import csv
import re
# from cStringIO import StringIO
from copy import deepcopy
from io import StringIO
from os import unlink, getenv
from os.path import join

import chardet
import geopy
import pandas
import sqlalchemy
from flask import (
    Blueprint,
    Response,
    render_template,
    request,
    flash,
    redirect,
    url_for,
    abort,
    send_from_directory,
)
# from pandas.compat import StringIO as PandasStringIO
from wtforms import validators
from yaml import safe_dump as yaml_dump

from flaskr.content import content, base_url
from flaskr.core import (
    get_emission_models,
    increment_hit_counter,
)
from flaskr.extensions import send_email
from flaskr.forms import EstimateForm
from flaskr.geocoder import CachedGeocoder
from flaskr.models import db, Estimation, StatusEnum, ScenarioEnum

main = Blueprint('main', __name__)


#OUT_ENCODING = 'utf-8'


# -----------------------------------------------------------------------------

pi_email = "didier.barret@gmail.com"  # todo: move to content YAML or .env
# pi_email = "goutte@protonmail.com"

# -----------------------------------------------------------------------------


@main.route('/favicon.ico')
# @cache.cached(timeout=10000)
def favicon():  # we want it served from the root, not from static/
    return send_from_directory(
        join(main.root_path, '..', 'static', 'img'),
        'favicon.ico', mimetype='image/vnd.microsoft.icon'
    )


@main.route('/')
@main.route('/home')
@main.route('/home.html')
# @cache.cached(timeout=1000)
def home():
    models = get_emission_models()
    models_dict = {}
    for model in models:
        models_dict[model.slug] = model.__dict__
    increment_hit_counter()
    return render_template(
        'home.html',
        models=models_dict,
        colors=[model.color for model in models],
        labels=[model.name for model in models],
    )


def gather_addresses(from_list, from_file):
    """
    Gather a list of addresses from the provided list and file.
    If the file is provided the list is ignored.
    """
    addresses = []
    if from_file:
        file_mimetype = from_file.mimetype
        file_contents_raw = from_file.read()
        detected = chardet.detect(file_contents_raw)
        if detected['encoding']:
            file_contents = file_contents_raw.decode(
                encoding=detected['encoding']
            )
        else:
            file_contents = file_contents_raw.decode()

        rows_dicts = None

        if 'text/csv' == file_mimetype:
            delimiter = ','
            if ';' in file_contents:
                delimiter = ';'
            rows_dicts = pandas \
                .read_csv(StringIO(file_contents), delimiter=delimiter) \
                .rename(str.lower, axis='columns') \
                .to_dict(orient="row")

        # Here are just *some* of the mimetypes that Microsoft's
        # garbage spreadsheet files may have.
        # application/vnd.ms-excel (official)
        # application/msexcel
        # application/x-msexcel
        # application/x-ms-excel
        # application/x-excel
        # application/x-dos_ms_excel
        # application/xls
        # application/x-xls
        # application/vnd.openxmlformats-officedocument.spreadsheetml.sheet
        # ... Let's check extension instead.

        elif from_file.filename.endswith('xls') \
                or from_file.filename.endswith('xlsx'):

            rows_dicts = pandas \
                .read_excel(StringIO(file_contents)) \
                .rename(str.lower, axis='columns') \
                .to_dict(orient="row")

        # Python 3.7 only
        # elif from_file.filename.endswith('ods'):
        #     rows_dicts = read_ods(PandasStringIO(file_contents), 1) \
        #         .rename(str.lower, axis='columns') \
        #         .to_dict(orient="row")

        if rows_dicts is not None:
            for row_dict in rows_dicts:
                if 'address' in row_dict:
                    addresses.append(row_dict['address'])
                    continue
                address = None
                if 'city' in row_dict:
                    address = row_dict['city']
                if 'country' in row_dict:
                    if address is None:
                        address = row_dict['country']
                    else:
                        address += "," + row_dict['country']
                if address is not None:
                    addresses.append(address)
                else:
                    raise validators.ValidationError(
                        "We could not find Address data in the spreadsheet."
                    )
        else:
            raise validators.ValidationError(
                "We could not find any data in the spreadsheet."
            )

    else:
        addresses = from_list.replace("\r", '').split("\n")

    clean_addresses = []
    # Ignore inevitable copy/paste bloopers
    to_ignore = re.compile(r"City\s*,\s*Country", re.I & re.U)
    for address in addresses:
        if not address:
            continue
        # if type(address).__name__ == 'str':
        #     address = str(address).encode('utf-8')
        address = str(address)
        if to_ignore.match(address) is not None:
            continue
        clean_addresses.append(address)
    addresses = clean_addresses

