# Copyright(C) 2019 Powens # # This file is part of a woob module. # # This woob module is free software: you can redistribute it and/or modify # it under the terms of the GNU Lesser General Public License as published by # the Free Software Foundation, either version 3 of the License, or # (at your option) any later version. # # This woob module is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU Lesser General Public License for more details. # # You should have received a copy of the GNU Lesser General Public License # along with this woob module. If not, see . # flake8: compatible import re from hashlib import sha1 from html import unescape from woob.browser.elements import DictElement, ItemElement, ListElement, method from woob.browser.filters.html import Attr, Link from woob.browser.filters.json import Dict from woob.browser.filters.standard import CleanDecimal, CleanText, Coalesce, Currency, Date, Env, Field, Format, Regexp from woob.browser.pages import HTMLPage, JsonPage, LoggedPage, PartialHTMLPage, RawPage from woob.capabilities.address import PostalAddress from woob.capabilities.base import NotAvailable from woob.capabilities.bill import Bill, Document, DocumentTypes, Subscription from woob.capabilities.profile import Person from woob.exceptions import BrowserUnavailable from woob.tools.date import parse_french_date from woob.tools.json import json class LoginPage(HTMLPage): def is_here(self): return self.doc.xpath('//form[contains(@id, "CompteForm")]') or self.doc.xpath('//div[@id="loginPage"]') def login(self, username, password, _ct): form = self.get_form(id='connexioncompte_2connexionCompteForm') form['connexioncompte_2numSecuriteSociale'] = username form['connexioncompte_2codeConfidentiel'] = password form['_ct'] = _ct form.submit() def get_error_message(self): return Coalesce( CleanText('//div[@id="loginPage"]//div[has-class("zone-alerte") and not(has-class("hidden"))]/span'), CleanText('//div[@class="centrepage compte_bloque"]//p[@class="msg_erreur"]'), default="" )(self.doc) def is_direct_login_disabled(self): info_message = ( 'Suite à une opération de maintenance, cliquez sur FranceConnect et ' + 'utilisez vos identifiants ameli pour accéder à votre compte.' ) return info_message in CleanText('//div[@id="idBlocCnx"]/div/p')(self.doc) class LoginContinuePage(HTMLPage): def get_action_needed_url(self): # Url we need is in a content attribute formatted like # "0; url=http..." return Regexp( Attr('//div[@class="wlp-bighorn-window-content"]//meta', 'content'), r'url=(.*)', default='', )(self.doc) class NewPasswordPage(HTMLPage): pass class AmeliConnectOpenIdPage(LoginPage): def login(self, username, password): """ Submits the form to login with username / password. """ form = self.get_form(id='connexioncompte_2connexionCompteForm') form['user'] = username form['password'] = password form.submit() def request_otp(self): """ Submits the form to request an OTP. """ form = self.get_form(id='connexioncompte_2connexionCompteForm', submit='//input[@type="submit" and @name="envoiOTP"]') form['authStep'] = 'ENVOI_OTP' form.submit() def otp_step(self): """ Returns OTP auth step ('', OTP_NECESSAIRE, SAISIE_OTP). :rtype: str """ form = self.get_form(id='connexioncompte_2connexionCompteForm', submit='//input[@type="submit" and @id="id_r_cnx_btn_submit"]') return form['authStep'] class CtPage(RawPage): # the page contains only _ct value def get_ct_value(self): return re.search(r'_ct:(.*)', self.text).group(1) class RedirectPage(LoggedPage, HTMLPage): REFRESH_MAX = 0 REFRESH_XPATH = '//meta[@http-equiv="refresh"]' class CguPage(LoggedPage, HTMLPage): def get_cgu_message(self): return CleanText('//div[@class="page_nouvelles_cgus"]/p')(self.doc) class ErrorPage(HTMLPage): def on_load(self): # message is: "Oups... votre compte ameli est momentanément indisponible. Il sera de retour en pleine forme très bientôt." # nothing we can do, but retry later raise BrowserUnavailable(unescape(CleanText('//div[@class="mobile"]/p')(self.doc))) class SubscriptionPage(LoggedPage, HTMLPage): def get_subscription(self): sub = Subscription() # DON'T TAKE social security number for id because it's a very confidential data, take birth date instead sub.id = CleanText('//button[@id="idAssure"]//td[@class="dateNaissance"]')(self.doc).replace('/', '') sub.label = sub.subscriber = CleanText('//div[@id="pageAssure"]//span[@class="NomEtPrenomLabel"]')(self.doc) return sub @method class get_profile(ItemElement): klass = Person # Other recipients can also be on this page. # The first one corresponds to the logged user. obj_name = CleanText('(//span[@class="NomEtPrenomLabel"])[1]') obj_birth_date = Date(CleanText('(//td[@class="dateNaissance"]/span)[1]'), parse_func=parse_french_date) obj_phone = CleanText(Coalesce( '//div[@class="infoGauche"][normalize-space()="Téléphone portable"]/following-sibling::div/span', '//div[@class="infoGauche"][normalize-space()="Téléphone fixe"]/following-sibling::div/span', default=NotAvailable )) class obj_postal_address(ItemElement): klass = PostalAddress def parse(self, obj): full_address = CleanText( '//div[@class="infoGauche"][normalize-space()="Adresse postale"]/following-sibling::div/span' )(self) self.env['full_address'] = full_address m = re.search(r'(\d{1,4}.*) (\d{5}) (.*)', full_address) if m: street, postal_code, city = m.groups() self.env['street'] = street self.env['postal_code'] = postal_code self.env['city'] = city obj_full_address = Env('full_address', default=NotAvailable) obj_street = Env('street', default=NotAvailable) obj_postal_code = Env('postal_code', default=NotAvailable) obj_city = Env('city', default=NotAvailable) class DocumentsDetailsPage(LoggedPage, PartialHTMLPage): ENCODING = 'utf-8' def build_doc(self, content): res = json.loads(content) return super(DocumentsDetailsPage, self).build_doc(res['tableauPaiement'].encode('utf-8')) @method class iter_documents(ListElement): item_xpath = '//ul[@id="unordered_list"]//li[has-class("rowitem")]' class item(ItemElement): klass = Bill def obj_id(self): _id = Regexp(Field('url'), r'idPaiement=(.*)')(self) # idPaiement is very long, about 192 char, and sometimes they change it, (even existing id) # to make it much longer, (because 120 char wasn't enough apparently) return '%s_%s' % (Env('subid')(self), sha1(_id.encode('utf-8')).hexdigest()) obj_label = CleanText('.//div[has-class("col-label")]') obj_total_price = CleanDecimal.French('.//div[has-class("col-montant")]/span') obj_currency = Currency('.//div[has-class("col-montant")]/span') obj_url = Link('.//div[@class="col-download"]/a') obj_format = 'pdf' def obj_date(self): year = Regexp( CleanText('./preceding-sibling::li[@class="rowdate"]//span[@class="mois"]'), r'(\d+)' )(self) day_month = CleanText('.//div[has-class("col-date")]/span')(self) return parse_french_date(day_month + ' ' + year) class DocumentsFirstSummaryPage(LoggedPage, HTMLPage): @method class iter_documents(ListElement): item_xpath = '//ul[@id="unordered_list"]//li[@class="rowdate" and .//span[@class="blocTelecharger"]]' class item(ItemElement): klass = Document obj_type = DocumentTypes.BILL obj_label = Format('%s %s', CleanText('.//span[@class="libelle"]'), CleanText('.//span[@class="mois"]')) obj_url = Link('.//div[@class="col-telechargement"]//a') obj_format = 'pdf' def obj_date(self): year = Regexp(CleanText('.//span[@class="mois"]'), r'(\d+)')(self) month = Regexp(CleanText('.//span[@class="mois"]'), r'(\D+)')(self) return parse_french_date(month + ' ' + year) def obj_id(self): year = Regexp(CleanText('.//span[@class="mois"]'), r'(\d+)')(self) month = Regexp(CleanText('.//span[@class="mois"]'), r'(\D+)')(self) return '%s_%s' % (Env('subid')(self), parse_french_date(month + ' ' + year).strftime('%Y%m')) class DocumentsLastSummaryPage(LoggedPage, JsonPage): @method class iter_documents(DictElement): def find_elements(self): for doc in self.el['listeDecomptes']: if doc['montant']: yield doc class item(ItemElement): klass = Document obj_type = DocumentTypes.BILL obj_url = Dict('urlPDF') obj_format = 'pdf' obj_label = Format('Relevé mensuel %s', CleanText(Dict('mois'))) def obj_date(self): year = Regexp(CleanText(Dict('mois')), r'(\d+)')(self) month = Regexp(CleanText(Dict('mois')), r'(\D+)')(self) return parse_french_date(month + ' ' + year) def obj_id(self): year = Regexp(CleanText(Dict('mois')), r'(\d+)')(self) month = Regexp(CleanText(Dict('mois')), r'(\D+)')(self) return '%s_%s' % (Env('subid')(self), parse_french_date(month + ' ' + year).strftime('%Y%m'))