
'res.lang', string='Language',
|
|
compute='_compute_lang_id', readonly=False, store=True)
|
|
lang_code = fields.Char(related='lang_id.code')
|
|
lang_active_count = fields.Integer(compute='_compute_lang_active_count')
|
|
# Address fields
|
|
street = fields.Char('Street', compute='_compute_partner_address_values', readonly=False, store=True)
|
|
street2 = fields.Char('Street2', compute='_compute_partner_address_values', readonly=False, store=True)
|
|
zip = fields.Char('Zip', change_default=True, compute='_compute_partner_address_values', readonly=False, store=True)
|
|
city = fields.Char('City', compute='_compute_partner_address_values', readonly=False, store=True)
|
|
state_id = fields.Many2one(
|
|
"res.country.state", string='State',
|
|
compute='_compute_partner_address_values', readonly=False, store=True,
|
|
domain="[('country_id', '=?', country_id)]")
|
|
country_id = fields.Many2one(
|
|
'res.country', string='Country',
|
|
compute='_compute_partner_address_values', readonly=False, store=True)
|
|
# Probability (Opportunity only)
|
|
probability = fields.Float(
|
|
'Probability', group_operator="avg", copy=False,
|
|
compute='_compute_probabilities', readonly=False, store=True)
|
|
automated_probability = fields.Float('Automated Probability', compute='_compute_probabilities', readonly=True, store=True)
|
|
is_automated_probability = fields.Boolean('Is automated probability?', compute="_compute_is_automated_probability")
|
|
# Won/Lost
|
|
lost_reason_id = fields.Many2one(
|
|
'crm.lost.reason', string='Lost Reason',
|
|
index=True, ondelete='restrict', tracking=True)
|
|
# Statistics
|
|
calendar_event_ids = fields.One2many('calendar.event', 'opportunity_id', string='Meetings')
|
|
duplicate_lead_ids = fields.Many2many("crm.lead", compute="_compute_potential_lead_duplicates", string="Potential Duplicate Lead", context={"active_test": False})
|
|
duplicate_lead_count = fields.Integer(compute="_compute_potential_lead_duplicates", string="Potential Duplicate Lead Count")
|
|
meeting_display_date = fields.Date(compute="_compute_meeting_display")
|
|
meeting_display_label = fields.Char(compute="_compute_meeting_display")
|
|
# UX
|
|
partner_email_update = fields.Boolean('Partner Email will Update', compute='_compute_partner_email_update')
|
|
partner_phone_update = fields.Boolean('Partner Phone will Update', compute='_compute_partner_phone_update')
|
|
is_partner_visible = fields.Boolean('Is Partner Visible', compute='_compute_is_partner_visible')
|
|
# UTMs - enforcing the fact that we want to 'set null' when relation is unlinked
|
|
campaign_id = fields.Many2one(ondelete='set null')
|
|
medium_id = fields.Many2one(ondelete='set null')
|
|
source_id = fields.Many2one(ondelete='set null')
|
|
|
|
_sql_constraints = [
|
|
('check_probability', 'check(probability >= 0 and probability <= 100)', 'The probability of closing the deal should be between 0% and 100%!')
|
|
]
|
|
|
|
@api.depends('activity_date_deadline')
|
|
def _compute_kanban_state(self):
|
|
today = date.today()
|
|
for lead in self:
|
|
kanban_state = 'grey'
|
|
if lead.activity_date_deadline:
|
|
lead_date = fields.Date.from_string(lead.activity_date_deadline)
|
|
if lead_date >= today:
|
|
kanban_state = 'green'
|
|
else:
|
|
kanban_state = 'red'
|
|
lead.kanban_state = kanban_state
|
|
|
|
@api.depends('company_id')
|
|
def _compute_user_company_ids(self):
|
|
all_companies = self.env['res.company'].search([])
|
|
for lead in self:
|
|
if not lead.company_id:
|
|
lead.user_company_ids = all_companies
|
|
else:
|
|
lead.user_company_ids = lead.company_id
|
|
|
|
@api.depends('company_id')
|
|
def _compute_company_currency(self):
|
|
for lead in self:
|
|
if not lead.company_id:
|
|
lead.company_currency = self.env.company.currency_id
|
|
else:
|
|
lead.company_currency = lead.company_id.currency_id
|
|
|
|
@api.depends('user_id', 'type')
|
|
def _compute_team_id(self):
|
|
""" When changing the user, also set a team_id or restrict team id
|
|
to the ones user_id is member of. """
|
|
for lead in self:
|
|
# setting user as void should not trigger a new team computation
|
|
if not lead.user_id:
|
|
continue
|
|
user = lead.user_id
|
|
if lead.team_id and user in (lead.team_id.member_ids | lead.team_id.user_id):
|
|
continue
|
|
team_domain = [('use_leads', '=', True)] if lead.type == 'lead' else [('use_opportunities', '=', True)]
|
|
team = self.env['crm.team']._get_default_team_id(user_id=user.id, domain=team_domain)
|
|
if lead.team_id != team:
|
|
lead.team_id = team.id
|
|
|
|
@api.depends('user_id', 'team_id', 'partner_id')
|
|
def _compute_company_id(self):
|
|
""" Compute company_id coherency. """
|
|
for lead in self:
|
|
proposal = lead.company_id
|
|
|
|
# invalidate wrong configuration
|
|
if proposal:
|
|
# company not in responsible companies
|
|
if lead.user_id and proposal not in lead.user_id.company_ids:
|
|
proposal = False
|
|
# inconsistent
|
|
elif lead.team_id.company_id and proposal != lead.team_id.company_id:
|
|
proposal = False
|
|
# void company on team and no assignee
|
|
elif lead.team_id and not lead.team_id.company_id and not lead.user_id:
|
|
proposal = False
|
|
# no user and no team -> void company and let assignment do its job
|
|
# unless customer has a company
|
|
elif not lead.team_id and not lead.user_id and \
|
|
(not lead.partner_id or lead.partner_id.company_id != proposal):
|
|
proposal = False
|
|
|
|
# propose a new company based on team > user (respecting context) > partner
|
|
if not proposal:
|
|
if lead.team_id.company_id:
|
|
lead.company_id = lead.team_id.company_id
|
|
elif lead.user_id:
|
|
if self.env.company in lead.user_id.company_ids:
|
|
lead.company_id = self.env.company
|
|
else:
|
|
lead.company_id = lead.user_id.company_id & self.env.companies
|
|
elif lead.partner_id:
|
|
lead.company_id = lead.partner_id.company_id
|
|
else:
|
|
lead.company_id = False
|
|
|
|
@api.depends('team_id', 'type')
|
|
def _compute_stage_id(self):
|
|
for lead in self:
|
|
if not lead.stage_id:
|
|
lead.stage_id = lead._stage_find(domain=[('fold', '=', False)]).id
|
|
|
|
@api.depends('user_id')
|
|
def _compute_date_open(self):
|
|
for lead in self:
|
|
if not lead.date_open and lead.user_id:
|
|
lead.date_open = self.env.cr.now()
|
|
|
|
@api.depends('stage_id')
|
|
def _compute_date_last_stage_update(self):
|
|
for lead in self:
|
|
if not lead.date_last_stage_update:
|
|
lead.date_last_stage_update = self.env.cr.now()
|
|
|
|
@api.depends('create_date', 'date_open')
|
|
def _compute_day_open(self):
|
|
""" Compute difference between create date and open date """
|
|
leads = self.filtered(lambda l: l.date_open and l.create_date)
|
|
others = self - leads
|
|
others.day_open = None
|
|
for lead in leads:
|
|
date_create = fields.Datetime.from_string(lead.create_date).replace(microsecond=0)
|
|
date_open = fields.Datetime.from_string(lead.date_open)
|
|
lead.day_open = abs((date_open - date_create).days)
|
|
|
|
@api.depends('create_date', 'date_closed')
|
|
def _compute_day_close(self):
|
|
""" Compute difference between current date and log date """
|
|
leads = self.filtered(lambda l: l.date_closed and l.create_date)
|
|
others = self - leads
|
|
others.day_close = None
|
|
for lead in leads:
|
|
date_create = fields.Datetime.from_string(lead.create_date)
|
|
date_close = fields.Datetime.from_string(lead.date_closed)
|
|
lead.day_close = abs((date_close - date_create).days)
|
|
|
|
@api.depends('partner_id')
|
|
def _compute_name(self):
|
|
for lead in self:
|
|
if not lead.name and lead.partner_id and lead.partner_id.name:
|
|
lead.name = _("%s's opportunity") % lead.partner_id.name
|
|
|
|
@api.depends('partner_id')
|
|
def _compute_contact_name(self):
|
|
""" compute the new values when partner_id has changed """
|
|
for lead in self:
|
|
lead.update(lead._prepare_contact_name_from_partner(lead.partner_id))
|
|
|
|
@api.depends('partner_id')
|
|
def _compute_partner_name(self):
|
|
""" compute the new values when partner_id has changed """
|
|
for lead in self:
|
|
lead.update(lead._prepare_partner_name_from_partner(lead.partner_id))
|
|
|
|
@api.depends('partner_id')
|
|
def _compute_function(self):
|
|
""" compute the new values when partner_id has changed """
|
|
for lead in self:
|
|
if not lead.function or lead.partner_id.function:
|
|
lead.function = lead.partner_id.function
|
|
|
|
@api.depends('partner_id')
|
|
def _compute_title(self):
|
|
""" compute the new values when partner_id has changed """
|
|
for lead in self:
|
|
if not lead.title or lead.partner_id.title:
|
|
lead.title = lead.partner_id.title
|
|
|
|
@api.depends('partner_id')
|
|
def _compute_mobile(self):
|
|
""" compute the new values when partner_id has changed """
|
|
for lead in self:
|
|
if not lead.mobile or lead.partner_id.mobile:
|
|
lead.mobile = lead.partner_id.mobile
|
|
|
|
@api.depends('partner_id')
|
|
def _compute_website(self):
|
|
""" compute the new values when partner_id has changed """
|
|
for lead in self:
|
|
if not lead.website or lead.partner_id.website:
|
|
lead.website = lead.partner_id.website
|
|
|
|
@api.depends('partner_id')
|
|
def _compute_lang_id(self):
|
|
""" compute the lang based on partner, erase any value to force the partner
|
|
one if set. """
|
|
# prepare cache
|
|
lang_codes = [code for code in self.mapped('partner_id.lang') if code]
|
|
if lang_codes:
|
|
lang_id_by_code = dict(
|
|
(code, self.env['res.lang']._lang_get_id(code))
|
|
for code in lang_codes
|
|
)
|
|
else:
|
|
lang_id_by_code = {}
|
|
for lead in self.filtered('partner_id'):
|
|
lead.lang_id = lang_id_by_code.get(lead.partner_id.lang, False)
|
|
|
|
@api.depends('lang_id')
|
|
def _compute_lang_active_count(self):
|
|
self.lang_active_count = len(self.env['res.lang'].get_installed())
|
|
|
|
@api.depends('partner_id')
|
|
def _compute_partner_address_values(self):
|
|
""" Sync all or none of address fields """
|
|
for lead in self:
|
|
lead.update(lead._prepare_address_values_from_partner(lead.partner_id))
|
|
|
|
@api.depends('partner_id.email')
|
|
def _compute_email_from(self):
|
|
for lead in self:
|
|
if lead.partner_id.email and lead._get_partner_email_update():
|
|
lead.email_from = lead.partner_id.email
|
|
|
|
def _inverse_email_from(self):
|
|
for lead in self:
|
|
if lead._get_partner_email_update():
|
|
lead.partner_id.email = lead.email_from
|
|
|
|
@api.depends('partner_id.phone')
|
|
def _compute_phone(self):
|
|
for lead in self:
|
|
if lead.partner_id.phone and lead._get_partner_phone_update():
|
|
lead.phone = lead.partner_id.phone
|
|
|
|
def _inverse_phone(self):
|
|
for lead in self:
|
|
if lead._get_partner_phone_update():
|
|
lead.partner_id.phone = lead.phone
|
|
|
|
@api.depends('phone', 'country_id.code')
|
|
def _compute_phone_state(self):
|
|
for lead in self:
|
|
phone_status = False
|
|
if lead.phone:
|
|
country_code = lead.country_id.code if lead.country_id and lead.country_id.code else None
|
|
try:
|
|
if phone_validation.phone_parse(lead.phone, country_code): # otherwise library not installed
|
|
phone_status = 'correct'
|
|
except UserError:
|
|
phone_status = 'incorrect'
|
|
lead.phone_state = phone_status
|
|
|
|
@api.depends('email_from')
|
|
def _compute_email_state(self):
|
|
for lead in self:
|
|
email_state = False
|
|
if lead.email_from:
|
|
email_state = 'incorrect'
|
|
for email in email_split(lead.email_from):
|
|
if mail_validation.mail_validate(email):
|
|
email_state = 'correct'
|
|
break
|
|
lead.email_state = email_state
|
|
|
|
@api.depends('probability', 'automated_probability')
|
|
def _compute_is_automated_probability(self):
|
|
""" If probability and automated_probability are equal probability computation
|
|
is considered as automatic, aka probability is sync with automated_probability """
|
|
for lead in self:
|
|
lead.is_automated_probability = tools.float_compare(lead.probability, lead.automated_probability, 2) == 0
|
|
|
|
@api.depends(lambda self: ['stage_id', 'team_id'] + self._pls_get_safe_fields())
|
|
def _compute_probabilities(self):
|
|
lead_probabilities = self._pls_get_naive_bayes_probabilities()
|
|
for lead in self:
|
|
if lead.id in lead_probabilities:
|
|
was_automated = lead.active and lead.is_automated_probability
|
|
lead.automated_probability = lead_probabilities[lead.id]
|
|
if was_automated:
|
|
lead.probability = lead.automated_probability
|
|
|
|
@api.depends('expected_revenue', 'probability')
|
|
def _compute_prorated_revenue(self):
|
|
for lead in self:
|
|
lead.prorated_revenue = round((lead.expected_revenue or 0.0) * (lead.probability or 0) / 100.0, 2)
|
|
|
|
@api.depends('recurring_revenue', 'recurring_plan.number_of_months')
|
|
def _compute_recurring_revenue_monthly(self):
|
|
for lead in self:
|
|
lead.recurring_revenue_monthly = (lead.recurring_revenue or 0.0) / (lead.recurring_plan.number_of_months or 1)
|
|
|
|
@api.depends('recurring_revenue_monthly', 'probability')
|
|
def _compute_recurring_revenue_monthly_prorated(self):
|
|
for lead in self:
|
|
lead.recurring_revenue_monthly_prorated = (lead.recurring_revenue_monthly or 0.0) * (lead.probability or 0) / 100.0
|
|
|
|
@api.depends('recurring_revenue', 'probability')
|
|
def _compute_recurring_revenue_prorated(self):
|
|
for lead in self:
|
|
lead.recurring_revenue_prorated = (lead.recurring_revenue or 0.0) * (lead.probability or 0) / 100.0
|
|
|
|
@api.depends('calendar_event_ids', 'calendar_event_ids.start')
|
|
def _compute_meeting_display(self):
|
|
now = fields.Datetime.now()
|
|
meeting_data = self.env['calendar.event'].sudo()._read_group([
|
|
('opportunity_id', 'in', self.ids),
|
|
], ['opportunity_id'], ['start:array_agg', 'start:max'])
|
|
mapped_data = {
|
|
lead: {
|
|
'last_meeting_date': last_meeting_date,
|
|
'next_meeting_date': min([dt for dt in meeting_start_dates if dt > now] or [False]),
|
|
} for lead, meeting_start_dates, last_meeting_date in meeting_data
|
|
}
|
|
for lead in self:
|
|
lead_meeting_info = mapped_data.get(lead)
|
|
if not lead_meeting_info:
|
|
lead.meeting_display_date = False
|
|
lead.meeting_display_label = _('No Meeting')
|
|
elif lead_meeting_info['next_meeting_date']:
|
|
lead.meeting_display_date = lead_meeting_info['next_meeting_date']
|
|
lead.meeting_display_label = _('Next Meeting')
|
|
else:
|
|
lead.meeting_display_date = lead_meeting_info['last_meeting_date']
|
|
lead.meeting_display_label = _('Last Meeting')
|
|
|
|
@api.depends('email_normalized', 'partner_id',
|
|
'phone_sanitized')
|
|
def _compute_potential_lead_duplicates(self):
|
|
""" Override potential lead duplicates computation to be more efficient
|
|
with high lead volume.