    # Remove empty lines (if any) and white characters
    addresses = [a.strip() for a in addresses if a]

    return "\n".join(addresses)


@main.route("/estimate", methods=["GET", "POST"])
@main.route("/estimate.html", methods=["GET", "POST"])
def estimate():  # register new estimation request, more accurately
    maximum_travels_to_compute = 1000000
    models = get_emission_models()
    form = EstimateForm()

    def show_form():
        return render_template("estimation-request.html", form=form, models=models)

    if form.validate_on_submit():

        estimation = Estimation()
        # estimation.email = form.email.data
        estimation.run_name = form.run_name.data
        estimation.first_name = form.first_name.data
        estimation.last_name = form.last_name.data
        estimation.institution = form.institution.data
        estimation.status = StatusEnum.pending

        try:
            estimation.origin_addresses = gather_addresses(
                form.origin_addresses.data,
                form.origin_addresses_file.data
            )
        except validators.ValidationError as e:
            form.origin_addresses_file.errors.append(str(e))
            return show_form()
        except UnicodeDecodeError as e:
            form.origin_addresses_file.errors.append(
                "We only accept UTF-8 and UTF-16 encoded files, \n" +
                "or files we can detect encoding from."
            )
            return show_form()

        try:
            estimation.destination_addresses = gather_addresses(
                form.destination_addresses.data,
                form.destination_addresses_file.data
            )
        except validators.ValidationError as e:
            form.destination_addresses_file.errors.append(str(e))
            return show_form()
        except UnicodeDecodeError as e:
            form.origin_addresses_file.errors.append(
                "We only accept UTF-8 and UTF-16 encoded files, \n" +
                "or files we can detect encoding from."
            )
            return show_form()

        estimation.use_train_below_km = form.use_train_below_km.data

        models_slugs = []
        models_count = 0
        for model in models:
            if getattr(form, 'use_model_%s' % model.slug).data:
                models_slugs.append(model.slug)
                models_count += 1
        estimation.models_slugs = u"\n".join(models_slugs)

        travels_to_compute = \
            models_count * \
            (estimation.origin_addresses.count("\n") + 1) * \
            (estimation.destination_addresses.count("\n") + 1)
        if travels_to_compute > maximum_travels_to_compute:
            message = """
            Too many travels to compute. (%d > %d)
            We're working on increasing this limitation.
            Please contact us directly if you wish to boost this issue
            or get a dedicated estimation.
            """ % (travels_to_compute, maximum_travels_to_compute)
            form.origin_addresses.errors.append(message)
            form.destination_addresses.errors.append(message)
            # form.origin_addresses_file.errors.append(message)
            # form.destination_addresses_file.errors.append(message)
            return show_form()

        db.session.add(estimation)
        db.session.commit()

        send_email(
            to_recipient=pi_email,
            subject="[TCFM] New Estimation Request: %s" % estimation.public_id,
            message=render_template(
                'email/run_requested.html',
                base_url=base_url,
                estimation=estimation,
            )
        )

        flash("Estimation request submitted successfully.", "success")
        return redirect(url_for(
            endpoint=".consult_estimation",
            public_id=estimation.public_id,
            extension='html'
        ))

    return show_form()


@main.route("/invalidate")
@main.route("/invalidate.html")
def invalidate():
    stuck_estimations = Estimation.query \
        .filter_by(status=StatusEnum.working) \
        .all()

    for estimation in stuck_estimations:
        estimation.status = StatusEnum.failure
        estimation.errors = "Invalidated. Try again."
        db.session.commit()

    return "Estimations invalidated: %d" % len(stuck_estimations)


@main.route("/invalidate-geocache")
@main.route("/invalidate-geocache.html")
def invalidate_geocache():
    geocache = 'geocache.db'

    unlink(geocache)

    return "Geocache invalidated."


