sito-hackit-18/plugins/talks.py

'''
Manage talks scheduling in a semantic way
'''


from __future__ import print_function
import os
import io
from functools import wraps
import logging
import re
import datetime
import shutil
from copy import copy
import locale
from contextlib import contextmanager
from babel.dates import format_date, format_datetime, format_time

from markdown import markdown
from docutils import nodes
from docutils.parsers.rst import directives, Directive
import six

from pelican import signals, generators
import jinja2

try:
    import ics
except ImportError:
    ICS_ENABLED = False
else:
    ICS_ENABLED = True
import unidecode
import dateutil

pelican = None  # This will be set during register()


def memoize(function):
    '''decorators to cache'''
    memo = {}

    @wraps(function)
    def wrapper(*args):
        if args in memo:
            return memo[args]
        else:
            rv = function(*args)
            memo[args] = rv
            return rv
    return wrapper


@contextmanager
def setlocale(name):
    saved = locale.setlocale(locale.LC_ALL)
    try:
        yield locale.setlocale(locale.LC_ALL, name)
    finally:
        locale.setlocale(locale.LC_ALL, saved)


@memoize
def get_talk_names():
    names = [name for name in os.listdir(pelican.settings['TALKS_PATH'])
             if not name.startswith('_') and
             get_talk_data(name) is not None
             ]
    names.sort()
    return names


def all_talks():
    return [get_talk_data(tn) for tn in get_talk_names()]


def unique_attr(iterable, attr):
    return {x[attr] for x in iterable
            if attr in x}


@memoize
def get_global_data():
    fname = os.path.join(pelican.settings['TALKS_PATH'], 'meta.yaml')
    if not os.path.isfile(fname):
        return None
    with io.open(fname, encoding='utf8') as buf:
        try:
            data = yaml.load(buf)
        except Exception:
            logging.exception("Syntax error reading %s; skipping", fname)
            return None
    if data is None:
        return None
    if 'startdate' not in data:
        logging.error("Missing startdate in global data")
        data['startdate'] = datetime.datetime.now()
    if 'rooms' not in data:
        data['rooms'] = {}
    return data


def _get_time_shift(timestring):
    ''' Il problema che abbiamo è che vogliamo dire che le 2 di notte del sabato sono in realtà "parte" del
    venerdì. Per farlo accettiamo orari che superano le 24, ad esempio 25.30 vuol dire 1.30.

    Questa funzione ritorna una timedelta in base alla stringa passata
    '''
    timeparts = re.findall(r'\d+', timestring)
    if not timeparts or len(timeparts) > 2:
        raise ValueError("Malformed time %s" % timestring)
    timeparts += [0,0]  # "padding" per essere sicuro ci siano anche [1] e [2]
    duration = datetime.timedelta(hours=int(timeparts[0]),
            minutes=int(timeparts[1]),
            seconds=int(timeparts[2]))
    if duration.total_seconds() > 3600 * 31 or duration.total_seconds() < 0:
        raise ValueError("Sforamento eccessivo: %d" % duration.hours)
    return duration