|
|
Criterions:
|
|
* email domain exact match;
|
|
* phone_sanitized exact match;
|
|
* same commercial entity;
|
|
"""
|
|
SEARCH_RESULT_LIMIT = 21
|
|
|
|
def return_if_relevant(model_name, domain):
|
|
""" Returns the recordset obtained by performing a search on the provided
|
|
model with the provided domain if the cardinality of that recordset is
|
|
below a given threshold (i.e: `SEARCH_RESULT_LIMIT`). Otherwise, returns
|
|
an empty recordset of the provided model as it indicates search term
|
|
was not relevant.
|
|
Note: The function will use the administrator privileges to guarantee
|
|
that a maximum amount of leads will be included in the search results
|
|
and transcend multi-company record rules. It also includes archived
|
|
records. Idea is that counter indicates duplicates are present and
|
|
the lead could be escalated to managers.
|
|
"""
|
|
model = self.env[model_name].sudo().with_context(active_test=False)
|
|
res = model.search(domain, limit=SEARCH_RESULT_LIMIT)
|
|
return res if len(res) < SEARCH_RESULT_LIMIT else model
|
|
|
|
for lead in self:
|
|
lead_id = lead._origin.id if isinstance(lead.id, models.NewId) else lead.id
|
|
common_lead_domain = [
|
|
('id', '!=', lead_id)
|
|
]
|
|
|
|
duplicate_lead_ids = self.env['crm.lead']
|
|
|
|
# check for "same commercial entity" duplicates
|
|
if lead.partner_id and lead.partner_id.commercial_partner_id:
|
|
duplicate_lead_ids |= lead.with_context(active_test=False).search(common_lead_domain + [
|
|
("partner_id", "child_of", lead.partner_id.commercial_partner_id.ids)
|
|
])
|
|
# check the phone number duplicates, based on phone_sanitized. Only
|
|
# exact matches are found, and the single one stored in phone_sanitized
|
|
# in case phone and mobile are both set.
|
|
if lead.phone_sanitized:
|
|
duplicate_lead_ids |= return_if_relevant('crm.lead', common_lead_domain + [
|
|
('phone_sanitized', '=', lead.phone_sanitized)
|
|
])
|
|
|
|
lead.duplicate_lead_ids = duplicate_lead_ids + lead
|
|
lead.duplicate_lead_count = len(duplicate_lead_ids)
|
|
|
|
@api.depends('email_from', 'partner_id')
|
|
def _compute_partner_email_update(self):
|
|
for lead in self:
|
|
lead.partner_email_update = lead._get_partner_email_update()
|
|
|
|
@api.depends('phone', 'partner_id')
|
|
def _compute_partner_phone_update(self):
|
|
for lead in self:
|
|
lead.partner_phone_update = lead._get_partner_phone_update()
|
|
|
|
@api.depends_context('uid')
|
|
@api.depends('partner_id', 'type')
|
|
def _compute_is_partner_visible(self):
|
|
""" When the crm.lead is of type 'lead', we don't want to display the "Customer" field on the form view
|
|
unless it's set (or debug mode).
|
|
|
|
Indeed, most of the times leads will not have this information set, since when we assign a Customer we
|
|
usually convert the lead to an opportunity as well.
|
|
|
|
This means that on the lead form, we don't want to display this field since it may be misleading for the
|
|
end user.
|
|
When it's set however, we want to display it, mainly because there are a few automatic synchronizations between
|
|
the lead and its partner (phone and email for examples), and this needs to be clear that modifying
|
|
one of those fields will in turn modify the linked partner."""
|
|
is_debug_mode = self.user_has_groups('base.group_no_one')
|
|
for lead in self:
|
|
lead.is_partner_visible = bool(lead.type == 'opportunity' or lead.partner_id or is_debug_mode)
|
|
|
|
@api.onchange('phone', 'country_id', 'company_id')
|
|
def _onchange_phone_validation(self):
|
|
if self.phone:
|
|
self.phone = self._phone_format(fname='phone', force_format='INTERNATIONAL') or self.phone
|
|
|
|
@api.onchange('mobile', 'country_id', 'company_id')
|
|
def _onchange_mobile_validation(self):
|
|
if self.mobile:
|
|
self.mobile = self._phone_format(fname='mobile', force_format='INTERNATIONAL') or self.mobile
|
|
|
|
def _prepare_values_from_partner(self, partner):
|
|
""" Get a dictionary with values coming from partner information to
|
|
copy on a lead. Non-address fields get the current lead
|
|
values to avoid being reset if partner has no value for them. """
|
|
|
|
# Sync all address fields from partner, or none, to avoid mixing them.
|
|
values = self._prepare_address_values_from_partner(partner)
|
|
|
|
# For other fields, get the info from the partner, but only if set
|
|
values.update({f: partner[f] or self[f] for f in PARTNER_FIELDS_TO_SYNC if f != 'lang'})
|
|
if partner.lang:
|
|
values['lang_id'] = self.env['res.lang']._lang_get_id(partner.lang)
|
|
|
|
# Fields with specific logic
|
|
values.update(self._prepare_contact_name_from_partner(partner))
|
|
values.update(self._prepare_partner_name_from_partner(partner))
|
|
|
|
return self._convert_to_write(values)
|
|
|
|
def _prepare_address_values_from_partner(self, partner):
|
|
# Sync all address fields from partner, or none, to avoid mixing them.
|
|
if any(partner[f] for f in PARTNER_ADDRESS_FIELDS_TO_SYNC):
|
|
values = {f: partner[f] for f in PARTNER_ADDRESS_FIELDS_TO_SYNC}
|
|
else:
|
|
values = {f: self[f] for f in PARTNER_ADDRESS_FIELDS_TO_SYNC}
|
|
return values
|
|
|
|
def _prepare_contact_name_from_partner(self, partner):
|
|
contact_name = False if partner.is_company else partner.name
|
|
return {'contact_name': contact_name or self.contact_name}
|
|
|
|
def _prepare_partner_name_from_partner(self, partner):
|
|
""" Company name: name of partner parent (if set) or name of partner
|
|
(if company) or company_name of partner (if not a company). """
|
|
partner_name = partner.parent_id.name
|
|
if not partner_name and partner.is_company:
|
|
partner_name = partner.name
|
|
elif not partner_name and partner.company_name:
|
|
partner_name = partner.company_name
|
|
return {'partner_name': partner_name or self.partner_name}
|
|
|
|
def _get_partner_email_update(self):
|
|
"""Calculate if we should write the email on the related partner. When
|
|
the email of the lead / partner is an empty string, we force it to False
|
|
to not propagate a False on an empty string.
|
|
|
|
Done in a separate method so it can be used in both ribbon and inverse
|
|
and compute of email update methods.
|
|
"""
|
|
self.ensure_one()
|
|
if self.partner_id and self.email_from != self.partner_id.email:
|
|
lead_email_normalized = tools.email_normalize(self.email_from) or self.email_from or False
|
|
partner_email_normalized = tools.email_normalize(self.partner_id.email) or self.partner_id.email or False
|
|
return lead_email_normalized != partner_email_normalized
|
|
return False
|
|
|
|
def _get_partner_phone_update(self):
|
|
"""Calculate if we should write the phone on the related partner. When
|
|
the phone of the lead / partner is an empty string, we force it to False
|
|
to not propagate a False on an empty string.
|
|
|
|
Done in a separate method so it can be used in both ribbon and inverse
|
|
and compute of phone update methods.
|
|
"""
|
|
self.ensure_one()
|
|
if self.partner_id and self.phone != self.partner_id.phone:
|
|
lead_phone_formatted = self._phone_format(fname='phone') or self.phone or False
|
|
partner_phone_formatted = self.partner_id._phone_format(fname='phone') or self.partner_id.phone or False
|
|
return lead_phone_formatted != partner_phone_formatted
|
|
return False
|
|
|
|
# ------------------------------------------------------------
|
|
# ORM
|
|
# ------------------------------------------------------------
|
|
|
|
def _auto_init(self):
|
|
super()._auto_init()
|
|
tools.create_index(self._cr, 'crm_lead_user_id_team_id_type_index',
|
|
self._table, ['user_id', 'team_id', 'type'])
|
|
tools.create_index(self._cr, 'crm_lead_create_date_team_id_idx',
|
|
self._table, ['create_date', 'team_id'])
|
|
phone_pattern = r'[\s\\./\(\)\-]'
|
|
for field_name in ('phone', 'mobile'):
|
|
index_name = f'crm_lead_{field_name}_partial_tgm'
|
|
if tools.index_exists(self._cr, index_name):
|
|
continue
|
|
regex_expression = f"regexp_replace(({field_name}::text), %s::text, ''::text, 'g'::text)"
|
|
self._cr.execute(
|
|
f'CREATE INDEX "{index_name}" ON "{self._table}" ({regex_expression}) WHERE {field_name} IS NOT NULL',
|
|
(phone_pattern,)
|
|
)
|
|
_schema.debug("Table %r: created index %r (%s)", self._table, index_name, regex_expression)
|
|
|
|
@api.model_create_multi
|
|
def create(self, vals_list):
|
|
for vals in vals_list:
|
|
if vals.get('website'):
|
|
vals['website'] = self.env['res.partner']._clean_website(vals['website'])
|
|
leads = super(Lead, self).create(vals_list)
|
|
|
|
for lead, values in zip(leads, vals_list):
|
|
if any(field in ['active', 'stage_id'] for field in values):
|
|
lead._handle_won_lost(values)
|
|
|
|
return leads
|
|
|
|
def write(self, vals):
|
|
if vals.get('website'):
|
|
vals['website'] = self.env['res.partner']._clean_website(vals['website'])
|
|
|
|
now = self.env.cr.now()
|
|
stage_updated, stage_is_won = False, False
|
|
# stage change (or reset): update date_last_stage_update if at least one
|
|
# lead does not have the same stage
|
|
if 'stage_id' in vals:
|
|
stage_updated = any(lead.stage_id.id != vals['stage_id'] for lead in self)
|
|
if stage_updated:
|
|
vals['date_last_stage_update'] = now
|
|
if stage_updated and vals.get('stage_id'):
|
|
stage = self.env['crm.stage'].browse(vals['stage_id'])
|
|
if stage.is_won:
|
|
vals.update({'probability': 100, 'automated_probability': 100})
|
|
stage_is_won = True
|
|
# user change; update date_open if at least one lead does not
|
|
# have the same user
|
|
if 'user_id' in vals and not vals.get('user_id'):
|
|
vals['date_open'] = False
|
|
elif vals.get('user_id'):
|
|
user_updated = any(lead.user_id.id != vals['user_id'] for lead in self)
|
|
if user_updated:
|
|
vals['date_open'] = now
|
|
|
|
# stage change with new stage: update probability and date_closed
|
|
if vals.get('probability', 0) >= 100 or not vals.get('active', True):
|
|
vals['date_closed'] = fields.Datetime.now()
|
|
elif vals.get('probability', 0) > 0:
|
|
vals['date_closed'] = False
|
|
elif stage_updated and not stage_is_won and not 'probability' in vals:
|
|
vals['date_closed'] = False
|
|
|
|
if any(field in ['active', 'stage_id'] for field in vals):
|
|
self._handle_won_lost(vals)
|
|
|
|
if not stage_is_won:
|
|
return super(Lead, self).write(vals)
|
|
|
|
# stage change between two won stages: does not change the date_closed
|
|
leads_already_won = self.filtered(lambda lead: lead.stage_id.is_won)
|
|
remaining = self - leads_already_won
|
|
if remaining:
|
|
result = super(Lead, remaining).write(vals)
|
|
if leads_already_won:
|
|
vals.pop('date_closed', False)
|
|
result = super(Lead, leads_already_won).write(vals)
|
|
return result
|
|
|
|
@api.model
|
|
def search_fetch(self, domain, field_names, offset=0, limit=None, order=None):
|
|
""" Override to support ordering on my_activity_date_deadline.
|
|
|
|
Ordering through web client calls search_read() with an order parameter
|
|
set. Method search_read() then calls search_fetch(). Here we override
|
|
search_fetch() to intercept a search with an order on field
|
|
my_activity_date_deadline. In that case we do the search in two steps.
|
|
|
|
First step: fill with deadline-based results
|
|
|
|
* Perform a read_group on my activities to get a mapping lead_id / deadline
|
|
Remember date_deadline is required, we always have a value for it. Only
|
|
the earliest deadline per lead is kept.
|
|
* Search leads linked to those activities that also match the asked domain
|
|
and order from the original search request.
|
|
* Results of that search will be at the top of returned results. Use limit
|
|
None because we have to search all leads linked to activities as ordering
|
|
on deadline is done in post processing.
|
|
* Reorder them according to deadline asc or desc depending on original
|
|
search ordering. Finally take only a subset of those leads to fill with
|
|
results matching asked offset / limit.
|
|
|
|
Second step: fill with other results. If first step does not gives results
|
|
enough to match offset and limit parameters we fill with a search on other
|
|
leads. We keep the asked domain and ordering while filtering out already
|
|
scanned leads to keep a coherent results.
|
|
|
|
All other search and search_read are left untouched by this override to avoid
|
|
side effects. Search_count is not affected by this override.