@main.route("/compute")
def compute():  # process the queue of estimation requests

    # maximum_addresses_to_compute = 30000

    def _respond(_msg):
        return "<pre>%s</pre>" % _msg

    def _handle_failure(_estimation, _failure_message):
        _estimation.status = StatusEnum.failure
        _estimation.errors = _failure_message
        db.session.commit()
        send_email(
            to_recipient=pi_email,
            subject="[TCFM] Run failed: %s" % _estimation.public_id,
            message=render_template(
                'email/run_failed.html',
                base_url=base_url,
                estimation=_estimation,
            )
        )

    def _handle_warning(_estimation, _warning_message):
        if not _estimation.warnings:
            _estimation.warnings = _warning_message
        else:
            _estimation.warnings += _warning_message
            # _estimation.warnings = u"%s\n%s" % \
            #                        (_estimation.warnings, _warning_message)
        db.session.commit()

    estimation = None
    try:
        response = ""

        count_working = Estimation.query \
            .filter_by(status=StatusEnum.working) \
            .count()

        if 0 < count_working:
            return _respond("Already working on estimation.")

        try:
            estimation = Estimation.query \
                .filter_by(status=StatusEnum.pending) \
                .order_by(Estimation.id.asc()) \
                .first()
        except sqlalchemy.orm.exc.NoResultFound:
            return _respond("No estimation in the queue.")
        except Exception as e:
            return _respond("Database error: %s" % (e,))

        if not estimation:
            return _respond("No estimation in the queue.")

        estimation.status = StatusEnum.working
        db.session.commit()

        response += u"Processing estimation `%s`...\n" % (
            estimation.public_id
        )

        # GEOCODE ADDRESSES ###################################################

        failed_addresses = []
        geocoder = CachedGeocoder()

        # GEOCODE ORIGINS #####################################################

        origins_addresses = estimation.origin_addresses.strip().split("\n")
        origins_addresses_count = len(origins_addresses)
        origins = []

        # if origins_addresses_count > maximum_addresses_to_compute:
        #     errmsg = u"Too many origins. (%d > %d) \n" \
        #              u"Please contact us " \
        #              u"for support of more origins." % \
        #              (origins_addresses_count, maximum_addresses_to_compute)
        #     _handle_failure(estimation, errmsg)
        #     return _respond(errmsg)

        for i in range(origins_addresses_count):

            origin_address = origins_addresses[i].strip()

            if not origin_address:
                continue

            if origin_address in failed_addresses:
                continue

            try:
                origin = geocoder.geocode(origin_address)
            except geopy.exc.GeopyError as e:
                warning = u"Ignoring origin `%s` " \
                          u"since we failed to geocode it.\n%s\n" % (
                            origin_address, e,
                )
                response += warning
                _handle_warning(estimation, warning)
                failed_addresses.append(origin_address)
                continue

            if origin is None:
                warning = u"Ignoring origin `%s` " \
                          u"since we failed to geocode it.\n" % (
                            origin_address,
                )
                response += warning
                _handle_warning(estimation, warning)
                failed_addresses.append(origin_address)
                continue

            origins.append(origin)

            response += u"Origin `%s` geocoded to `%s` (%f, %f).\n" % (
                origin_address, origin.address,
                origin.latitude, origin.longitude,
            )

        # GEOCODE DESTINATIONS ################################################

        destinations_addresses = estimation.destination_addresses.strip().split("\n")
        destinations_addresses_count = len(destinations_addresses)
        destinations = []

        # if destinations_addresses_count > maximum_addresses_to_compute:
        #     errmsg = u"Too many destinations. (%d > %d) \n" \
        #              u"Please contact us " \
        #              u"for support of that many destinations." \
        #              % (
        #                  destinations_addresses_count,
        #                  maximum_addresses_to_compute,
        #              )
        #     _handle_failure(estimation, errmsg)
        #     return _respond(errmsg)

        for i in range(destinations_addresses_count):

            destination_address = destinations_addresses[i].strip()

            if not destination_address:
                continue

            if destination_address in failed_addresses:
                continue

            try:
                destination = geocoder.geocode(destination_address)
            except geopy.exc.GeopyError as e:
                warning = u"Ignoring destination `%s` " \
                          u"since we failed to geocode it.\n%s\n" % (
                            destination_address, e,
                )
                response += warning
                _handle_warning(estimation, warning)
                failed_addresses.append(destination_address)
                continue

            if destination is None:
                warning = u"Ignoring destination `%s` " \
                          u"since we failed to geocode it.\n" % (
                            destination_address,
                )
                response += warning
                _handle_warning(estimation, warning)
                failed_addresses.append(destination_address)
                continue