@memoize
def get_talk_data(talkname):
    fname = os.path.join(pelican.settings['TALKS_PATH'], talkname, 'meta.yaml')
    if not os.path.isfile(fname):
        return None
    with io.open(fname, encoding='utf8') as buf:
        try:
            data = yaml.load(buf)
        except Exception:
            logging.exception("Syntax error reading %s; skipping", fname)
            return None
    if data is None:
        return None
    try:
        gridstep = pelican.settings['TALKS_GRID_STEP']
        data.setdefault('nooverlap', [])
        if 'title' not in data:
            logging.warn("Talk <{}> has no `title` field".format(talkname))
            data['title'] = six.text_type(talkname)
        else:
            data['title'] = six.text_type(data['title'])
        if 'text' not in data:
            logging.warn("Talk <{}> has no `text` field".format(talkname))
            data['text'] = ''
        else:
            data['text'] = six.text_type(data['text'])
        if 'duration' not in data:
            logging.info("Talk <{}> has no `duration` field (50min used)"
                         .format(talkname))
            data['duration'] = 50
        data['duration'] = int(data['duration'])
        if data['duration'] < gridstep:
            logging.info("Talk <{}> lasts only {} minutes; changing to {}"
                         .format(talkname, data['duration'], gridstep))
            data['duration'] = gridstep
        if 'links' not in data or not data['links']:
            data['links'] = []
        if 'contacts' not in data or not data['contacts']:
            data['contacts'] = []
        if 'needs' not in data or not data['needs']:
            data['needs'] = []
        if 'room' not in data:
            logging.warn("Talk <{}> has no `room` field".format(talkname))
        else:
            if data['room'] in get_global_data()['rooms']:
                data['room'] = get_global_data()['rooms'][data['room']]
        if 'time' not in data or 'day' not in data:
            logging.warn("Talk <{}> has no `time` or `day`".format(talkname))
            if 'time' in data:
                del data['time']
            if 'day' in data:
                del data['day']
        else:
            data['day'] = get_global_data()['startdate'] + \
                    datetime.timedelta(days=data['day'])
            try:
                shift = _get_time_shift(str(data['time']))
            except ValueError:
                logging.error("Talk <%s> has malformed `time`", talkname)

            data['delta'] = shift
            data['time'] = datetime.datetime.combine(
                                      data['day'], datetime.time(0,0,0))
            data['time'] += shift
            data['time'] = data['time'].replace(tzinfo=dateutil.tz.gettz('Europe/Rome'))


        data['id'] = talkname
        resdir = os.path.join(pelican.settings['TALKS_PATH'], talkname,
                              pelican.settings['TALKS_ATTACHMENT_PATH'])
        if os.path.isdir(resdir) and os.listdir(resdir):
            data['resources'] = resdir
        return data
    except Exception:
        logging.exception("Error on talk %s", talkname)
        raise


def overlap(interval_a, interval_b):
    '''how many minutes do they overlap?'''
    return max(0, min(interval_a[1], interval_b[1]) -
               max(interval_a[0], interval_b[0]))


def get_talk_overlaps(name):
    data = get_talk_data(name)
    overlapping_talks = set()
    if 'time' not in data:
        return overlapping_talks
    start = int(data['time'].strftime('%s'))
    end = start + data['duration'] * 60
    for other in get_talk_names():
        if other == name:
            continue
        if 'time' not in get_talk_data(other):
            continue
        other_start = int(get_talk_data(other)['time'].strftime('%s'))
        other_end = other_start + get_talk_data(other)['duration'] * 60

        minutes = overlap((start, end), (other_start, other_end))
        if minutes > 0:
            overlapping_talks.add(other)
    return overlapping_talks


@memoize
def check_overlaps():
    for t in get_talk_names():
        over = get_talk_overlaps(t)
        noover = get_talk_data(t)['nooverlap']
        contacts = set(get_talk_data(t)['contacts'])
        for overlapping in over:
            if overlapping in noover or \
               set(get_talk_data(overlapping)['contacts']).\
               intersection(contacts):
                logging.warning('Talk %s overlaps with %s' % (t, overlapping))


@memoize
def jinja_env():
    env = jinja2.Environment(
        loader=jinja2.FileSystemLoader(os.path.join(
            pelican.settings['TALKS_PATH'],
            '_templates')),
        autoescape=True,
    )
    env.filters['markdown'] = lambda text: jinja2.Markup(markdown(text))
    env.filters['dateformat'] = format_date
    env.filters['datetimeformat'] = format_datetime
    env.filters['timeformat'] = format_time
    return env


class TalkListDirective(Directive):
    final_argument_whitespace = True
    has_content = True
    option_spec = {
        'lang': directives.unchanged
    }

    def run(self):
        lang = self.options.get('lang', 'C')
        tmpl = jinja_env().get_template('talk.html')

        def _sort_date(name):
            '''
            This function is a helper to sort talks by start date