|
|
"""
|
|
if not order or 'my_activity_date_deadline' not in order:
|
|
return super().search_fetch(domain, field_names, offset, limit, order)
|
|
order_items = [order_item.strip().lower() for order_item in (order or self._order).split(',')]
|
|
|
|
# Perform a read_group on my activities to get a mapping lead_id / deadline
|
|
# Remember date_deadline is required, we always have a value for it. Only
|
|
# the earliest deadline per lead is kept.
|
|
activity_asc = any('my_activity_date_deadline asc' in item for item in order_items)
|
|
my_lead_activities = self.env['mail.activity']._read_group(
|
|
[('res_model', '=', self._name), ('user_id', '=', self.env.uid)],
|
|
['res_id'],
|
|
['date_deadline:min'],
|
|
order='date_deadline:min ASC, res_id',
|
|
)
|
|
my_lead_mapping = dict(my_lead_activities)
|
|
my_lead_ids = list(my_lead_mapping.keys())
|
|
my_lead_domain = expression.AND([[('id', 'in', my_lead_ids)], domain])
|
|
my_lead_order = ', '.join(item for item in order_items if 'my_activity_date_deadline' not in item)
|
|
|
|
# Search leads linked to those activities and order them. See docstring
|
|
# of this method for more details.
|
|
search_res = super().search_fetch(my_lead_domain, field_names, order=my_lead_order)
|
|
my_lead_ids_ordered = sorted(search_res.ids, key=lambda lead_id: my_lead_mapping[lead_id], reverse=not activity_asc)
|
|
# keep only requested window (offset + limit, or offset+)
|
|
my_lead_ids_keep = my_lead_ids_ordered[offset:(offset + limit)] if limit else my_lead_ids_ordered[offset:]
|
|
# keep list of already skipped lead ids to exclude them from future search
|
|
my_lead_ids_skip = my_lead_ids_ordered[:(offset + limit)] if limit else my_lead_ids_ordered
|
|
|
|
# do not go further if limit is achieved
|
|
if limit and len(my_lead_ids_keep) >= limit:
|
|
return self.browse(my_lead_ids_keep)
|
|
|
|
# Fill with remaining leads. If a limit is given, simply remove count of
|
|
# already fetched. Otherwise keep none. If an offset is set we have to
|
|
# reduce it by already fetch results hereabove. Order is updated to exclude
|
|
# my_activity_date_deadline when calling super() .
|
|
lead_limit = (limit - len(my_lead_ids_keep)) if limit else None
|
|
if offset:
|
|
lead_offset = max((offset - len(search_res), 0))
|
|
else:
|
|
lead_offset = 0
|
|
lead_order = ', '.join(item for item in order_items if 'my_activity_date_deadline' not in item)
|
|
|
|
other_lead_res = super().search_fetch(
|
|
expression.AND([[('id', 'not in', my_lead_ids_skip)], domain]),
|
|
field_names, lead_offset, lead_limit, lead_order,
|
|
)
|
|
return self.browse(my_lead_ids_keep) + other_lead_res
|
|
|
|
def _handle_won_lost(self, vals):
|
|
""" This method handle the state changes :
|
|
- To lost : We need to increment corresponding lost count in scoring frequency table
|
|
- To won : We need to increment corresponding won count in scoring frequency table
|
|
- From lost to Won : We need to decrement corresponding lost count + increment corresponding won count
|
|
in scoring frequency table.
|
|
- From won to lost : We need to decrement corresponding won count + increment corresponding lost count
|
|
in scoring frequency table."""
|
|
Lead = self.env['crm.lead']
|
|
leads_reach_won = Lead
|
|
leads_leave_won = Lead
|
|
leads_reach_lost = Lead
|
|
leads_leave_lost = Lead
|
|
won_stage_ids = self.env['crm.stage'].search([('is_won', '=', True)]).ids
|
|
for lead in self:
|
|
if 'stage_id' in vals:
|
|
if vals['stage_id'] in won_stage_ids:
|
|
if lead.probability == 0:
|
|
leads_leave_lost += lead
|
|
leads_reach_won += lead
|
|
elif lead.stage_id.id in won_stage_ids and lead.active: # a lead can be lost at won_stage
|
|
leads_leave_won += lead
|
|
if 'active' in vals:
|
|
if not vals['active'] and lead.active: # archive lead
|
|
if lead.stage_id.id in won_stage_ids and lead not in leads_leave_won:
|
|
leads_leave_won += lead
|
|
leads_reach_lost += lead
|
|
elif vals['active'] and not lead.active: # restore lead
|
|
leads_leave_lost += lead
|
|
|
|
leads_reach_won._pls_increment_frequencies(to_state='won')
|
|
leads_leave_won._pls_increment_frequencies(from_state='won')
|
|
leads_reach_lost._pls_increment_frequencies(to_state='lost')
|
|
leads_leave_lost._pls_increment_frequencies(from_state='lost')
|
|
|
|
@api.returns('self', lambda value: value.id)
|
|
def copy(self, default=None):
|
|
self.ensure_one()
|
|
# set default value in context, if not already set (Put stage to 'new' stage)
|
|
context = dict(self._context)
|
|
context.setdefault('default_type', self.type)
|
|
context.setdefault('default_team_id', self.team_id.id)
|
|
# Set date_open to today if it is an opp
|
|
default = default or {}
|
|
default['date_open'] = self.env.cr.now() if self.type == 'opportunity' else False
|
|
# Do not assign to an archived user
|
|
if not self.user_id.active:
|
|
default['user_id'] = False
|
|
if not self.env.user.has_group('crm.group_use_recurring_revenues'):
|
|
default['recurring_revenue'] = 0
|
|
default['recurring_plan'] = False
|
|
return super(Lead, self.with_context(context)).copy(default=default)
|
|
|
|
def unlink(self):
|
|
""" Update meetings when removing opportunities, otherwise you have
|
|
a link to a record that does not lead anywhere. """
|
|
meetings = self.env['calendar.event'].search([
|
|
('res_id', 'in', self.ids),
|
|
('res_model', '=', self._name),
|
|
])
|
|
if meetings:
|
|
meetings.write({
|
|
'res_id': False,
|
|
'res_model_id': False,
|
|
})
|
|
return super(Lead, self).unlink()
|
|
|
|
@api.model
|
|
def _read_group_stage_ids(self, stages, domain, order):
|
|
# retrieve team_id from the context and write the domain
|
|
# - ('id', 'in', stages.ids): add columns that should be present
|
|
# - OR ('fold', '=', False): add default columns that are not folded
|
|
# - OR ('team_ids', '=', team_id), ('fold', '=', False) if team_id: add team columns that are not folded
|
|
team_id = self._context.get('default_team_id')
|
|
if team_id:
|
|
search_domain = ['|', ('id', 'in', stages.ids), '|', ('team_id', '=', False), ('team_id', '=', team_id)]
|
|
else:
|
|
search_domain = ['|', ('id', 'in', stages.ids), ('team_id', '=', False)]
|
|
|
|
# perform search
|
|
stage_ids = stages._search(search_domain, order=order, access_rights_uid=SUPERUSER_ID)
|
|
return stages.browse(stage_ids)
|
|
|
|
def _stage_find(self, team_id=False, domain=None, order='sequence, id', limit=1):
|
|
""" Determine the stage of the current lead with its teams, the given domain and the given team_id
|
|
:param team_id
|
|
:param domain : base search domain for stage
|
|
:param order : base search order for stage
|
|
:param limit : base search limit for stage
|
|
:returns crm.stage recordset
|
|
"""
|
|
# collect all team_ids by adding given one, and the ones related to the current leads
|
|
team_ids = set()
|
|
if team_id:
|
|
team_ids.add(team_id)
|
|
for lead in self:
|
|
if lead.team_id:
|
|
team_ids.add(lead.team_id.id)
|
|
# generate the domain
|
|
if team_ids:
|
|
search_domain = ['|', ('team_id', '=', False), ('team_id', 'in', list(team_ids))]
|
|
else:
|
|
search_domain = [('team_id', '=', False)]
|
|
# AND with the domain in parameter
|
|
if domain:
|
|
search_domain += list(domain)
|
|
# perform search, return the first found
|
|
return self.env['crm.stage'].search(search_domain, order=order, limit=limit)
|
|
|
|
# ------------------------------------------------------------
|
|
# ACTIONS
|
|
# ------------------------------------------------------------
|
|
|
|
def toggle_active(self):
|
|
""" When archiving: mark probability as 0. When re-activating
|
|
update probability again, for leads and opportunities. """
|
|
res = super(Lead, self).toggle_active()
|
|
activated = self.filtered(lambda lead: lead.active)
|
|
archived = self.filtered(lambda lead: not lead.active)
|
|
if activated:
|
|
activated.write({'lost_reason_id': False})
|
|
activated._compute_probabilities()
|
|
if archived:
|
|
archived.write({'probability': 0, 'automated_probability': 0})
|
|
return res
|
|
|
|
def action_set_lost(self, **additional_values):
|
|
""" Lost semantic: probability = 0 or active = False """
|
|
res = self.action_archive()
|
|
if additional_values:
|
|
self.write(dict(additional_values))
|
|
return res
|
|
|
|
def action_set_won(self):
|
|
""" Won semantic: probability = 100 (active untouched) """
|
|
self.action_unarchive()
|
|
# group the leads by team_id, in order to write once by values couple (each write leads to frequency increment)
|
|
leads_by_won_stage = {}
|
|
for lead in self:
|
|
won_stages = self._stage_find(domain=[('is_won', '=', True)], limit=None)
|
|
# ABD : We could have a mixed pipeline, with "won" stages being separated by "standard"
|
|
# stages. In the future, we may want to prevent any "standard" stage to have a higher
|
|
# sequence than any "won" stage. But while this is not the case, searching
|
|
# for the "won" stage while alterning the sequence order (see below) will correctly
|
|
# handle such a case :
|
|
# stage sequence : [x] [x (won)] [y] [y (won)] [z] [z (won)]
|
|
# when in stage [y] and marked as "won", should go to the stage [y (won)],
|
|
# not in [x (won)] nor [z (won)]
|
|
stage_id = next((stage for stage in won_stages if stage.sequence > lead.stage_id.sequence), None)
|
|
if not stage_id:
|
|
stage_id = next((stage for stage in reversed(won_stages) if stage.sequence <= lead.stage_id.sequence), won_stages)
|
|
if stage_id in leads_by_won_stage:
|
|
leads_by_won_stage[stage_id] += lead
|
|
else:
|
|
leads_by_won_stage[stage_id] = lead
|
|
for won_stage_id, leads in leads_by_won_stage.items():
|
|
leads.write({'stage_id': won_stage_id.id, 'probability': 100})
|
|
return True
|
|
|
|
def action_set_automated_probability(self):
|
|
self.write({'probability': self.automated_probability})
|
|
|
|
def action_set_won_rainbowman(self):
|
|
self.ensure_one()
|
|
self.action_set_won()
|
|
|
|
message = self._get_rainbowman_message()
|
|
if message:
|
|
return {
|
|
'effect': {
|
|
'fadeout': 'slow',
|
|
'message': message,
|
|
'img_url': '/web/image/%s/%s/image_1024' % (self.team_id.user_id._name, self.team_id.user_id.id) if self.team_id.user_id.image_1024 else '/web/static/img/smile.svg',
|
|
'type': 'rainbow_man',
|
|
}
|
|
}
|
|
return True
|
|
|
|
def get_rainbowman_message(self):
|
|
self.ensure_one()
|
|
if self.stage_id.is_won:
|
|
return self._get_rainbowman_message()
|
|
return False
|
|
|
|
def _get_rainbowman_message(self):
|
|
if not self.user_id or not self.team_id:
|
|
return False
|
|
if not self.expected_revenue:
|
|
# Show rainbow man for the first won lead of a salesman, even if expected revenue is not set. It is not
|
|
# very often that leads without revenues are marked won, so simply get count using ORM instead of query
|
|
today = fields.Datetime.today()
|
|
user_won_leads_count = self.search_count([
|
|
('type', '=', 'opportunity'),
|
|
('user_id', '=', self.user_id.id),
|
|
('probability', '=', 100),
|
|
('date_closed', '>=', date_utils.start_of(today, 'year')),
|
|
('date_closed', '<', date_utils.end_of(today, 'year')),
|
|
])
|
|
if user_won_leads_count == 1:
|
|
return _('Go, go, go! Congrats for your first deal.')
|
|
return False
|
|
|
|
self.flush_model() # flush fields to make sure DB is up to date
|
|
query = """
|
|
SELECT
|
|
SUM(CASE WHEN user_id = %(user_id)s THEN 1 ELSE 0 END) as total_won,
|
|
MAX(CASE WHEN date_closed >= CURRENT_DATE - INTERVAL '30 days' AND user_id = %(user_id)s THEN expected_revenue ELSE 0 END) as max_user_30,
|
|
MAX(CASE WHEN date_closed >= CURRENT_DATE - INTERVAL '7 days' AND user_id = %(user_id)s THEN expected_revenue ELSE 0 END) as max_user_7,
|
|
MAX(CASE WHEN date_closed >= CURRENT_DATE - INTERVAL '30 days' AND team_id = %(team_id)s THEN expected_revenue ELSE 0 END) as max_team_30,
|
|
MAX(CASE WHEN date_closed >= CURRENT_DATE - INTERVAL '7 days' AND team_id = %(team_id)s THEN expected_revenue ELSE 0 END) as max_team_7
|
|
FROM crm_lead
|
|
WHERE
|
|
type = 'opportunity'
|
|
AND
|
|
active = True
|
|
AND
|
|
probability = 100
|
|
AND
|
|
DATE_TRUNC('year', date_closed) = DATE_TRUNC('year', CURRENT_DATE)
|
|
AND
|
|
(user_id = %(user_id)s OR team_id = %(team_id)s)
|
|
"""
|
|
self.env.cr.execute(query, {'user_id': self.user_id.id,
|
|
'team_id': self.team_id.id})
|
|
query_result = self.env.cr.dictfetchone()
|
|
|
|
message = False
|
|
if query_result['total_won'] == 1:
|
|
message = _('Go, go, go! Congrats for your first deal.')
|
|
elif query_result['max_team_30'] == self.expected_revenue:
|
|
message = _('Boom! Team record for the past 30 days.')
|
|
elif query_result['max_team_7'] == self.expected_revenue:
|
|
message = _('Yeah! Deal of the last 7 days for the team.')
|
|
elif query_result['max_user_30'] == self.expected_revenue:
|
|
message = _('You just beat your personal record for the past 30 days.')
|
|
elif query_result['max_user_7'] == self.expected_revenue:
|
|
message = _('You just beat your personal record for the past 7 days.')
|
|
return message
|
|
|
|
def action_schedule_meeting(self, smart_calendar=True):
|
|
""" Open meeting's calendar view to schedule meeting on current opportunity.
|
|
|
|
:param smart_calendar: boolean, to set to False if the view should not try to choose relevant
|
|
mode and initial date for calendar view, see ``_get_opportunity_meeting_view_parameters``
|
|
:return dict: dictionary value for created Meeting view
|
|
"""
|
|
self.ensure_one()
|
|
action = self.env["ir.actions.actions"]._for_xml_id("calendar.action_calendar_event")
|
|
partner_ids = self.env.user.partner_id.ids
|
|
if self.partner_id:
|
|
partner_ids.append(self.partner_id.id)
|
|
current_opportunity_id = self.id if self.type == 'opportunity' else False
|
|
action['context'] = {
|
|
'search_default_opportunity_id': current_opportunity_id,
|
|
'default_opportunity_id': current_opportunity_id,
|
|
'default_partner_id': self.partner_id.id,
|
|
'default_partner_ids': partner_ids,
|
|
'default_team_id': self.team_id.id,
|
|
'default_name': self.name,
|
|
}
|
|
|
|
# 'Smart' calendar view : get the most relevant time period to display to the user.
|
|
if current_opportunity_id and smart_calendar:
|
|
mode, initial_date = self._get_opportunity_meeting_view_parameters()
|
|
action['context'].update({'default_mode': mode, 'initial_date': initial_date})
|
|
|
|
return action
|
|
|
|
def _get_opportunity_meeting_view_parameters(self):
|
|
""" Return the most relevant parameters for calendar view when viewing meetings linked to an opportunity.
