MonScoDocEssai/sco_news.py
2021-02-03 22:00:41 +01:00

306 lines
10 KiB
Python

# -*- mode: python -*-
# -*- coding: utf-8 -*-
##############################################################################
#
# Gestion scolarite IUT
#
# Copyright (c) 1999 - 2021 Emmanuel Viennet. All rights reserved.
#
# This program is free software; you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation; either version 2 of the License, or
# (at your option) any later version.
#
# This program 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 General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program; if not, write to the Free Software
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
#
# Emmanuel Viennet emmanuel.viennet@viennet.net
#
##############################################################################
"""Gestions des "nouvelles"
"""
import PyRSS2Gen # pylint: disable=import-error
from cStringIO import StringIO
import datetime, re
import time
from stripogram import html2text, html2safehtml
from email.MIMEMultipart import ( # pylint: disable=no-name-in-module,import-error
MIMEMultipart,
)
from email.MIMEText import MIMEText # pylint: disable=no-name-in-module,import-error
from email.Header import Header # pylint: disable=no-name-in-module,import-error
from email import Encoders # pylint: disable=no-name-in-module,import-error
import notesdb as ndb # pylint: disable=unused-wildcard-import
from notes_log import log
import scolars
from sco_utils import SCO_ENCODING, SCO_ANNONCES_WEBSITE
import sco_formsemestre
import sco_moduleimpl
_scolar_news_editor = ndb.EditableTable(
"scolar_news",
"news_id",
("date", "authenticated_user", "type", "object", "text", "url"),
sortkey="date desc",
output_formators={"date": ndb.DateISOtoDMY},
input_formators={"date": ndb.DateDMYtoISO},
html_quote=False, # no user supplied data, needed to store html links
)
NEWS_INSCR = "INSCR" # inscription d'étudiants (object=None ou formsemestre_id)
NEWS_NOTE = "NOTES" # saisie note (object=moduleimpl_id)
NEWS_FORM = "FORM" # modification formation (object=formation_id)
NEWS_SEM = "SEM" # creation semestre (object=None)
NEWS_MISC = "MISC" # unused
NEWS_MAP = {
NEWS_INSCR: "inscription d'étudiants",
NEWS_NOTE: "saisie note",
NEWS_FORM: "modification formation",
NEWS_SEM: "création semestre",
NEWS_MISC: "opération", # unused
}
NEWS_TYPES = NEWS_MAP.keys()
scolar_news_create = _scolar_news_editor.create
scolar_news_list = _scolar_news_editor.list
_LAST_NEWS = {} # { (authuser_name, type, object) : time }
def add(context, REQUEST, typ, object=None, text="", url=None, max_frequency=False):
"""Ajoute une nouvelle.
Si max_frequency, ne genere pas 2 nouvelles identiques à moins de max_frequency
secondes d'intervalle.
"""
authuser_name = str(REQUEST.AUTHENTICATED_USER)
cnx = context.GetDBConnexion()
args = {
"authenticated_user": authuser_name,
"user_info": context.Users.user_info(user_name=authuser_name),
"type": typ,
"object": object,
"text": text,
"url": url,
}
log("news: %s" % args)
t = time.time()
if max_frequency:
last_news_time = _LAST_NEWS.get((authuser_name, typ, object), False)
if last_news_time and (t - last_news_time < max_frequency):
log("not recording")
return
_LAST_NEWS[(authuser_name, typ, object)] = t
_send_news_by_mail(context, args)
return scolar_news_create(cnx, args, has_uniq_values=False)
def scolar_news_summary(context, n=5):
"""Return last n news.
News are "compressed", ie redondant events are joined.