            # print(repr(destination.raw))

            destinations.append(destination)

            response += u"Destination `%s` geocoded to `%s` (%f, %f).\n" % (
                destination_address, destination.address,
                destination.latitude, destination.longitude,
            )

        geocoder.close()

        # GTFO IF NO ORIGINS OR NO DESTINATIONS ###############################

        if 0 == len(origins):
            response += u"Failed to geocode ALL the origin(s).\n"
            _handle_failure(estimation, response)
            return _respond(response)
        if 0 == len(destinations):
            response += u"Failed to geocode ALL the destination(s).\n"
            _handle_failure(estimation, response)
            return _respond(response)

        # GRAB AND CONFIGURE THE EMISSION MODELS ##############################

        emission_models = estimation.get_models()
        # print(emission_models)

        extra_config = {
            'use_train_below_distance': estimation.use_train_below_km,
            # 'use_train_below_distance': 300,
        }

        # PREPARE RESULT DICTIONARY THAT WILL BE STORED #######################

        results = {}

        # UTILITY PRIVATE FUNCTION(S) #########################################

        # _locations
        def _get_location_key(_location):
            return "%s, %s" % (
                _get_city_key(_location),
                _get_country_key(_location),
            )

        def _get_city_key(_location):
            return _location.address.split(',')[0]

            # _city_key = _location.address
            # # if 'address100' in _location.raw['address']:
            # #     _city_key = _location.raw['address']['address100']
            # if 'city' in _location.raw['address']:
            #     _city_key = _location.raw['address']['city']
            # elif 'state' in _location.raw['address']:
            #     _city_key = _location.raw['address']['state']
            # return _city_key

        def _get_country_key(_location):
            return _location.address.split(',')[-1]

        def compute_one_to_many(
                _origin,
                _destinations,
                _extra_config=None
        ):
            _results = {}
            footprints = {}

            destinations_by_key = {}

            cities_sum_foot = {}
            cities_sum_dist = {}
            cities_dict_first_model = None
            for model in emission_models:
                cities_dict = {}
                for _destination in _destinations:
                    footprint = model.compute_travel_footprint(
                        origin_latitude=_origin.latitude,
                        origin_longitude=_origin.longitude,
                        destination_latitude=_destination.latitude,
                        destination_longitude=_destination.longitude,
                        extra_config=_extra_config,
                    )

                    _key = _get_location_key(_destination)
                    destinations_by_key[_key] = _destination

                    if _key not in cities_dict:
                        cities_dict[_key] = {
                            'location': _get_location_key(_destination),
                            'city': _get_city_key(_destination),
                            'country': _get_country_key(_destination),
                            'address': _destination.address,
                            'latitude': _destination.latitude,
                            'longitude': _destination.longitude,
                            'footprint': 0.0,
                            'distance': 0.0,
                            'train_trips': 0,
                            'plane_trips': 0,
                        }
                    cities_dict[_key]['footprint'] += footprint['co2eq_kg']
                    cities_dict[_key]['distance'] += footprint['distance_km']
                    cities_dict[_key]['train_trips'] += footprint['train_trips']
                    cities_dict[_key]['plane_trips'] += footprint['plane_trips']
                    if _key not in cities_sum_foot:
                        cities_sum_foot[_key] = 0.0
                    cities_sum_foot[_key] += footprint['co2eq_kg']
                    if _key not in cities_sum_dist:
                        cities_sum_dist[_key] = 0.0
                    cities_sum_dist[_key] += footprint['distance_km']

                cities = sorted(cities_dict.values(), key=lambda c: c['footprint'])

                footprints[model.slug] = {
                    'cities': cities,
                }

                if cities_dict_first_model is None:
                    cities_dict_first_model = deepcopy(cities_dict)