|
|
If there are any meetings that are not finished yet, only consider those meetings,
|
|
since the user would prefer no to see past meetings. Otherwise, consider all meetings.
|
|
Allday events datetimes are used without taking tz into account.
|
|
-If there is no event, return week mode and false (The calendar will target 'now' by default)
|
|
-If there is only one, return week mode and date of the start of the event.
|
|
-If there are several events entirely on the same week, return week mode and start of first event.
|
|
-Else, return month mode and the date of the start of first event as initial date. (If they are
|
|
on the same month, this will display that month and therefore show all of them, which is expected)
|
|
|
|
:return tuple(mode, initial_date)
|
|
- mode: selected mode of the calendar view, 'week' or 'month'
|
|
- initial_date: date of the start of the first relevant meeting. The calendar will target that date.
|
|
"""
|
|
self.ensure_one()
|
|
meeting_results = self.env["calendar.event"].search_read([('opportunity_id', '=', self.id)], ['start', 'stop', 'allday'])
|
|
if not meeting_results:
|
|
return "week", False
|
|
|
|
user_tz = self.env.user.tz or self.env.context.get('tz')
|
|
user_pytz = pytz.timezone(user_tz) if user_tz else pytz.utc
|
|
|
|
# meeting_dts will contain one tuple of datetimes per meeting : (Start, Stop)
|
|
# meetings_dts and now_dt are as per user time zone.
|
|
meeting_dts = []
|
|
now_dt = datetime.now().astimezone(user_pytz).replace(tzinfo=None)
|
|
|
|
# When creating an allday meeting, whatever the TZ, it will be stored the same e.g. 00.00.00->23.59.59 in utc or
|
|
# 08.00.00->18.00.00. Therefore we must not put it back in the user tz but take it raw.
|
|
for meeting in meeting_results:
|
|
if meeting.get('allday'):
|
|
meeting_dts.append((meeting.get('start'), meeting.get('stop')))
|
|
else:
|
|
meeting_dts.append((meeting.get('start').astimezone(user_pytz).replace(tzinfo=None),
|
|
meeting.get('stop').astimezone(user_pytz).replace(tzinfo=None)))
|
|
|
|
# If there are meetings that are still ongoing or to come, only take those.
|
|
unfinished_meeting_dts = [meeting_dt for meeting_dt in meeting_dts if meeting_dt[1] >= now_dt]
|
|
relevant_meeting_dts = unfinished_meeting_dts if unfinished_meeting_dts else meeting_dts
|
|
relevant_meeting_count = len(relevant_meeting_dts)
|
|
|
|
if relevant_meeting_count == 1:
|
|
return "week", relevant_meeting_dts[0][0].date()
|
|
else:
|
|
# Range of meetings
|
|
earliest_start_dt = min(relevant_meeting_dt[0] for relevant_meeting_dt in relevant_meeting_dts)
|
|
latest_stop_dt = max(relevant_meeting_dt[1] for relevant_meeting_dt in relevant_meeting_dts)
|
|
|
|
# The week start day depends on language. We fetch the week_start of user's language. 1 is monday.
|
|
lang_week_start = self.env["res.lang"].search_read([('code', '=', self.env.user.lang)], ['week_start'])
|
|
# We substract one to make week_start_index range 0-6 instead of 1-7
|
|
week_start_index = int(lang_week_start[0].get('week_start', '1')) - 1
|
|
|
|
# We compute the weekday of earliest_start_dt according to week_start_index. earliest_start_dt_index will be 0 if we are on the
|
|
# first day of the week and 6 on the last. weekday() returns 0 for monday and 6 for sunday. For instance, Tuesday in UK is the
|
|
# third day of the week, so earliest_start_dt_index is 2, and remaining_days_in_week includes tuesday, so it will be 5.
|
|
# The first term 7 is there to avoid negative left side on the modulo, improving readability.
|
|
earliest_start_dt_weekday = (7 + earliest_start_dt.weekday() - week_start_index) % 7
|
|
remaining_days_in_week = 7 - earliest_start_dt_weekday
|
|
|
|
# We compute the start of the week following the one containing the start of the first meeting.
|
|
next_week_start_date = earliest_start_dt.date() + timedelta(days=remaining_days_in_week)
|
|
|
|
# Latest_stop_dt must be before the start of following week. Limit is therefore set at midnight of first day, included.
|
|
meetings_in_same_week = latest_stop_dt <= datetime(next_week_start_date.year, next_week_start_date.month, next_week_start_date.day, 0, 0, 0)
|
|
|
|
if meetings_in_same_week:
|
|
return "week", earliest_start_dt.date()
|
|
else:
|
|
return "month", earliest_start_dt.date()
|
|
|
|
def action_reschedule_meeting(self):
|
|
self.ensure_one()
|
|
action = self.action_schedule_meeting(smart_calendar=False)
|
|
next_activity = self.activity_ids.filtered(lambda activity: activity.user_id == self.env.user)[:1]
|
|
if next_activity.calendar_event_id:
|
|
action['context']['initial_date'] = next_activity.calendar_event_id.start
|
|
return action
|
|
|
|
def action_show_potential_duplicates(self):
|
|
""" Open kanban view to display duplicate leads or opportunity.
|
|
:return dict: dictionary value for created kanban view
|
|
"""
|
|
self.ensure_one()
|
|
action = self.env["ir.actions.actions"]._for_xml_id("crm.crm_lead_opportunities")
|
|
action['domain'] = [('id', 'in', self.duplicate_lead_ids.ids)]
|
|
action['context'] = {
|
|
'active_test': False,
|
|
'create': False
|
|
}
|
|
return action
|
|
|
|
def action_snooze(self):
|
|
self.ensure_one()
|
|
today = date.today()
|
|
my_next_activity = self.activity_ids.filtered(lambda activity: activity.user_id == self.env.user)[:1]
|
|
if my_next_activity:
|
|
if my_next_activity.date_deadline < today:
|
|
date_deadline = today + timedelta(days=7)
|
|
else:
|
|
date_deadline = my_next_activity.date_deadline + timedelta(days=7)
|
|
my_next_activity.write({
|
|
'date_deadline': date_deadline
|
|
})
|
|
return True
|
|
|
|
# ------------------------------------------------------------
|
|
# VIEWS
|
|
# ------------------------------------------------------------
|
|
|
|
def redirect_lead_opportunity_view(self):
|
|
self.ensure_one()
|
|
return {
|
|
'name': _('Lead or Opportunity'),
|
|
'view_mode': 'form',
|
|
'res_model': 'crm.lead',
|
|
'domain': [('type', '=', self.type)],
|
|
'res_id': self.id,
|
|
'view_id': False,
|
|
'type': 'ir.actions.act_window',
|
|
'context': {'default_type': self.type}
|
|
}
|
|
|
|
@api.model
|
|
def get_empty_list_help(self, help_message):
|
|
""" This method returns the action helpers for the leads. If help is already provided
|
|
on the action, the same is returned. Otherwise, we build the help message which
|
|
contains the alias responsible for creating the lead (if available) and return it.
|
|
"""
|
|
if not is_html_empty(help_message):
|
|
return help_message
|
|
|
|
help_title, sub_title = "", ""
|
|
if self._context.get('default_type') == 'lead':
|
|
help_title = _('Create a new lead')
|
|
else:
|
|
help_title = _('Create an opportunity to start playing with your pipeline.')
|
|
alias_domain = [
|
|
('company_id', '=', self.env.company.id),
|
|
('alias_id.alias_name', '!=', False),
|
|
('alias_id.alias_name', '!=', ''),
|
|
('alias_id.alias_model_id.model', '=', 'crm.lead'),
|
|
]
|
|
# sort by use_leads, then by our membership of the team
|
|
alias_records = self.env['crm.team'].search(alias_domain).sorted(
|
|
lambda r: (r.use_leads, self.env.user in r.member_ids), reverse=True
|
|
)
|
|
alias_record = alias_records[0] if alias_records else None
|
|
if alias_record and alias_record.alias_domain and alias_record.alias_name:
|
|
sub_title = Markup(_('Use the <i>New</i> button, or send an email to %(email_link)s to test the email gateway.')) % {
|
|
'email_link': Markup("<b><a href='mailto:%s'>%s</a></b>") % (alias_record.display_name, alias_record.display_name),
|
|
}
|
|
return super().get_empty_list_help(
|
|
f'<p class="o_view_nocontent_smiling_face">{help_title}</p><p class="oe_view_nocontent_alias">{sub_title}</p>'
|
|
)
|
|
|
|
# ------------------------------------------------------------
|
|
# BUSINESS
|
|
# ------------------------------------------------------------
|
|
|
|
def log_meeting(self, meeting):
|
|
""" Log the meeting info with a link to it in the chatter
|
|
:param record meeting: the meeting we want to log
|
|
"""
|
|
if not meeting.duration:
|
|
duration = _('unknown')
|
|
else:
|
|
duration = self.env['ir.qweb.field.duration'].value_to_html(meeting.duration, {'unit': 'hour'})
|
|
meeting_usertime = fields.Datetime.to_string(fields.Datetime.context_timestamp(self, meeting.start))
|
|
meeting_time = Markup("<time datetime='%(meeting_start)s+00:00'>%(meeting_user_time)s</time>") % {
|
|
'meeting_start': meeting.start,
|
|
'meeting_user_time': meeting_usertime,
|
|
}
|
|
message = Markup("<p>%(meeting)s<br/>%(subject_string)s %(subject_link)s<br/>%(duration)s<p>") % {
|
|
'meeting': _("Meeting scheduled at %s", meeting_time),
|
|
'subject_string': _("Subject: "),
|
|
'subject_link': meeting._get_html_link(),
|
|
'duration': _("Duration: %s", duration),
|
|
}
|
|
return self.message_post(body=message)
|
|
|
|
# ------------------------------------------------------------
|
|
# MERGE AND CONVERT LEADS / OPPORTUNITIES
|
|
# ------------------------------------------------------------
|
|
|
|
def _merge_data(self, fnames=None):
|
|
""" Prepare lead/opp data into a dictionary for merging. Different types
|
|
of fields are processed in different ways:
|
|
- text: all the values are concatenated
|
|
- m2m and o2m: those fields aren't processed
|
|
- m2o: the first not null value prevails (the other are dropped)
|
|
- any other type of field: same as m2o
|
|
|
|
:param fields: list of fields to process
|
|
:return dict data: contains the merged values of the new opportunity
|
|
"""
|
|
if fnames is None:
|
|
fnames = self._merge_get_fields()
|
|
fcallables = self._merge_get_fields_specific()
|
|
address_values = self._merge_get_fields_address()
|
|
|
|
# helpers
|
|
def _get_first_not_null(attr, opportunities):
|
|
value = False
|
|
for opp in opportunities:
|
|
if opp[attr]:
|
|
value = opp[attr].id if isinstance(opp[attr], models.BaseModel) else opp[attr]
|
|
break
|
|
return value
|
|
|
|
# process the field's values
|
|
data = {}
|
|
for field_name in fnames:
|
|
field = self._fields.get(field_name)
|
|
if field is None:
|
|
continue
|
|
|
|
fcallable = fcallables.get(field_name)
|
|
if fcallable and callable(fcallable):
|
|
data[field_name] = fcallable(field_name, self)
|
|
elif field_name in address_values:
|
|
data[field_name] = address_values[field_name]
|
|
elif not fcallable and field.type in ('many2many', 'one2many'):
|
|
continue
|
|
else:
|
|
data[field_name] = _get_first_not_null(field_name, self) # take the first not null
|
|
|
|
return data
|
|
|
|
def merge_opportunity(self, user_id=False, team_id=False, auto_unlink=True):
|
|
""" Merge opportunities in one. Different cases of merge:
|
|
- merge leads together = 1 new lead
|
|
- merge at least 1 opp with anything else (lead or opp) = 1 new opp
|
|
The resulting lead/opportunity will be the most important one (based on its confidence level)
|
|
updated with values from other opportunities to merge.
|
|
|
|
:param user_id : the id of the saleperson. If not given, will be determined by `_merge_data`.
|
|
:param team : the id of the Sales Team. If not given, will be determined by `_merge_data`.
|
|
|
|
:return crm.lead record resulting of th merge
|
|
"""
|
|
return self._merge_opportunity(user_id=user_id, team_id=team_id, auto_unlink=auto_unlink)
|
|
|
|
def _merge_opportunity(self, user_id=False, team_id=False, auto_unlink=True, max_length=5):
|
|
""" Private merging method. This one allows to relax rules on record set
|
|
length allowing to merge more than 5 opportunities at once if requested.
|
|
This should not be called by action buttons.
|
|
|
|
See ``merge_opportunity`` for more details. """
|
|
if len(self.ids) <= 1:
|
|
raise UserError(_('Please select more than one element (lead or opportunity) from the list view.'))
|
|
|
|
if max_length and len(self.ids) > max_length and not self.env.is_superuser():
|
|
raise UserError(_("To prevent data loss, Leads and Opportunities can only be merged by groups of %(max_length)s.", max_length=max_length))
|
|
|
|
opportunities = self._sort_by_confidence_level(reverse=True)
|
|
|
|
# get SORTED recordset of head and tail, and complete list
|
|
opportunities_head = opportunities[0]
|
|
opportunities_tail = opportunities[1:]
|
|
|
|
# merge all the sorted opportunity. This means the value of
|
|
# the first (head opp) will be a priority.
|
|
merged_data = opportunities._merge_data(self._merge_get_fields())
|
|
|
|
# force value for saleperson and Sales Team
|
|
if user_id:
|
|
merged_data['user_id'] = user_id
|
|
if team_id:
|
|
merged_data['team_id'] = team_id
|
|
|
|
merged_followers = opportunities_head._merge_followers(opportunities_tail)
|
|
|
|
# log merge message
|
|
opportunities_head._merge_log_summary(merged_followers, opportunities_tail)
|
|
# merge other data (mail.message, attachments, ...) from tail into head
|
|
opportunities_head._merge_dependences(opportunities_tail)
|
|
|
|
# check if the stage is in the stages of the Sales Team. If not, assign the stage with the lowest sequence
|
|
if merged_data.get('team_id'):
|
|
team_stage_ids = self.env['crm.stage'].search(['|', ('team_id', '=', merged_data['team_id']), ('team_id', '=', False)], order='sequence, id')
|
|
if merged_data.get('stage_id') not in team_stage_ids.ids:
|
|
merged_data['stage_id'] = team_stage_ids[0].id if team_stage_ids else False
|
|
|
|
# write merged data into first opportunity; remove some keys if already
|
|
# set on opp to avoid useless recomputes
|
|
if 'user_id' in merged_data and opportunities_head.user_id.id == merged_data['user_id']:
|
|
merged_data.pop('user_id')
|
|
if 'team_id' in merged_data and opportunities_head.team_id.id == merged_data['team_id']:
|
|
merged_data.pop('team_id')
|
|
opportunities_head.write(merged_data)
|
|
|
|
# delete tail opportunities
|
|
# we use the SUPERUSER to avoid access rights issues because as the user had the rights to see the records it should be safe to do so
|
|
if auto_unlink:
|
|
opportunities_tail.sudo().unlink()
|
|
|
|
return opportunities_head
|
|
|
|
def _merge_get_fields_address(self):
|
|
"""The address fields are propagated as a whole.