            When no date is available, put at the beginning
            '''
            d = get_talk_data(name)
            room = d.get('room', '')
            time = d.get('time', datetime.datetime(1, 1, 1).replace(
                tzinfo=dateutil.tz.gettz('Europe/Rome')))
            title = d.get('title', '')
            return (time, room, title)

        return [
            nodes.raw('', tmpl.render(lang=lang, **get_talk_data(n)),
                      format='html')
            for n in sorted(get_talk_names(),
                            key=_sort_date)
                                                                ]


class TalkDirective(Directive):
    required_arguments = 1
    final_argument_whitespace = True
    has_content = True
    option_spec = {
        'lang': directives.unchanged
    }

    def run(self):
        lang = self.options.get('lang', 'C')
        tmpl = jinja_env().get_template('talk.html')
        data = get_talk_data(self.arguments[0])
        if data is None:
            return []
        return [
            nodes.raw('', tmpl.render(lang=lang, **data),
                      format='html')
        ]


def _delta_to_position(delta):
    gridstep = pelican.settings['TALKS_GRID_STEP']
    sec = delta.total_seconds() // gridstep * gridstep
    return int('%2d%02d' % (sec // 3600, (sec % 3600) // 60))


def _delta_inc_position(delta, i):
    gridstep = pelican.settings['TALKS_GRID_STEP']
    delta = delta + datetime.timedelta(minutes=i*gridstep)
    sec = delta.total_seconds() // gridstep * gridstep
    return int('%2d%02d' % (sec // 3600, (sec % 3600) // 60))

def _approx_timestr(timestr):
    gridstep = pelican.settings['TALKS_GRID_STEP']
    t = str(timestr)
    minutes = int(t[-2:])
    hours = t[:-2]
    minutes = minutes // gridstep * gridstep
    return int('%s%02d' % (hours, minutes))


class TalkGridDirective(Directive):
    '''A complete grid'''
    final_argument_whitespace = True
    has_content = True
    option_spec = {
        'lang': directives.unchanged
    }

    def run(self):
        try:
            lang = self.options.get('lang', 'C')
            tmpl = jinja_env().get_template('grid.html')
            output = []
            days = unique_attr(all_talks(), 'day')
            gridstep = pelican.settings['TALKS_GRID_STEP']
            for day in sorted(days):
                talks = {talk['id'] for talk in all_talks()
                         if talk.get('day', None) == day
                         and 'time' in talk
                         and 'room' in talk}
                if not talks:
                    continue
                talks = [get_talk_data(t) for t in talks]
                rooms = set()
                for t in talks:
                    if type(t['room']) is list:
                        for r in t['room']:
                            rooms.add(r)
                    else:
                        rooms.add(t['room'])
                # TODO: ordina in base a qualcosa nel meta.yaml globale
                rooms = list(sorted(rooms))

                # room=* is not a real room.
                # Remove it unless that day only has special rooms
                if '*' in rooms and len(rooms) > 1:
                    del rooms[rooms.index('*')]
                mintimedelta = min({talk['delta'] for talk in talks})
                maxtimedelta = max({talk['delta'] + datetime.timedelta(minutes=talk['duration']) for talk in talks})
                mintime = _delta_to_position(mintimedelta)
                maxtime = _delta_to_position(maxtimedelta)
                times = {}

                t = mintimedelta
                while t <= maxtimedelta:
                    times[_delta_to_position(t)] = [None] * len(rooms)
                    t += datetime.timedelta(minutes=gridstep)
                print(list(times.keys()))
                for talk in sorted(talks, key=lambda x: x['delta']):
                    talktime = _delta_to_position(talk['delta'])
                    position = _approx_timestr(talktime)
                    assert position in times, 'pos=%d,time=%d' % (position, talktime)
                    if talk['room'] == '*':
                        roomnums = range(len(rooms))
                    elif type(talk['room']) is list:
                        roomnums = [rooms.index(r) for r in talk['room']]
                    else:
                        roomnums = [rooms.index(talk['room'])]
                    for roomnum in roomnums:
                        if times[position][roomnum] is not None:
                            logging.error("Talk %s and %s overlap! (room %s)",
                                          times[position][roomnum]['id'],
                                          talk['id'],
                                          rooms[roomnum]
                                          )
                            continue
                        times[position][roomnum] = copy(talk)
                        times[position][roomnum]['skip'] = False
                        for i in range(1, talk['duration'] // gridstep):
                            p = _approx_timestr(_delta_inc_position(talk['delta'], i))
                            times[p][roomnum] = copy(talk)
                            times[p][roomnum]['skip'] = True