            _results['footprints'] = footprints

            total_foot = 0.0
            total_dist = 0.0
            total_train_trips = 0
            total_plane_trips = 0

            cities_mean_dict = {}
            for city in cities_sum_foot.keys():
                city_mean_foot = 1.0 * cities_sum_foot[city] / len(emission_models)
                city_mean_dist = 1.0 * cities_sum_dist[city] / len(emission_models)
                city_train_trips = cities_dict_first_model[city]['train_trips']
                city_plane_trips = cities_dict_first_model[city]['plane_trips']
                cities_mean_dict[city] = {
                    'location': _get_location_key(destinations_by_key[city]),
                    'city': _get_city_key(destinations_by_key[city]),
                    'country': _get_country_key(destinations_by_key[city]),
                    'address': destinations_by_key[city].address,
                    'latitude': destinations_by_key[city].latitude,
                    'longitude': destinations_by_key[city].longitude,
                    'footprint': city_mean_foot,
                    'distance': city_mean_dist,
                    'train_trips': city_train_trips,
                    'plane_trips': city_plane_trips,
                }
                total_foot += city_mean_foot
                total_dist += city_mean_dist
                total_train_trips += city_train_trips
                total_plane_trips += city_plane_trips

            cities_mean = [cities_mean_dict[k] for k in cities_mean_dict.keys()]
            cities_mean = sorted(cities_mean, key=lambda c: c['footprint'])

            _results['mean_footprint'] = {  # DEPRECATED?
                'cities': cities_mean
            }
            _results['cities'] = cities_mean

            _results['total'] = total_foot  # DEPRECATED
            _results['footprint'] = total_foot

            _results['distance'] = total_dist
            _results['train_trips'] = total_train_trips
            _results['plane_trips'] = total_plane_trips

            return _results

        # SCENARIO A : One Origin, At Least One Destination ###################
        #
        # We compute the sum of each of the travels' footprint,
        # for each of the Emission Models, and present a mean of all Models.
        #
        if 1 == len(origins):
            estimation.scenario = ScenarioEnum.one_to_many
            results = compute_one_to_many(
                _origin=origins[0],
                _destinations=destinations,
                _extra_config=extra_config,
            )

        # SCENARIO B : At Least One Origin, One Destination ###################
        #
        # Same as A for now.
        #
        elif 1 == len(destinations):
            estimation.scenario = ScenarioEnum.many_to_one
            results = compute_one_to_many(
                _origin=destinations[0],
                _destinations=origins,
                _extra_config=extra_config,
            )

        # SCENARIO C : At Least One Origin, At Least One Destination ##########
        #
        # Run Scenario A for each Destination, and expose optimum Destination.
        # Skip destinations already visited.  (collapse duplicate destinations)
        #
        else:
            estimation.scenario = ScenarioEnum.many_to_many
            unique_location_keys = []
            result_cities = []
            for destination in destinations:
                location_key = _get_location_key(destination)
                city_key = _get_city_key(destination)
                country_key = _get_country_key(destination)

                if location_key in unique_location_keys:
                    continue
                else:
                    unique_location_keys.append(location_key)

                city_results = compute_one_to_many(
                    _origin=destination,
                    _destinations=origins,
                    _extra_config=extra_config,
                )
                city_results['city'] = city_key
                city_results['country'] = country_key
                city_results['location'] = location_key
                city_results['address'] = destination.address
                city_results['latitude'] = destination.latitude
                city_results['longitude'] = destination.longitude
                result_cities.append(city_results)

            result_cities = sorted(result_cities, key=lambda c: int(c['footprint']))
            results = {
                'cities': result_cities,
            }

        # WRITE RESULTS INTO THE DATABASE #####################################

        estimation.status = StatusEnum.success
        # Don't use YAML, it is too slow for big data.
        # estimation.output_yaml = u"%s" % yaml_dump(results)
        estimation.informations = response
        estimation.set_output_dict(results)
        db.session.commit()