|
|
|
|
The address is taken from the lead with the most non-empty address field
|
|
(sorted by highest rank if multiple lead have the same amount of non-empty
|
|
fields).
|
|
"""
|
|
source_lead = max(self, key=lambda lead: len(list(
|
|
lead[field] for field in PARTNER_ADDRESS_FIELDS_TO_SYNC
|
|
if lead[field]
|
|
)))
|
|
return {fname: source_lead[fname] for fname in PARTNER_ADDRESS_FIELDS_TO_SYNC}
|
|
|
|
def _merge_get_fields_specific(self):
|
|
return {
|
|
'description': lambda fname, leads: '<br/><br/>'.join(desc for desc in leads.mapped('description') if not is_html_empty(desc)),
|
|
'type': lambda fname, leads: 'opportunity' if any(lead.type == 'opportunity' for lead in leads) else 'lead',
|
|
'priority': lambda fname, leads: max(leads.mapped('priority')) if leads else False,
|
|
'tag_ids': lambda fname, leads: leads.mapped('tag_ids'),
|
|
'lost_reason_id': lambda fname, leads:
|
|
False if leads and leads[0].probability
|
|
else next((lead.lost_reason_id for lead in leads if lead.lost_reason_id), False),
|
|
}
|
|
|
|
def _merge_get_fields(self):
|
|
return (
|
|
CRM_LEAD_FIELDS_TO_MERGE
|
|
+ list(self._merge_get_fields_specific().keys())
|
|
+ PARTNER_ADDRESS_FIELDS_TO_SYNC
|
|
)
|
|
|
|
def _merge_dependences(self, opportunities):
|
|
""" Merge dependences (messages, attachments,activities, calendar events,
|
|
...). These dependences will be transfered to `self` considered as the
|
|
master lead.
|
|
|
|
:param opportunities : recordset of opportunities to transfer. Does not
|
|
include `self` which is the target crm.lead being the result of the
|
|
merge;
|
|
"""
|
|
self.ensure_one()
|
|
self._merge_dependences_history(opportunities)
|
|
self._merge_dependences_attachments(opportunities)
|
|
self._merge_dependences_calendar_events(opportunities)
|
|
|
|
def _merge_dependences_history(self, opportunities):
|
|
""" Move history from the given opportunities to the current one. `self`
|
|
is the crm.lead record destination for message of `opportunities`.
|
|
|
|
This method moves
|
|
* messages
|
|
* activities
|
|
|
|
:param opportunities: see ``_merge_dependences``
|
|
"""
|
|
self.ensure_one()
|
|
for opportunity in opportunities:
|
|
for message in opportunity.message_ids:
|
|
if message.subject:
|
|
subject = _("From %(source_name)s: %(source_subject)s", source_name=opportunity.name, source_subject=message.subject)
|
|
else:
|
|
subject = _("From %(source_name)s", source_name=opportunity.name)
|
|
message.write({
|
|
'res_id': self.id,
|
|
'subject': subject,
|
|
})
|
|
|
|
opportunities.activity_ids.write({
|
|
'res_id': self.id,
|
|
})
|
|
|
|
return True
|
|
|
|
def _merge_dependences_attachments(self, opportunities):
|
|
""" Move attachments of given opportunities to the current one `self`, and rename
|
|
the attachments having same name than native ones.
|
|
|
|
:param opportunities: see ``_merge_dependences``
|
|
"""
|
|
self.ensure_one()
|
|
|
|
all_attachments = self.env['ir.attachment'].search([
|
|
('res_model', '=', self._name),
|
|
('res_id', 'in', opportunities.ids)
|
|
])
|
|
|
|
for opportunity in opportunities:
|
|
attachments = all_attachments.filtered(lambda attach: attach.res_id == opportunity.id)
|
|
for attachment in attachments:
|
|
attachment.write({
|
|
'res_id': self.id,
|
|
'name': _("%(attach_name)s (from %(lead_name)s)",
|
|
attach_name=attachment.name,
|
|
lead_name=opportunity.name[:20]
|
|
)
|
|
})
|
|
return True
|
|
|
|
def _merge_dependences_calendar_events(self, opportunities):
|
|
""" Move calender.event from the given opportunities to the current one. `self` is the
|
|
crm.lead record destination for event of `opportunities`.
|
|
:param opportunities: see ``merge_dependences``
|
|
"""
|
|
self.ensure_one()
|
|
meetings = self.env['calendar.event'].search([('opportunity_id', 'in', opportunities.ids)])
|
|
return meetings.write({
|
|
'res_id': self.id,
|
|
'opportunity_id': self.id,
|
|
})
|
|
|
|
def _merge_followers(self, opportunities):
|
|
"""Add the followers into the destination lead if they post a message in the last 30 days.
|
|
|
|
:param opportunities : Record<crm.lead> of opportunities to transfer
|
|
:return: {old_lead_id: Record<mail.followers>} Followers which have been added in
|
|
the destination lead grouped by source lead ID.
|
|
"""
|
|
self.ensure_one()
|
|
|
|
self.env['mail.message'].flush_model()
|
|
self.env['mail.followers'].flush_model()
|
|
|
|
# Get the active followers (followers whose partner post a message on the
|
|
# leads in the last 30 days) which should be moved on the destination lead
|
|
self.env.cr.execute(
|
|
'''
|
|
SELECT MAX(mf.id) AS id
|
|
FROM mail_followers AS mf
|
|
JOIN mail_message AS mm
|
|
ON mm.author_id = mf.partner_id
|
|
AND mm.res_id = mf.res_id
|
|
AND mm.model = 'crm.lead'
|
|
AND mm.date > NOW() - INTERVAL '30 DAY'
|
|
/* Check if the partner is already
|
|
following the destination lead */
|
|
LEFT JOIN mail_followers AS destf
|
|
ON destf.res_model = 'crm.lead'
|
|
AND destf.res_id = %(lead_id)s
|
|
AND destf.partner_id = mf.partner_id
|
|
/* Select only once each partner
|
|
to not create duplicated followers */
|
|
WHERE mf.res_model = 'crm.lead'
|
|
AND mf.res_id IN %(lead_ids)s
|
|
AND destf IS NULL
|
|
GROUP BY mf.partner_id
|
|
''',
|
|
{'lead_ids': tuple(opportunities.ids), 'lead_id': self.id},
|
|
)
|
|
followers_to_update = [r[0] for r in self.env.cr.fetchall()]
|
|
followers_to_update = self.env['mail.followers'].browse(followers_to_update).sudo()
|
|
followers_by_old_lead = dict(groupby(followers_to_update, lambda f: f.res_id))
|
|
followers_to_update.write({'res_id': self.id})
|
|
return followers_by_old_lead
|
|
|
|
def _merge_log_summary(self, merged_followers, opportunities_tail):
|
|
"""Log the merge message on the lead."""
|
|
self.ensure_one()
|
|
self.message_post_with_source(
|
|
"crm.crm_lead_merge_summary",
|
|
render_values={
|
|
"merged_followers": merged_followers,
|
|
"opportunities": opportunities_tail,
|
|
"is_html_empty": is_html_empty,
|
|
},
|
|
subtype_xmlid='mail.mt_note',
|
|
)
|
|
|
|
def _format_properties(self):
|
|
"""Format the properties to build the merge message.
|
|
|
|
Return a list of dict containing the label, and a value key if there's only
|
|
one value, or a "values" key if we have multiple values (e.g. many2many, tags).
|
|
|
|
E.G.
|
|
[{
|
|
'label': 'My Partner',
|
|
'value': 'Alice',
|
|
}, {
|
|
'label': 'My Partners',
|
|
'values': [
|
|
{'name': 'Alice'},
|
|
{'name': 'Bob'},
|
|
],
|
|
}, {
|
|
'label': 'My Tags',
|
|
'values': [
|
|
{'name': 'A', 'color': 1},
|
|
{'name': 'C', 'color': 3},
|
|
],
|
|
}]
|
|
"""
|
|
self.ensure_one()
|
|
# read to have the display names already in the value
|
|
properties = self.read(['lead_properties'])[0]['lead_properties']
|
|
|
|
formatted = []
|
|
for definition in properties:
|
|
label = definition.get('string')
|
|
value = definition.get('value')
|
|
property_type = definition['type']
|
|
if not value and property_type != 'boolean':
|
|
continue
|
|
|
|
property_dict = {'label': label}
|
|
if property_type == 'boolean':
|
|
property_dict['value'] = _('Yes') if value else _('No')
|
|
elif value and property_type == 'many2one':
|
|
property_dict['value'] = value[1]
|
|
elif value and property_type == 'many2many':
|
|
# show many2many in badge
|
|
property_dict['values'] = [{'name': rec[1]} for rec in value]
|
|
elif value and property_type in ['selection', 'tags']:
|
|
# retrieve the option label from the value
|
|
options = {
|
|
option[0]: option[1:]
|
|
for option in (definition.get(property_type) or [])
|
|
}
|
|
if property_type == 'selection':
|
|
value = options.get(value)
|
|
property_dict['value'] = value[0] if value else None
|
|
else:
|
|
property_dict['values'] = [{
|
|
'name': options[tag][0],
|
|
'color': options[tag][1],
|
|
} for tag in value if tag in options
|
|
]
|
|
else:
|
|
property_dict['value'] = value
|
|
|
|
formatted.append(property_dict)
|
|
|
|
return formatted
|
|
|
|
# CONVERT
|
|
# ----------------------------------------------------------------------
|
|
|
|
def _convert_opportunity_data(self, customer, team_id=False):
|
|
""" Extract the data from a lead to create the opportunity
|
|
:param customer : res.partner record
|
|
:param team_id : identifier of the Sales Team to determine the stage
|
|
"""
|
|
new_team_id = team_id if team_id else self.team_id.id
|
|
upd_values = {
|
|
'type': 'opportunity',
|
|
'date_conversion': self.env.cr.now(),
|
|
}
|
|
if customer != self.partner_id:
|
|
upd_values['partner_id'] = customer.id if customer else False
|
|
if not self.stage_id:
|
|
stage = self._stage_find(team_id=new_team_id)
|
|
upd_values['stage_id'] = stage.id
|
|
return upd_values
|
|
|
|
def convert_opportunity(self, partner, user_ids=False, team_id=False):
|
|
customer = partner if partner else self.env['res.partner']
|
|
for lead in self:
|
|
if not lead.active or lead.probability == 100:
|
|
continue
|
|
vals = lead._convert_opportunity_data(customer, team_id)
|
|
lead.write(vals)
|
|
|
|
if user_ids or team_id:
|
|
self._handle_salesmen_assignment(user_ids=user_ids, team_id=team_id)
|
|
|
|
return True
|
|
|
|
def _handle_partner_assignment(self, force_partner_id=False, create_missing=True):
|
|
""" Update customer (partner_id) of leads. Purpose is to set the same
|
|
partner on most leads; either through a newly created partner either
|
|
through a given partner_id.
|
|
|
|
:param int force_partner_id: if set, update all leads to that customer;
|
|
:param create_missing: for leads without customer, create a new one
|
|
based on lead information;
|
|
"""
|
|
for lead in self:
|
|
if force_partner_id:
|
|
lead.partner_id = force_partner_id
|
|
if not lead.partner_id and create_missing:
|
|
partner = lead._create_customer()
|
|
lead.partner_id = partner.id
|
|
|
|
def _handle_salesmen_assignment(self, user_ids=False, team_id=False):
|
|
""" Assign salesmen and salesteam to a batch of leads. If there are more
|
|
leads than salesmen, these salesmen will be assigned in round-robin. E.g.
|
|
4 salesmen (S1, S2, S3, S4) for 6 leads (L1, L2, ... L6) will assigned as
|
|
following: L1 - S1, L2 - S2, L3 - S3, L4 - S4, L5 - S1, L6 - S2.
|
|
|
|
:param list user_ids: salesmen to assign
|
|
:param int team_id: salesteam to assign
|
|
"""
|
|
update_vals = {'team_id': team_id} if team_id else {}
|
|
if not user_ids and team_id:
|
|
self.write(update_vals)
|
|
else:
|
|
lead_ids = self.ids
|
|
steps = len(user_ids)
|
|
# pass 1 : lead_ids[0:6:3] = [L1,L4]
|
|
# pass 2 : lead_ids[1:6:3] = [L2,L5]
|
|
# pass 3 : lead_ids[2:6:3] = [L3,L6]
|
|
# ...
|
|
for idx in range(0, steps):
|
|
subset_ids = lead_ids[idx:len(lead_ids):steps]
|
|
update_vals['user_id'] = user_ids[idx]
|
|
self.env['crm.lead'].browse(subset_ids).write(update_vals)
|
|
|
|
# ------------------------------------------------------------
|
|
# MERGE / CONVERT TOOLS
|
|
# ---------------------------------------------------------
|
|
|
|
# CLASSIFICATION TOOLS
|
|
# --------------------------------------------------
|
|
|
|
def _get_lead_duplicates(self, partner=None, email=None, include_lost=False):
|
|
""" Search for leads that seem duplicated based on partner / email.
|
|
|
|
:param partner : optional customer when searching duplicated
|
|
:param email: email (possibly formatted) to search
|
|
:param boolean include_lost: if True, search includes archived opportunities
|
|
(still only active leads are considered). If False, search for active
|
|
and not won leads and opportunities;
|
|
"""
|
|
if not email and not partner:
|
|
return self.env['crm.lead']
|
|
|
|
domain = []
|
|
for normalized_email in [tools.email_normalize(email) for email in tools.email_split(email)]:
|
|
domain.append(('email_normalized', '=', normalized_email))
|
|
if partner:
|
|
domain.append(('partner_id', '=', partner.id))
|
|
|
|
if not domain:
|
|
return self.env['crm.lead']
|
|
|
|
domain = ['|'] * (len(domain) - 1) + domain
|
|
if include_lost:
|
|
domain += ['|', ('type', '=', 'opportunity'), ('active', '=', True)]
|
|
else:
|
|
domain += ['&', ('active', '=', True), '|', ('stage_id', '=', False), ('stage_id.is_won', '=', False)]
|
|
|
|
return self.with_context(active_test=False).search(domain)
|
|
|
|
def _sort_by_confidence_level(self, reverse=False):
|
|
""" Sorting the leads/opps according to the confidence level to it
|
|
being won. It is sorted following this incremental heuristics :
|
|
|
|
* "not lost" first (inactive leads are lost); normally all leads
|
|
should be active but in case lost one, they are always last.