                render = tmpl.render(times=times,
                                     rooms=rooms,
                                     mintime=mintime, maxtime=maxtime,
                                     timestep=gridstep,
                                     lang=lang,
                                     )
                output.append(nodes.raw(
                    '', u'<h4>%s</h4>' % format_date(day, format='full',
                                                     locale=lang),
                    format='html'))
                output.append(nodes.raw('', render, format='html'))
        except:
            logging.exception("Error on talk grid")
            import traceback
            traceback.print_exc()
            return []
        return output


def talks_to_ics():
    c = ics.Calendar()
    c.creator = u'pelican'
    for t in all_talks():
        e = talk_to_ics(t)
        if e is not None:
            c.events.add(e)
    return six.text_type(c)


def talk_to_ics(talk):
    def _decode(s):
        if six.PY2:
            return unidecode.unidecode(s)
        else:
            return s

    if 'time' not in talk or 'duration' not in talk:
        return None
    e = ics.Event(
        uid="%s@%d.hackmeeting.org" % (talk['id'],
                                       get_global_data()['startdate'].year),
        name=_decode(talk['title']),
        begin=talk['time'],
        duration=datetime.timedelta(minutes=talk['duration']),
        transparent=True,
    )
    # ics.py has some problems with unicode
    # unidecode replaces letters with their most similar ASCII counterparts
    # (ie: accents get stripped)
    if 'text' in talk:
        e.description = _decode(talk['text'])
    e.url = pelican.settings['SCHEDULEURL'] + '#talk-' + talk['id']
    if 'room' in talk:
        e.location = talk['room']

    return e


class TalksGenerator(generators.Generator):
    def __init__(self, *args, **kwargs):
        self.talks = []
        super(TalksGenerator, self).__init__(*args, **kwargs)

    def generate_context(self):
        self.talks = {n: get_talk_data(n) for n in get_talk_names()}
        self._update_context(('talks',))
        check_overlaps()

    def generate_output(self, writer=None):
        for talkname in sorted(self.talks):
            if 'resources' in self.talks[talkname]:
                outdir = os.path.join(self.output_path,
                                      pelican.settings['TALKS_PATH'], talkname,
                                      pelican.settings['TALKS_ATTACHMENT_PATH']
                                     )
                if os.path.isdir(outdir):
                    shutil.rmtree(outdir)
                shutil.copytree(self.talks[talkname]['resources'], outdir)
        if ICS_ENABLED:
            with io.open(os.path.join(self.output_path,
                                      pelican.settings.get('TALKS_ICS')),
                         'w', encoding='utf8') as buf:
                buf.write(talks_to_ics())
        else:
            logging.warning('module `ics` not found. '
                            'ICS calendar will not be generated')


def add_talks_option_defaults(pelican):
    pelican.settings.setdefault('TALKS_PATH', 'talks')
    pelican.settings.setdefault('TALKS_ATTACHMENT_PATH', 'res')
    pelican.settings.setdefault('TALKS_ICS', 'schedule.ics')
    pelican.settings.setdefault('TALKS_GRID_STEP', 30)


def get_generators(gen):
    return TalksGenerator


def pelican_init(pelicanobj):
    global pelican
    pelican = pelicanobj


try:
    import yaml
except ImportError:
    print('ERROR: yaml not found. Talks plugins will be disabled')

    def register():
        pass
else:

    def register():
        signals.initialized.connect(pelican_init)
        signals.get_generators.connect(get_generators)
        signals.initialized.connect(add_talks_option_defaults)
        directives.register_directive('talklist', TalkListDirective)
        directives.register_directive('talk', TalkDirective)
        directives.register_directive('talkgrid', TalkGridDirective)