        # SEND AN EMAIL #######################################################

        send_email(
            to_recipient=pi_email,
            subject="[TCFM] Run completed: %s" % estimation.public_id,
            message=render_template(
                'email/run_completed.html',
                base_url=base_url,
                estimation=estimation,
            )
        )

        # FINALLY, RESPOND ####################################################

        # YAML is too expensive, let's not
        # response += yaml_dump(results) + "\n"

        return _respond(response)

    except Exception as e:
        errmsg = u"Computation failed : %s" % (e,)
        if 'production' != getenv('FLASK_ENV', 'production'):
            import traceback
            errmsg = u"%s\n\n%s" % (errmsg, traceback.format_exc())
        if estimation:
            _handle_failure(estimation, errmsg)
        return _respond(errmsg)


@main.route("/estimation/<public_id>.<extension>")
def consult_estimation(public_id, extension):
    try:
        estimation = Estimation.query \
            .filter_by(public_id=public_id) \
            .one()
    except sqlalchemy.orm.exc.NoResultFound:
        return abort(404)
    except Exception as e:
        # log? (or not)
        return abort(500)

    # allowed_formats = ['html']
    # if format not in allowed_formats:
    #     abort(404)

    if extension in ['xhtml', 'html', 'htm']:

        if estimation.status in estimation.unavailable_statuses:
            return render_template(
                "estimation-queue-wait.html",
                estimation=estimation
            )
        else:
            try:
                estimation_output = estimation.get_output_dict()
            except Exception as e:
                return abort(404)
            estimation_sum = 0
            if estimation_output:
                for city in estimation_output['cities']:
                    estimation_sum += city['footprint']

            return render_template(
                "estimation.html",
                estimation=estimation,
                estimation_output=estimation_output,
                estimation_sum=estimation_sum,
            )

    elif extension in ['yaml', 'yml']:

        if not estimation.is_available():
            return abort(404)

        return u"%s" % yaml_dump(estimation.get_output_dict())
        # return estimation.output_yaml

    elif 'csv' == extension:

        if not estimation.is_available():
            return abort(404)

        si = StringIO()
        cw = csv.writer(si, quoting=csv.QUOTE_ALL)
        cw.writerow([
            u"location",
            u"city", u"country", u"address",
            u"latitude", u"longitude",
            u"co2_kg",
            u"distance_km",
            u"plane trips_amount",
            u'train trips_amount',
        ])

        results = estimation.get_output_dict()
        for city in results['cities']:
            cw.writerow([
                # city['location'].encode(OUT_ENCODING),
                city['location'],
                city['city'],
                city['country'],
                city['address'],
                city.get('latitude', 0.0),
                city.get('longitude', 0.0),
                round(city['footprint'], 3),
                round(city['distance'], 3),
                city['plane_trips'],
                city['train_trips'],
            ])

        # return si.getvalue().strip('\r\n')
        return Response(
            response=si.getvalue().strip('\r\n'),
            headers={
                'Content-type': 'text/csv',
                'Content-disposition': "attachment; filename=%s.csv"%public_id,
            },
        )

    else:
        return abort(404)


def get_locations(addresses):
    geocoder = CachedGeocoder()

    warnings = []
    addresses_count = len(addresses)
    failed_addresses = []
    locations = []

    for i in range(addresses_count):

        address = addresses[i].strip()
        unicode_address = address.encode('utf-8')

        if not address:
            continue

        if address in failed_addresses:
            continue

        try:
            location = geocoder.geocode(unicode_address)
        except geopy.exc.GeopyError as e:
            warning = u"Ignoring address `%s` " \
                      u"since we failed to geocode it.\n%s\n" % (
                          address, e,
                      )
            warnings.append(warning)
            failed_addresses.append(address)
            continue

        if location is None:
            warning = u"Ignoring address `%s` " \
                      u"since we failed to geocode it.\n" % (
                          address,
                      )
            warnings.append(warning)
            failed_addresses.append(address)
            failed_addresses.append(address)
            continue

        print("Geocoded Location:\n", repr(location.raw))
        locations.append(location)