|
|
Inactive opportunities are considered as valid;
|
|
* opportunity is more reliable than a lead which is a pre-stage
|
|
used mainly for first classification;
|
|
* stage sequence: the higher the better as it indicates we are moving
|
|
towards won stage;
|
|
* probability: the higher the better as it is more likely to be won;
|
|
* ID: the higher the better when all other parameters are equal. We
|
|
consider newer leads to be more reliable;
|
|
"""
|
|
def opps_key(opportunity):
|
|
return opportunity.type == 'opportunity' or opportunity.active, \
|
|
opportunity.type == 'opportunity', \
|
|
opportunity.stage_id.sequence, \
|
|
opportunity.probability, \
|
|
-opportunity._origin.id
|
|
|
|
return self.sorted(key=opps_key, reverse=reverse)
|
|
|
|
# CUSTOMER TOOLS
|
|
# --------------------------------------------------
|
|
|
|
def _find_matching_partner(self, email_only=False):
|
|
""" Try to find a matching partner with available information on the
|
|
lead, using notably customer's name, email, ...
|
|
|
|
:param email_only: Only find a matching based on the email. To use
|
|
for automatic process where ilike based on name can be too dangerous
|
|
:return: partner browse record
|
|
"""
|
|
self.ensure_one()
|
|
partner = self.partner_id
|
|
|
|
if not partner and self.email_from:
|
|
partner = self.env['res.partner'].search([('email', '=', self.email_from)], limit=1)
|
|
|
|
if not partner and not email_only:
|
|
# search through the existing partners based on the lead's partner or contact name
|
|
# to be aligned with _create_customer, search on lead's name as last possibility
|
|
for customer_potential_name in [self[field_name] for field_name in ['partner_name', 'contact_name', 'name'] if self[field_name]]:
|
|
partner = self.env['res.partner'].search([('name', 'ilike', customer_potential_name)], limit=1)
|
|
if partner:
|
|
break
|
|
|
|
return partner
|
|
|
|
def _create_customer(self):
|
|
""" Create a partner from lead data and link it to the lead.
|
|
|
|
:return: newly-created partner browse record
|
|
"""
|
|
Partner = self.env['res.partner']
|
|
contact_name = self.contact_name
|
|
if not contact_name:
|
|
contact_name = parse_contact_from_email(self.email_from)[0] if self.email_from else False
|
|
|
|
if self.partner_name:
|
|
partner_company = Partner.create(self._prepare_customer_values(self.partner_name, is_company=True))
|
|
elif self.partner_id:
|
|
partner_company = self.partner_id
|
|
else:
|
|
partner_company = None
|
|
|
|
if contact_name:
|
|
return Partner.create(self._prepare_customer_values(contact_name, is_company=False, parent_id=partner_company.id if partner_company else False))
|
|
|
|
if partner_company:
|
|
return partner_company
|
|
return Partner.create(self._prepare_customer_values(self.name, is_company=False))
|
|
|
|
def _get_customer_information(self):
|
|
email_normalized_to_values = super()._get_customer_information()
|
|
Partner = self.env['res.partner']
|
|
|
|
for record in self.filtered('email_normalized'):
|
|
values = email_normalized_to_values.setdefault(record.email_normalized, {})
|
|
contact_name = record.contact_name or record.partner_name or parse_contact_from_email(record.email_from)[0]
|
|
# Note that we don't attempt to create the parent company even if partner name is set
|
|
values.update(record._prepare_customer_values(contact_name, is_company=False))
|
|
values['company_name'] = record.partner_name
|
|
if contact_name == record.partner_name:
|
|
values['company_type'] = 'company'
|
|
return email_normalized_to_values
|
|
|
|
def _prepare_customer_values(self, partner_name, is_company=False, parent_id=False):
|
|
""" Extract data from lead to create a partner.
|
|
|
|
:param name : furtur name of the partner
|
|
:param is_company : True if the partner is a company
|
|
:param parent_id : id of the parent partner (False if no parent)
|
|
|
|
:return: dictionary of values to give at res_partner.create()
|
|
"""
|
|
email_parts = tools.email_split(self.email_from)
|
|
res = {
|
|
'name': partner_name,
|
|
'user_id': self.env.context.get('default_user_id') or self.user_id.id,
|
|
'comment': self.description,
|
|
'team_id': self.team_id.id,
|
|
'parent_id': parent_id,
|
|
'phone': self.phone,
|
|
'mobile': self.mobile,
|
|
'email': email_parts[0] if email_parts else False,
|
|
'title': self.title.id,
|
|
'function': self.function,
|
|
'street': self.street,
|
|
'street2': self.street2,
|
|
'zip': self.zip,
|
|
'city': self.city,
|
|
'country_id': self.country_id.id,
|
|
'state_id': self.state_id.id,
|
|
'website': self.website,
|
|
'is_company': is_company,
|
|
'type': 'contact'
|
|
}
|
|
if self.lang_id.active:
|
|
res['lang'] = self.lang_id.code
|
|
return res
|
|
|
|
# ------------------------------------------------------------
|
|
# MAILING
|
|
# ------------------------------------------------------------
|
|
|
|
def _creation_subtype(self):
|
|
return self.env.ref('crm.mt_lead_create')
|
|
|
|
def _track_subtype(self, init_values):
|
|
self.ensure_one()
|
|
if 'stage_id' in init_values and self.probability == 100 and self.stage_id:
|
|
return self.env.ref('crm.mt_lead_won')
|
|
elif 'lost_reason_id' in init_values and self.lost_reason_id:
|
|
return self.env.ref('crm.mt_lead_lost')
|
|
elif 'stage_id' in init_values:
|
|
return self.env.ref('crm.mt_lead_stage')
|
|
elif 'active' in init_values and self.active:
|
|
return self.env.ref('crm.mt_lead_restored')
|
|
elif 'active' in init_values and not self.active:
|
|
return self.env.ref('crm.mt_lead_lost')
|
|
return super(Lead, self)._track_subtype(init_values)
|
|
|
|
def _notify_by_email_prepare_rendering_context(self, message, msg_vals=False, model_description=False,
|
|
force_email_company=False, force_email_lang=False):
|
|
render_context = super()._notify_by_email_prepare_rendering_context(
|
|
message, msg_vals, model_description=model_description,
|
|
force_email_company=force_email_company, force_email_lang=force_email_lang
|
|
)
|
|
if self.date_deadline:
|
|
render_context['subtitles'].append(
|
|
_('Deadline: %s', self.date_deadline.strftime(get_lang(self.env).date_format)))
|
|
return render_context
|
|
|
|
def _notify_get_recipients_groups(self, message, model_description, msg_vals=None):
|
|
""" Handle salesman recipients that can convert leads into opportunities
|
|
and set opportunities as won / lost. """
|
|
groups = super()._notify_get_recipients_groups(
|
|
message, model_description, msg_vals=msg_vals
|
|
)
|
|
if not self:
|
|
return groups
|
|
|
|
local_msg_vals = dict(msg_vals or {})
|
|
|
|
self.ensure_one()
|
|
if self.type == 'lead':
|
|
convert_action = self._notify_get_action_link('controller', controller='/lead/convert', **local_msg_vals)
|
|
salesman_actions = [{'url': convert_action, 'title': _('Convert to opportunity')}]
|
|
else:
|
|
won_action = self._notify_get_action_link('controller', controller='/lead/case_mark_won', **local_msg_vals)
|
|
lost_action = self._notify_get_action_link('controller', controller='/lead/case_mark_lost', **local_msg_vals)
|
|
salesman_actions = [
|
|
{'url': won_action, 'title': _('Mark Won')},
|
|
{'url': lost_action, 'title': _('Mark Lost')}]
|
|
|
|
salesman_group_id = self.env.ref('sales_team.group_sale_salesman').id
|
|
new_group = (
|
|
'group_sale_salesman',
|
|
lambda pdata: pdata['type'] == 'user' and salesman_group_id in pdata['groups'],
|
|
{
|
|
'actions': salesman_actions,
|
|
'active': True,
|
|
'has_button_access': True,
|
|
}
|
|
)
|
|
|
|
return [new_group] + groups
|
|
|
|
def _notify_get_reply_to(self, default=None):
|
|
""" Override to set alias of lead and opportunities to their sales team if any. """
|
|
aliases = self.mapped('team_id').sudo()._notify_get_reply_to(default=default)
|
|
res = {lead.id: aliases.get(lead.team_id.id) for lead in self}
|
|
leftover = self.filtered(lambda rec: not rec.team_id)
|
|
if leftover:
|
|
res.update(super(Lead, leftover)._notify_get_reply_to(default=default))
|
|
return res
|
|
|
|
def _message_get_default_recipients(self):
|
|
return {
|
|
r.id: {
|
|
'partner_ids': [],
|
|
'email_to': ','.join(tools.email_normalize_all(r.email_from)) or r.email_from,
|
|
'email_cc': False,
|
|
} for r in self
|
|
}
|
|
|
|
def _message_get_suggested_recipients(self):
|
|
recipients = super(Lead, self)._message_get_suggested_recipients()
|
|
try:
|
|
for lead in self:
|
|
# check if that language is correctly installed (and active) before using it
|
|
lang_code = lead.lang_code if lead.lang_code and self.env['res.lang']._lang_get(lead.lang_code) else None
|
|
if lead.partner_id:
|
|
lead._message_add_suggested_recipient(
|
|
recipients, partner=lead.partner_id, lang=lang_code, reason=_('Customer'))
|
|
elif lead.email_from:
|
|
lead._message_add_suggested_recipient(
|
|
recipients, email=lead.email_from, lang=lang_code, reason=_('Customer Email'))
|
|
except AccessError: # no read access rights -> just ignore suggested recipients because this imply modifying followers
|
|
pass
|
|
return recipients
|
|
|
|
@api.model
|
|
def message_new(self, msg_dict, custom_values=None):
|
|
""" Overrides mail_thread message_new that is called by the mailgateway
|
|
through message_process.
|
|
This override updates the document according to the email.
|
|
"""
|
|
# remove default author when going through the mail gateway. Indeed we
|
|
# do not want to explicitly set an user as responsible. We prefer that
|
|
# assignment is done automatically (scoring) or manually. Otherwise it
|
|
# would always be root (gateway user). It also allows to exclude portal
|
|
# and public users.
|
|
self = self.with_context(default_user_id=False)
|
|
|
|
if custom_values is None:
|
|
custom_values = {}
|
|
defaults = {
|
|
'name': msg_dict.get('subject') or _("No Subject"),
|
|
'email_from': msg_dict.get('from'),
|
|
'partner_id': msg_dict.get('author_id', False),
|
|
}
|
|
if msg_dict.get('priority') in dict(crm_stage.AVAILABLE_PRIORITIES):
|
|
defaults['priority'] = msg_dict.get('priority')
|
|
defaults.update(custom_values)
|
|
|
|
return super(Lead, self).message_new(msg_dict, custom_values=defaults)
|
|
|
|
def _message_post_after_hook(self, message, msg_vals):
|
|
if self.email_from and not self.partner_id:
|
|
# we consider that posting a message with a specified recipient (not a follower, a specific one)
|
|
# on a document without customer means that it was created through the chatter using
|
|
# suggested recipients. This heuristic allows to avoid ugly hacks in JS.
|
|
new_partner = message.partner_ids.filtered(
|
|
lambda partner: partner.email == self.email_from or (self.email_normalized and partner.email_normalized == self.email_normalized)
|
|
)
|
|
if new_partner:
|
|
if new_partner[0].email_normalized:
|
|
email_domain = ('email_normalized', '=', new_partner[0].email_normalized)
|
|
else:
|
|
email_domain = ('email_from', '=', new_partner[0].email)
|
|
self.search([
|
|
('partner_id', '=', False), email_domain, ('stage_id.fold', '=', False)
|
|
]).write({'partner_id': new_partner[0].id})
|
|
return super(Lead, self)._message_post_after_hook(message, msg_vals)
|
|
|
|
def _message_partner_info_from_emails(self, emails, link_mail=False):
|
|
""" Try to propose a better recipient when having only an email by populating
|
|
it with the partner_name / contact_name field of the lead e.g. if lead
|
|
contact_name is "Raoul" and email is "raoul@raoul.fr", suggest
|
|
"Raoul" <raoul@raoul.fr> as recipient. """
|
|
result = super(Lead, self)._message_partner_info_from_emails(emails, link_mail=link_mail)
|
|
if not (self.partner_name or self.contact_name) or not self.email_from:
|
|
return result
|
|
for email, partner_info in zip(emails, result):
|
|
if partner_info.get('partner_id') or not email:
|
|
continue
|
|
# reformat email if no name information
|
|
name_emails = tools.email_split_tuples(email)
|
|
name_from_email = name_emails[0][0] if name_emails else False
|
|
if name_from_email:
|
|
continue # already containing name + email
|
|
name_from_email = self.partner_name or self.contact_name
|
|
emails_normalized = tools.email_normalize_all(email)
|
|
email_normalized = emails_normalized[0] if emails_normalized else False
|
|
if email.lower() == self.email_from.lower() or (email_normalized and self.email_normalized == email_normalized):
|
|
partner_info['full_name'] = tools.formataddr((
|
|
name_from_email,
|
|
','.join(emails_normalized) if emails_normalized else email))
|
|
break
|
|
return result
|
|
|
|
@api.model
|
|
def get_import_templates(self):
|
|
return [{
|
|
'label': _('Import Template for Leads & Opportunities'),
|
|
'template': '/crm/static/xls/crm_lead.xls'
|
|
}]
|
|
|
|
# ------------------------------------------------------------
|
|
# PLS
|
|
# ------------------------------------------------------------
|
|
# Predictive lead scoring is computing the lead probability, based on won and lost leads from the past
|
|
# Each won/lost lead increments a frequency table, where we store, for each field/value couple, the number of
|
|
# won and lost leads.
|
|
# E.g. : A won lead from Belgium will increase the won count of the frequency country_id='Belgium' by 1.
|
|
# The frequencies are split by team_id, so each team has its own frequencies environment. (Team A doesn't impact B)
|
|
# There are two main ways to build the frequency table:
|
|
# - Live Increment: At each Won/lost, we increment directly the frequencies based on the lead values.
|
|
# Done right BEFORE writing the lead as won or lost.
|
|
# We consider a lead that will be marked as won or lost.
|
|
# Used each time a lead is won or lost, to ensure frequency table is always up to date
|
|
# - One shot Rebuild: empty the frequency table and rebuild it from scratch, based on every already won/lost leads
|
|
# Done during cron process.
|
|
# We consider all the leads that have been already won or lost.
|
|
# Used in one shot, when modifying the criteria to take into account (fields or reference date)
|
|
|
|
# ---------------------------------
|
|
# PLS: Probability Computation
|
|
# ---------------------------------
|
|
def _pls_get_naive_bayes_probabilities(self, batch_mode=False):
|
|
"""
|
|
In machine learning, naive Bayes classifiers (NBC) are a family of simple "probabilistic classifiers" based on
|
|
applying Bayes theorem with strong (naive) independence assumptions between the variables taken into account.
|
|
E.g: will TDE eat m&m's depending on his sleep status, the amount of work he has and the fullness of his stomach?
|
|
As we use experience to compute the statistics, every day, we will register the variables state + the result.
|
|
As the days pass, we will be able to determine, with more and more precision, if TDE will eat m&m's
|
|
for a specific combination :
|
|
- did sleep very well, a lot of work and stomach full > Will never happen !
|
|
- didn't sleep at all, no work at all and empty stomach > for sure !