"""
cnx = context.GetDBConnexion()
cursor = cnx.cursor(cursor_factory=ndb.ScoDocCursor)
cursor.execute("select * from scolar_news order by date desc limit 100")
selected_news = {} # (type,object) : news dict
news = cursor.dictfetchall() # la plus récente d'abord
for r in reversed(news): # la plus ancienne d'abord
# si on a deja une news avec meme (type,object)
# et du meme jour, on la remplace
dmy = ndb.DateISOtoDMY(r["date"]) # round
key = (r["type"], r["object"], dmy)
selected_news[key] = r
news = selected_news.values()
# sort by date, descending
news.sort(lambda x, y: cmp(y["date"], x["date"]))
news = news[:n]
# mimic EditableTable.list output formatting:
for n in news:
n["date822"] = n["date"].strftime("%a, %d %b %Y %H:%M:%S %z")
# heure
n["hm"] = n["date"].strftime("%Hh%M")
n["rssdate"] = n["date"].strftime("%d/%m %Hh%M") # pour affichage
for k in n.keys():
if n[k] is None:
n[k] = ""
if _scolar_news_editor.output_formators.has_key(k):
n[k] = _scolar_news_editor.output_formators[k](n[k])
# date resumee
j, m = n["date"].split("/")[:2]
mois = scolars.abbrvmonthsnames[int(m) - 1]
n["formatted_date"] = "%s %s %s" % (j, mois, n["hm"])
# indication semestre si ajout notes:
infos = _get_formsemestre_infos_from_news(context, n)
if infos:
n["text"] += (
' (<a href="Notes/formsemestre_status?formsemestre_id=%(formsemestre_id)s">%(descr_sem)s</a>)'
% infos
)
n["text"] += (
" par "
+ context.Users.user_info(user_name=n["authenticated_user"])["nomcomplet"]
)
return news
def _get_formsemestre_infos_from_news(context, n):
"""Informations sur le semestre concerné par la nouvelle n
{} si inexistant
"""
formsemestre_id = None
if n["type"] == NEWS_INSCR:
formsemestre_id = n["object"]
elif n["type"] == NEWS_NOTE:
moduleimpl_id = n["object"]
mods = sco_moduleimpl.do_moduleimpl_list(context, moduleimpl_id=moduleimpl_id)
if not mods:
return {} # module does not exists anymore
mod = mods[0]
formsemestre_id = mod["formsemestre_id"]
if not formsemestre_id:
return {}
try:
sem = sco_formsemestre.get_formsemestre(context, formsemestre_id)
except:
# semestre n'existe plus
return {}
if sem["semestre_id"] > 0:
descr_sem = "S%d" % sem["semestre_id"]
else:
descr_sem = ""
if sem["modalite"]:
descr_sem += " " + sem["modalite"]
return {"formsemestre_id": formsemestre_id, "sem": sem, "descr_sem": descr_sem}
def scolar_news_summary_html(context, n=5, rssicon=None):
"""News summary, formated in HTML"""
news = scolar_news_summary(context, n=n)
if not news:
return ""
H = ['<div class="news"><span class="newstitle">Dernières opérations']
if rssicon: # 2020-12-30 plus utilisé
H.append('<a href="rssnews">' + rssicon + "</a>")
H.append('</span><ul class="newslist">')
for n in news:
H.append(
'<li class="newslist"><span class="newsdate">%(formatted_date)s</span><span class="newstext">%(text)s</span></li>'
% n
)
H.append("</ul>")
# Informations générales
H.append(
"""<div>
Pour être informé des évolutions de ScoDoc,
vous pouvez vous
<a class="stdlink" href="%s">
abonner à la liste de diffusion</a>.
</div>
"""
% SCO_ANNONCES_WEBSITE
)
H.append("</div>")
return "\n".join(H)
def scolar_news_summary_rss(context, title, sco_url, n=5):
"""rss feed for scolar news"""
news = scolar_news_summary(context, n=n)
items = []
for n in news:
text = html2text(n["text"])
items.append(
PyRSS2Gen.RSSItem(
title=unicode("%s %s" % (n["rssdate"], text), SCO_ENCODING),
link=sco_url + "/" + n["url"],
pubDate=n["date822"],
)
)
rss = PyRSS2Gen.RSS2(
title=unicode(title, SCO_ENCODING),
link=sco_url,
description=unicode(title, SCO_ENCODING),
lastBuildDate=datetime.datetime.now(),
items=items,
)
f = StringIO()
rss.write_xml(f, encoding=SCO_ENCODING)
f.seek(0)
data = f.read()
f.close()
return data
def _send_news_by_mail(context, n):
"""Notify by email"""
infos = _get_formsemestre_infos_from_news(context, n)
formsemestre_id = infos.get("formsemestre_id", None)
prefs = context.get_preferences(formsemestre_id=formsemestre_id)
destinations = prefs["emails_notifications"] or ""
destinations = [x.strip() for x in destinations.split(",")]
destinations = [x for x in destinations if x]
if not destinations:
return
#
txt = n["text"]
if infos:
txt += "\n\nSemestre %(titremois)s\n\n" % infos["sem"]
txt += (
"""<a href="Notes/formsemestre_status?formsemestre_id=%(formsemestre_id)s">%(descr_sem)s</a>
"""
% infos
)
txt += "\n\nEffectué par: %(nomcomplet)s\n" % n["user_info"]
txt = (
"\n"
+ txt
+ """\n
--- Ceci est un message de notification automatique issu de ScoDoc
--- vous recevez ce message car votre adresse est indiquée dans les paramètres de ScoDoc.
"""
)
# Transforme les URL en URL absolue
base = context.ScoURL()
txt = re.sub('href=.*?"', 'href="' + base + "/", txt)
# Transforme les liens HTML en texte brut: '<a href="url">texte</a>' devient 'texte: url'
# (si on veut des messages non html)
txt = re.sub(r'<a.*?href\s*=\s*"(.*?)".*?>(.*?)</a>', r"\2: \1", txt)
msg = MIMEMultipart()
msg["Subject"] = Header("[ScoDoc] " + NEWS_MAP.get(n["type"], "?"), SCO_ENCODING)
msg["From"] = prefs["email_from_addr"]
txt = MIMEText(txt, "plain", SCO_ENCODING)
msg.attach(txt)
for email_addr in destinations:
if email_addr:
del msg["To"]
msg["To"] = email_addr
# log('xxx mail: %s' % msg)
context.sendEmail(msg)