        # response += u"Location `%s` geocoded to `%s` (%f, %f).\n" % (
        #     location_address, location.address,
        #     location.latitude, location.longitude,
        # )

    return locations, warnings


@main.route("/estimation/<public_id>/trips_to_destination_<destination_index>.csv")
def get_trips_csv(public_id, destination_index=0):
    destination_index = int(destination_index)
    try:
        estimation = Estimation.query \
            .filter_by(public_id=public_id) \
            .one()
    except sqlalchemy.orm.exc.NoResultFound:
        return abort(404)
    except Exception as e:
        return abort(500)

    if not estimation.is_available():
        return abort(404)

    si = StringIO()
    cw = csv.writer(si, quoting=csv.QUOTE_ALL)
    cw.writerow([
        u"origin_lon",
        u"origin_lat",
        u"destination_lon",
        u"destination_lat",
    ])
    results = estimation.get_output_dict()

    if not 'cities' in results:
        return abort(500)

    cities_length = len(results['cities'])

    if 0 == cities_length:
        return abort(500, Response("No cities in results."))

    destination_index = min(destination_index, cities_length - 1)
    destination_index = max(destination_index, 0)

    city = results['cities'][destination_index]
    # >>> yaml_dump(city)
    # address: Paris, Ile - de - France, Metropolitan
    # France, France
    # city: Paris
    # country: ' France'
    # distance: 1752.7481921181325
    # footprint: 824.9628320703453
    # plane_trips: 1
    # train_trips: 0

    geocoder = CachedGeocoder()
    try:
        city_location = geocoder.geocode(city['address'].encode('utf-8'))
    except geopy.exc.GeopyError as e:
        return Response(
            response=si.getvalue().strip('\r\n'),
        )

    other_locations, _warnings = get_locations(estimation.origin_addresses.split("\n"))
    # destination_locations = get_locations(estimation.destination_addresses.split("\n"))
    for other_location in other_locations:
        cw.writerow([
            u"%.8f" % city_location.longitude,
            u"%.8f" % city_location.latitude,
            u"%.8f" % other_location.longitude,
            u"%.8f" % other_location.latitude,
        ])

    filename = "trips_to_destination_%d.csv" % destination_index
    return Response(
        response=si.getvalue().strip('\r\n'),
        headers={
            'Content-type': 'text/csv',
            'Content-disposition': 'attachment; filename=%s' % filename,
        },
    )


@main.route("/scaling_laws.csv")
def get_scaling_laws_csv():
    distances = content.laws_plot.distances
    models = get_emission_models()

    si = StringIO()
    cw = csv.writer(si, quoting=csv.QUOTE_ALL)

    header = ['distance'] + [model.slug for model in models]
    cw.writerow(header)

    for distance in distances:
        row = [distance]
        for model in models:
            row.append(model.compute_airplane_distance_footprint(distance))
        cw.writerow(row)

    return Response(
        response=si.getvalue().strip('\r\n'),
        headers={
            'Content-type': 'text/csv',
            'Content-disposition': 'attachment; filename=scaling_laws.csv',
        },
    )


@main.route('/geocode')
@main.route('/geocode.html')
def query_geocode():
    requested = request.args.getlist('address')
    if not requested:
        requested = request.args.getlist('address[]')
    if not requested:
        requested = request.args.getlist('a')
    if not requested:
        requested = request.args.getlist('a[]')
    # requested = _collect_request_args_list(('address', 'a'))
    if not requested:
        return Response(
            response="""
<p>
Usage example: <a href="/geocode.html?address=Toulouse,France&amp;address=Paris,France">/geocode?address=Toulouse,France</a>
</p>

<p>
Please do not request this endpoint more than every two seconds.
</p>
"""
        )

    response = u""

    geocoder = CachedGeocoder()
    for address in requested:
        location = geocoder.geocode(address)

        response += """
<pre>
Requested: `%s'
Geocoded: `%s'
Longitude: `%f`
Latitude: `%f`
Altitude: `%f` (unreliable)
</pre>
""" % (address, location, location.longitude, location.latitude, location.altitude)

    return Response(response=response)


@main.route("/test")
# @basic_auth.required
def dev_test():
    # email_content = render_template(
    #     'email/run_completed.html',
    #     # run=run,
    # )
    # send_email(
    #     'goutte@protonmail.com',
    #     subject=u"[TCFC] New run request",
    #     message=email_content
    # )

    return "ok"