|
|
Following Bayes' Theorem: the probability that an event occurs (to win) under certain conditions is proportional
|
|
to the probability to win under each condition separately and the probability to win. We compute a 'Win score'
|
|
-> P(Won | A∩B) ∝ P(A∩B | Won)*P(Won) OR S(Won | A∩B) = P(A∩B | Won)*P(Won)
|
|
To compute a percentage of probability to win, we also compute the 'Lost score' that is proportional to the
|
|
probability to lose under each condition separately and the probability to lose.
|
|
-> Probability = S(Won | A∩B) / ( S(Won | A∩B) + S(Lost | A∩B) )
|
|
See https://www.youtube.com/watch?v=CPqOCI0ahss can help to get a quick and simple example.
|
|
One issue about NBC is when a event occurence is never observed.
|
|
E.g: if when TDE has an empty stomach, he always eat m&m's, than the "not eating m&m's when empty stomach' event
|
|
will never be observed.
|
|
This is called 'zero frequency' and that leads to division (or at least multiplication) by zero.
|
|
To avoid this, we add 0.1 in each frequency. With few data, the computation is than not really realistic.
|
|
The more we have records to analyse, the more the estimation will be precise.
|
|
:return: probability in percent (and integer rounded) that the lead will be won at the current stage.
|
|
"""
|
|
lead_probabilities = {}
|
|
if not self:
|
|
return lead_probabilities
|
|
|
|
# Get all leads values, no matter the team_id
|
|
domain = []
|
|
if batch_mode:
|
|
domain = [
|
|
'&',
|
|
('active', '=', True), ('id', 'in', self.ids),
|
|
'|',
|
|
('probability', '=', None),
|
|
'&',
|
|
('probability', '<', 100), ('probability', '>', 0)
|
|
]
|
|
leads_values_dict = self._pls_get_lead_pls_values(domain=domain)
|
|
|
|
if not leads_values_dict:
|
|
return lead_probabilities
|
|
|
|
# Get unique couples to search in frequency table and won leads.
|
|
leads_fields = set() # keep unique fields, as a lead can have multiple tag_ids
|
|
won_leads = set()
|
|
won_stage_ids = self.env['crm.stage'].search([('is_won', '=', True)]).ids
|
|
for lead_id, values in leads_values_dict.items():
|
|
for field, value in values['values']:
|
|
if field == 'stage_id' and value in won_stage_ids:
|
|
won_leads.add(lead_id)
|
|
leads_fields.add(field)
|
|
leads_fields = sorted(leads_fields)
|
|
# get all variable related records from frequency table, no matter the team_id
|
|
frequencies = self.env['crm.lead.scoring.frequency'].search([('variable', 'in', list(leads_fields))], order="team_id asc, id")
|
|
|
|
# get all team_ids from frequencies
|
|
frequency_teams = frequencies.mapped('team_id')
|
|
frequency_team_ids = [team.id for team in frequency_teams]
|
|
|
|
# 1. Compute each variable value count individually
|
|
# regroup each variable to be able to compute their own probabilities
|
|
# As all the variable does not enter into account (as we reject unset values in the process)
|
|
# each value probability must be computed only with their own variable related total count
|
|
# special case: for lead for which team_id is not in frequency table or lead with no team_id,
|
|
# we consider all the records, independently from team_id (this is why we add a result[-1])
|
|
result = dict((team_id, dict((field, dict(won_total=0, lost_total=0)) for field in leads_fields)) for team_id in frequency_team_ids)
|
|
result[-1] = dict((field, dict(won_total=0, lost_total=0)) for field in leads_fields)
|
|
for frequency in frequencies:
|
|
field = frequency['variable']
|
|
value = frequency['value']
|
|
|
|
# To avoid that a tag take too much importance if its subset is too small,
|
|
# we ignore the tag frequencies if we have less than 50 won or lost for this tag.
|
|
if field == 'tag_id' and (frequency['won_count'] + frequency['lost_count']) < 50:
|
|
continue
|
|
|
|
if frequency.team_id:
|
|
team_result = result[frequency.team_id.id]
|
|
team_result[field][value] = {'won': frequency['won_count'], 'lost': frequency['lost_count']}
|
|
team_result[field]['won_total'] += frequency['won_count']
|
|
team_result[field]['lost_total'] += frequency['lost_count']
|
|
|
|
if value not in result[-1][field]:
|
|
result[-1][field][value] = {'won': 0, 'lost': 0}
|
|
result[-1][field][value]['won'] += frequency['won_count']
|
|
result[-1][field][value]['lost'] += frequency['lost_count']
|
|
result[-1][field]['won_total'] += frequency['won_count']
|
|
result[-1][field]['lost_total'] += frequency['lost_count']
|
|
|
|
# Get all won, lost and total count for all records in frequencies per team_id
|
|
for team_id in result:
|
|
result[team_id]['team_won'], \
|
|
result[team_id]['team_lost'], \
|
|
result[team_id]['team_total'] = self._pls_get_won_lost_total_count(result[team_id])
|
|
|
|
save_team_id = None
|
|
p_won, p_lost = 1, 1
|
|
for lead_id, lead_values in leads_values_dict.items():
|
|
# if stage_id is null, return 0 and bypass computation
|
|
lead_fields = [value[0] for value in lead_values.get('values', [])]
|
|
if not 'stage_id' in lead_fields:
|
|
lead_probabilities[lead_id] = 0
|
|
continue
|
|
# if lead stage is won, return 100
|
|
elif lead_id in won_leads:
|
|
lead_probabilities[lead_id] = 100
|
|
continue
|
|
|
|
# team_id not in frequency Table -> convert to -1
|
|
lead_team_id = lead_values['team_id'] if lead_values['team_id'] in result else -1
|
|
if lead_team_id != save_team_id:
|
|
save_team_id = lead_team_id
|
|
team_won = result[save_team_id]['team_won']
|
|
team_lost = result[save_team_id]['team_lost']
|
|
team_total = result[save_team_id]['team_total']
|
|
# if one count = 0, we cannot compute lead probability
|
|
if not team_won or not team_lost:
|
|
continue
|
|
p_won = team_won / team_total
|
|
p_lost = team_lost / team_total
|
|
|
|
# 2. Compute won and lost score using each variable's individual probability
|
|
s_lead_won, s_lead_lost = p_won, p_lost
|
|
for field, value in lead_values['values']:
|
|
field_result = result.get(save_team_id, {}).get(field)
|
|
value = value.origin if hasattr(value, 'origin') else value
|
|
value_result = field_result.get(str(value)) if field_result else False
|
|
if value_result:
|
|
total_won = team_won if field == 'stage_id' else field_result['won_total']
|
|
total_lost = team_lost if field == 'stage_id' else field_result['lost_total']
|
|
|
|
# if one count = 0, we cannot compute lead probability
|
|
if not total_won or not total_lost:
|
|
continue
|
|
s_lead_won *= value_result['won'] / total_won
|
|
s_lead_lost *= value_result['lost'] / total_lost
|
|
|
|
# 3. Compute Probability to win
|
|
probability = s_lead_won / (s_lead_won + s_lead_lost)
|
|
lead_probabilities[lead_id] = min(max(round(100 * probability, 2), 0.01), 99.99)
|
|
return lead_probabilities
|
|
|
|
# ---------------------------------
|
|
# PLS: Live Increment
|
|
# ---------------------------------
|
|
def _pls_increment_frequencies(self, from_state=None, to_state=None):
|
|
"""
|
|
When losing or winning a lead, this method is called to increment each PLS parameter related to the lead
|
|
in won_count (if won) or in lost_count (if lost).
|
|
|
|
This method is also used when reactivating a mistakenly lost lead (using the decrement argument).
|
|
In this case, the lost count should be de-increment by 1 for each PLS parameter linked to the lead.
|
|
|
|
Live increment must be done before writing the new values because we need to know the state change (from and to).
|
|
This would not be an issue for the reach won or reach lost as we just need to increment the frequencies with the
|
|
final state of the lead.
|
|
This issue is when the lead leaves a closed state because once the new values have been writen, we do not know
|
|
what was the previous state that we need to decrement.
|
|
This is why 'is_won' and 'decrement' parameters are used to describe the from / to change of its state.
|
|
"""
|
|
new_frequencies_by_team, existing_frequencies_by_team = self._pls_prepare_update_frequency_table(target_state=from_state or to_state)
|
|
|
|
# update frequency table
|
|
self._pls_update_frequency_table(new_frequencies_by_team, 1 if to_state else -1,
|
|
existing_frequencies_by_team=existing_frequencies_by_team)
|
|
|
|
# ---------------------------------
|
|
# PLS: One shot rebuild
|
|
# ---------------------------------
|
|
def _cron_update_automated_probabilities(self):
|
|
""" This cron will :
|
|
- rebuild the lead scoring frequency table
|
|
- recompute all the automated_probability and align probability if both were aligned
|
|
"""
|
|
cron_start_date = datetime.now()
|
|
self._rebuild_pls_frequency_table()
|
|
self._update_automated_probabilities()
|
|
_logger.info("Predictive Lead Scoring : Cron duration = %d seconds" % ((datetime.now() - cron_start_date).total_seconds()))
|
|
|
|
def _rebuild_pls_frequency_table(self):
|
|
# Clear the frequencies table (in sql to speed up the cron)
|
|
try:
|
|
self.check_access_rights('unlink')
|
|
except AccessError:
|
|
raise UserError(_("You don't have the access needed to run this cron."))
|
|
else:
|
|
self._cr.execute('TRUNCATE TABLE crm_lead_scoring_frequency')
|
|
|
|
new_frequencies_by_team, unused = self._pls_prepare_update_frequency_table(rebuild=True)
|
|
# update frequency table
|
|
self._pls_update_frequency_table(new_frequencies_by_team, 1)
|
|
|
|
_logger.info("Predictive Lead Scoring : crm.lead.scoring.frequency table rebuilt")
|
|
|
|
def _update_automated_probabilities(self):
|
|
""" Recompute all the automated_probability (and align probability if both were aligned) for all the leads
|
|
that are active (not won, nor lost).
|
|
|
|
For performance matter, as there can be a huge amount of leads to recompute, this cron proceed by batch.
|
|
Each batch is performed into its own transaction, in order to minimise the lock time on the lead table
|
|
(and to avoid complete lock if there was only 1 transaction that would last for too long -> several minutes).
|
|
If a concurrent update occurs, it will simply be put in the queue to get the lock.
|
|
"""
|
|
pls_start_date = self._pls_get_safe_start_date()
|
|
if not pls_start_date:
|
|
return
|
|
|
|
# 1. Get all the leads to recompute created after pls_start_date that are nor won nor lost
|
|
# (Won : probability = 100 | Lost : probability = 0 or inactive. Here, inactive won't be returned anyway)
|
|
# Get also all the lead without probability --> These are the new leads. Activate auto probability on them.
|
|
pending_lead_domain = [
|
|
'&',
|
|
'&',
|
|
('stage_id', '!=', False), ('create_date', '>=', pls_start_date),
|
|
'|',
|
|
('probability', '=', False),
|
|
'&',
|
|
('probability', '<', 100), ('probability', '>', 0)
|
|
]
|
|
leads_to_update = self.env['crm.lead'].search(pending_lead_domain)
|
|
leads_to_update_count = len(leads_to_update)
|
|
|
|
# 2. Compute by batch to avoid memory error
|
|
lead_probabilities = {}
|
|
for i in range(0, leads_to_update_count, PLS_COMPUTE_BATCH_STEP):
|
|
leads_to_update_part = leads_to_update[i:i + PLS_COMPUTE_BATCH_STEP]
|
|
lead_probabilities.update(leads_to_update_part._pls_get_naive_bayes_probabilities(batch_mode=True))
|
|
_logger.info("Predictive Lead Scoring : New automated probabilities computed")
|
|
|
|
# 3. Group by new probability to reduce server roundtrips when executing the update
|
|
probability_leads = defaultdict(list)
|
|
for lead_id, probability in sorted(lead_probabilities.items()):
|
|
probability_leads[probability].append(lead_id)
|
|
|
|
# 4. Update automated_probability (+ probability if both were equal)
|
|
update_sql = """UPDATE crm_lead
|
|
SET automated_probability = %s,
|
|
probability = CASE WHEN (probability = automated_probability OR probability is null)
|
|
THEN (%s)
|
|
ELSE (probability)
|
|
END
|
|
WHERE id in %s"""
|
|
|
|
# Update by a maximum number of leads at the same time, one batch by transaction :
|
|
# - avoid memory errors
|
|
# - avoid blocking the table for too long with a too big transaction
|
|
transactions_count, transactions_failed_count = 0, 0
|
|
cron_update_lead_start_date = datetime.now()
|
|
auto_commit = not getattr(threading.current_thread(), 'testing', False)
|
|
self.flush_model()
|
|
for probability, probability_lead_ids in probability_leads.items():
|
|
for lead_ids_current in tools.split_every(PLS_UPDATE_BATCH_STEP, probability_lead_ids):
|
|
transactions_count += 1
|
|
try:
|
|
self.env.cr.execute(update_sql, (probability, probability, tuple(lead_ids_current)))
|
|
# auto-commit except in testing mode
|
|
if auto_commit:
|
|
self.env.cr.commit()
|
|
except Exception as e:
|
|
_logger.warning("Predictive Lead Scoring : update transaction failed. Error: %s" % e)
|
|
transactions_failed_count += 1
|
|
self.invalidate_model()
|
|
|
|
_logger.info(
|
|
"Predictive Lead Scoring : All automated probabilities updated (%d leads / %d transactions (%d failed) / %d seconds)" % (
|
|
leads_to_update_count,
|
|
transactions_count,
|
|
transactions_failed_count,
|
|
(datetime.now() - cron_update_lead_start_date).total_seconds(),
|
|
)
|
|
)
|
|
|
|
# ---------------------------------
|
|
# PLS: Common parts for both mode
|
|
# ---------------------------------
|
|
def _pls_prepare_update_frequency_table(self, rebuild=False, target_state=False):
|
|
"""
|
|
This method is common to Live Increment or Full Rebuild mode, as it shares the main steps.
|
|
This method will prepare the frequency dict needed to update the frequency table:
|
|
- New frequencies: frequencies that we need to add in the frequency table.
|
|
- Existing frequencies: frequencies that are already in the frequency table.
|
|
In rebuild mode, only the new frequencies are needed as existing frequencies are truncated.
|
|
For each team, each dict contains the frequency in won and lost for each field/value couple
|
|
of the target leads.
|
|
Target leads are :
|
|
- in Live increment mode : given ongoing leads (self)
|
|
- in Full rebuild mode : all the closed (won and lost) leads in the DB.
|
|
During the frequencies update, with both new and existing frequencies, we can split frequencies to update
|
|
and frequencies to add. If a field/value couple already exists in the frequency table, we just update it.
|
|
Otherwise, we need to insert a new one.
|
|
"""
|
|
# Keep eligible leads
|
|
pls_start_date = self._pls_get_safe_start_date()
|
|
if not pls_start_date:
|
|
return {}, {}
|
|
|
|
if rebuild: # rebuild will treat every closed lead in DB, increment will treat current ongoing leads
|
|
pls_leads = self
|
|
else:
|
|
# Only treat leads created after the PLS start Date
|
|
pls_leads = self.filtered(
|
|
lambda lead: fields.Date.to_date(pls_start_date) <= fields.Date.to_date(lead.create_date))
|
|
if not pls_leads:
|
|
return {}, {}
|
|
|
|
# Extract target leads values
|
|
if rebuild: # rebuild is ok
|
|
domain = [
|
|
'&',
|
|
('create_date', '>=', pls_start_date),
|
|
'|',
|
|
('probability', '=', 100),
|
|
'&',
|
|
('probability', '=', 0), ('active', '=', False)
|
|
]
|
|
team_ids = self.env['crm.team'].with_context(active_test=False).search([]).ids + [0] # If team_id is unset, consider it as team 0
|
|
else: # increment
|
|
domain = [('id', 'in', pls_leads.ids)]
|
|
team_ids = pls_leads.mapped('team_id').ids + [0]
|
|
|
|
leads_values_dict = pls_leads._pls_get_lead_pls_values(domain=domain)
|
|
|
|
# split leads values by team_id
|
|
# get current frequencies related to the target leads
|
|
leads_frequency_values_by_team = dict((team_id, []) for team_id in team_ids)
|
|
leads_pls_fields = set() # ensure to keep each field unique (can have multiple tag_id leads_values_dict)
|
|
for lead_id, values in leads_values_dict.items():
|
|
team_id = values.get('team_id', 0) # If team_id is unset, consider it as team 0
|
|
lead_frequency_values = {'count': 1}
|
|
for field, value in values['values']:
|
|
if field != "probability": # was added to lead values in batch mode to know won/lost state, but is not a pls fields.
|
|
leads_pls_fields.add(field)
|
|
else: # extract lead probability - needed to increment tag_id frequency. (proba always before tag_id)
|
|
lead_probability = value
|
|
if field == 'tag_id': # handle tag_id separatelly (as in One Shot rebuild mode)
|
|
leads_frequency_values_by_team[team_id].append({field: value, 'count': 1, 'probability': lead_probability})
|
|
else:
|
|
lead_frequency_values[field] = value
|
|
leads_frequency_values_by_team[team_id].append(lead_frequency_values)
|
|
leads_pls_fields = sorted(leads_pls_fields)
|
|
|
|
# get new frequencies
|
|
new_frequencies_by_team = {}
|
|
for team_id in team_ids:
|
|
# prepare fields and tag values for leads by team
|
|
new_frequencies_by_team[team_id] = self._pls_prepare_frequencies(
|
|
leads_frequency_values_by_team[team_id], leads_pls_fields, target_state=target_state)
|
|
|
|
# get existing frequencies
|
|
existing_frequencies_by_team = {}
|
|
if not rebuild: # there is no existing frequency in rebuild mode as they were all deleted.
|
|
# read all fields to get everything in memory in one query (instead of having query + prefetch)
|
|
existing_frequencies = self.env['crm.lead.scoring.frequency'].search_read(
|
|
['&', ('variable', 'in', leads_pls_fields),
|
|
'|', ('team_id', 'in', pls_leads.mapped('team_id').ids), ('team_id', '=', False)])
|
|
for frequency in existing_frequencies:
|
|
team_id = frequency['team_id'][0] if frequency.get('team_id') else 0
|
|
if team_id not in existing_frequencies_by_team:
|
|
existing_frequencies_by_team[team_id] = dict((field, {}) for field in leads_pls_fields)
|
|
|
|
existing_frequencies_by_team[team_id][frequency['variable']][frequency['value']] = {
|
|
'frequency_id': frequency['id'],
|
|
'won': frequency['won_count'],
|
|
'lost': frequency['lost_count']
|
|
}
|
|
|
|
return new_frequencies_by_team, existing_frequencies_by_team
|
|
|
|
def _pls_update_frequency_table(self, new_frequencies_by_team, step, existing_frequencies_by_team=None):
|
|
""" Create / update the frequency table in a cross company way, per team_id"""
|
|
values_to_update = {}
|
|
values_to_create = []
|
|
if not existing_frequencies_by_team:
|
|
existing_frequencies_by_team = {}
|
|
# build the create multi + frequencies to update
|
|
for team_id, new_frequencies in new_frequencies_by_team.items():
|
|
for field, value in new_frequencies.items():
|
|
# frequency already present ?
|
|
current_frequencies = existing_frequencies_by_team.get(team_id, {})
|
|
for param, result in value.items():
|
|
current_frequency_for_couple = current_frequencies.get(field, {}).get(param, {})
|
|
# If frequency already present : UPDATE IT
|
|
if current_frequency_for_couple:
|
|
new_won = current_frequency_for_couple['won'] + (result['won'] * step)
|
|
new_lost = current_frequency_for_couple['lost'] + (result['lost'] * step)
|
|
# ensure to have always positive frequencies
|
|
values_to_update[current_frequency_for_couple['frequency_id']] = {
|
|
'won_count': new_won if new_won > 0 else 0.1,
|
|
'lost_count': new_lost if new_lost > 0 else 0.1
|
|
}
|
|
continue
|
|
|
|
# Else, CREATE a new frequency record.
|
|
# We add + 0.1 in won and lost counts to avoid zero frequency issues
|
|
# should be +1 but it weights too much on small recordset.
|
|
values_to_create.append({
|
|
'variable': field,
|
|
'value': param,
|
|
'won_count': result['won'] + 0.1,
|
|
'lost_count': result['lost'] + 0.1,
|
|
'team_id': team_id if team_id else None # team_id = 0 means no team_id
|
|
})
|
|
|
|
LeadScoringFrequency = self.env['crm.lead.scoring.frequency'].sudo()
|
|
for frequency_id, values in values_to_update.items():
|
|
LeadScoringFrequency.browse(frequency_id).write(values)
|
|
|
|
if values_to_create:
|
|
LeadScoringFrequency.create(values_to_create)
|
|
|
|
# ---------------------------------
|
|
# Utility Tools for PLS
|
|
# ---------------------------------
|
|
|
|
# PLS: Config Parameters
|
|
# ---------------------
|
|
def _pls_get_safe_start_date(self):
|
|
""" As config_parameters does not accept Date field,
|
|
we get directly the date formated string stored into the Char config field,
|
|
as we directly use this string in the sql queries.
|
|
To avoid sql injections when using this config param,
|
|
we ensure the date string can be effectively a date."""
|
|
str_date = self.env['ir.config_parameter'].sudo().get_param('crm.pls_start_date')
|
|
if not fields.Date.to_date(str_date):
|
|
return False
|
|
return str_date
|
|
|
|
def _pls_get_safe_fields(self):
|
|
""" As config_parameters does not accept M2M field,
|
|
we the fields from the formated string stored into the Char config field.
|
|
To avoid sql injections when using that list, we return only the fields
|
|
that are defined on the model. """
|
|
pls_fields_config = self.env['ir.config_parameter'].sudo().get_param('crm.pls_fields')
|
|
pls_fields = pls_fields_config.split(',') if pls_fields_config else []
|
|
pls_safe_fields = [field for field in pls_fields if field in self._fields.keys()]
|
|
return pls_safe_fields
|
|
|
|
# Compute Automated Probability Tools
|
|
# -----------------------------------
|
|
def _pls_get_won_lost_total_count(self, team_results):
|
|
""" Get all won and all lost + total :
|
|
first stage can be used to know how many lost and won there is
|
|
as won count are equals for all stage
|
|
and first stage is always incremented in lost_count
|
|
:param frequencies: lead_scoring_frequencies
|
|
:return: won count, lost count and total count for all records in frequencies
|
|
"""
|
|
# TODO : check if we need to handle specific team_id stages [for lost count] (if first stage in sequence is team_specific)
|
|
first_stage_id = self.env['crm.stage'].search([('team_id', '=', False)], order='sequence, id', limit=1)
|
|
if str(first_stage_id.id) not in team_results.get('stage_id', []):
|
|
return 0, 0, 0
|
|
stage_result = team_results['stage_id'][str(first_stage_id.id)]
|
|
return stage_result['won'], stage_result['lost'], stage_result['won'] + stage_result['lost']
|
|
|
|
# PLS: Rebuild Frequency Table Tools
|
|
# ----------------------------------
|
|
def _pls_prepare_frequencies(self, lead_values, leads_pls_fields, target_state=None):
|
|
"""new state is used when getting frequencies for leads that are changing to lost or won.
|
|
Stays none if we are checking frequencies for leads already won or lost."""
|
|
pls_fields = leads_pls_fields.copy()
|
|
frequencies = dict((field, {}) for field in pls_fields)
|
|
|
|
stage_ids = self.env['crm.stage'].search_read([], ['sequence', 'name', 'id'], order='sequence, id')
|
|
stage_sequences = {stage['id']: stage['sequence'] for stage in stage_ids}
|
|
|
|
# Increment won / lost frequencies by criteria (field / value couple)
|
|
for values in lead_values:
|
|
if target_state: # ignore probability values if target state (as probability is the old value)
|
|
won_count = values['count'] if target_state == 'won' else 0
|
|
lost_count = values['count'] if target_state == 'lost' else 0
|
|
else:
|
|
won_count = values['count'] if values.get('probability', 0) == 100 else 0
|
|
lost_count = values['count'] if values.get('probability', 1) == 0 else 0
|
|
|
|
if 'tag_id' in values:
|
|
frequencies = self._pls_increment_frequency_dict(frequencies, 'tag_id', values['tag_id'], won_count, lost_count)
|
|
continue
|
|
|
|
# Else, treat other fields
|
|
if 'tag_id' in pls_fields: # tag_id already treated here above.
|
|
pls_fields.remove('tag_id')
|
|
for field in pls_fields:
|
|
if field not in values:
|
|
continue
|
|
value = values[field]
|
|
if value or field in ('email_state', 'phone_state'):
|
|
if field == 'stage_id':
|
|
if won_count: # increment all stages if won
|
|
stages_to_increment = [stage['id'] for stage in stage_ids]
|
|
else: # increment only current + previous stages if lost
|
|
current_stage_sequence = stage_sequences[value]
|
|
stages_to_increment = [stage['id'] for stage in stage_ids if stage['sequence'] <= current_stage_sequence]
|
|
for stage_id in stages_to_increment:
|
|
frequencies = self._pls_increment_frequency_dict(frequencies, field, stage_id, won_count, lost_count)
|
|
else:
|
|
frequencies = self._pls_increment_frequency_dict(frequencies, field, value, won_count, lost_count)
|
|
|
|
return frequencies
|
|
|
|
def _pls_increment_frequency_dict(self, frequencies, field, value, won, lost):
|
|
value = str(value) # Ensure we will always compare strings.
|
|
if value not in frequencies[field]:
|
|
frequencies[field][value] = {'won': won, 'lost': lost}
|
|
else:
|
|
frequencies[field][value]['won'] += won
|
|
frequencies[field][value]['lost'] += lost
|
|
return frequencies
|
|
|
|
# Common PLS Tools
|
|
# ----------------
|
|
def _pls_get_lead_pls_values(self, domain=[]):
|
|
"""
|
|
This methods builds a dict where, for each lead in self or matching the given domain,
|
|
we will get a list of field/value couple.
|
|
Due to onchange and create, we don't always have the id of the lead to recompute.
|
|
When we update few records (one, typically) with onchanges, we build the lead_values (= couple field/value)
|
|
using the ORM.
|
|
To speed up the computation and avoid making too much DB read inside loops,
|
|
we can give a domain to make sql queries to bypass the ORM.
|
|
This domain will be used in sql queries to get the values for every lead matching the domain.
|
|
:param domain: If set, we get all the leads values via unique sql queries (one for tags, one for other fields),
|
|
using the given domain on leads.
|
|
If not set, get lead values lead by lead using the ORM.
|
|
:return: {lead_id: [(field1: value1), (field2: value2), ...], ...}
|
|
"""
|
|
leads_values_dict = OrderedDict()
|
|
pls_fields = ["stage_id", "team_id"] + self._pls_get_safe_fields()
|
|
|
|
# Check if tag_ids is in the pls_fields and removed it from the list. The tags will be managed separately.
|
|
use_tags = 'tag_ids' in pls_fields
|
|
if use_tags:
|
|
pls_fields.remove('tag_ids')
|
|
|
|
if domain:
|
|
# active_test = False as domain should take active into 'active' field it self
|
|
from_clause, where_clause, where_params = self.env['crm.lead'].with_context(active_test=False)._where_calc(domain).get_sql()
|
|
str_fields = ", ".join(["{}"] * len(pls_fields))
|
|
args = [sql.Identifier(field) for field in pls_fields]
|
|
|
|
# Get leads values
|
|
self.flush_model()
|
|
query = """SELECT id, probability, %s
|
|
FROM %s
|
|
WHERE %s order by team_id asc, id desc"""
|
|
query = sql.SQL(query % (str_fields, from_clause, where_clause)).format(*args)
|
|
self._cr.execute(query, where_params)
|
|
lead_results = self._cr.dictfetchall()
|
|
|
|
if use_tags:
|
|
# Get tags values
|
|
query = """SELECT crm_lead.id as lead_id, t.id as tag_id
|
|
FROM %s
|
|
LEFT JOIN crm_tag_rel rel ON crm_lead.id = rel.lead_id
|
|
LEFT JOIN crm_tag t ON rel.tag_id = t.id
|
|
WHERE %s order by crm_lead.team_id asc, crm_lead.id"""
|
|
args.append(sql.Identifier('tag_id'))
|
|
query = sql.SQL(query % (from_clause, where_clause)).format(*args)
|
|
self._cr.execute(query, where_params)
|
|
tag_results = self._cr.dictfetchall()
|
|
else:
|
|
tag_results = []
|
|
|
|
# get all (variable, value) couple for all in self
|
|
for lead in lead_results:
|
|
lead_values = []
|
|
for field in pls_fields + ['probability']: # add probability as used in _pls_prepare_frequencies (needed in rebuild mode)
|
|
value = lead[field]
|
|
if field == 'team_id': # ignore team_id as stored separately in leads_values_dict[lead_id][team_id]
|
|
continue
|
|
if value or field == 'probability': # 0 is a correct value for probability
|
|
lead_values.append((field, value))
|
|
elif field in ('email_state', 'phone_state'): # As ORM reads 'None' as 'False', do the same here
|
|
lead_values.append((field, False))
|
|
leads_values_dict[lead['id']] = {'values': lead_values, 'team_id': lead['team_id'] or 0}
|
|
|
|
for tag in tag_results:
|
|
if tag['tag_id']:
|
|
leads_values_dict[tag['lead_id']]['values'].append(('tag_id', tag['tag_id']))
|
|
return leads_values_dict
|
|
else:
|
|
for lead in self:
|
|
lead_values = []
|
|
for field in pls_fields:
|
|
if field == 'team_id': # ignore team_id as stored separately in leads_values_dict[lead_id][team_id]
|
|
continue
|
|
value = lead[field].id if isinstance(lead[field], models.BaseModel) else lead[field]
|
|
if value or field in ('email_state', 'phone_state'):
|
|
lead_values.append((field, value))
|
|
if use_tags:
|
|
for tag in lead.tag_ids:
|
|
lead_values.append(('tag_id', tag.id))
|
|
leads_values_dict[lead.id] = {'values': lead_values, 'team_id': lead['team_id'].id}
|
|
return leads_values_dict
|