import collections
import csv
import logging
import numbers
from collections import OrderedDict
from decimal import Decimal
from importlib import import_module

import xlsxwriter
from django.db.models import BinaryField, ForeignKey
from django.db.models import Max
from django.utils.encoding import force_text

import otree
from otree.currency import Currency, RealWorldCurrency
from otree.common import get_models_module
from otree.models.group import BaseGroup
from otree.models.participant import Participant
from otree.models.player import BasePlayer
from otree.models.session import Session
from otree.models.subsession import BaseSubsession
from otree.models_concrete import PageCompletion
from otree.session import SessionConfig

logger = logging.getLogger(__name__)


def inspect_field_names(Model):
    # filter out BinaryField, because it's not useful for CSV export or
    # live results. could be very big, and causes problems with utf-8 export

    # I tried .get_fields() instead of .fields, but that method returns
    # fields that cause problems, like saying group has an attribute 'player'
    field_names = []
    for f in Model._meta.fields:
        if not isinstance(f, BinaryField):
            if isinstance(f, ForeignKey):
                field_names.append('{}_id'.format(f.name))
            else:
                field_names.append(f.name)

    return field_names


def get_field_names_for_live_update(Model):
    return _get_table_fields(Model, for_export=False)


def get_field_names_for_csv(Model):
    return _get_table_fields(Model, for_export=True)


def _get_table_fields(Model, for_export=False):

    if Model is Session:
        # only data export
        return [
            'code',
            'label',
            'experimenter_name',
            'mturk_HITId',
            'mturk_HITGroupId',
            'comment',
            'is_demo',
        ]

    if Model is Participant:
        if for_export:
            return [
                'id_in_session',
                'code',
                'label',
                '_is_bot',
                '_index_in_pages',
                '_max_page_index',
                '_current_app_name',
                # this could be confusing because it will be in every row,
                # even rows for different rounds.
                #'_round_number',
                '_current_page_name',
                'time_started',
                'visited',
                'mturk_worker_id',
                'mturk_assignment_id',
                # last so that it will be next to payoff_plus_participation_fee
                'payoff',
            ]
        else:
            return [
                '_id_in_session',
                'code',
                'label',
                '_current_page',
                '_current_app_name',
                '_round_number',
                '_current_page_name',
                'status',
                '_last_page_timestamp',
            ]

    if issubclass(Model, BasePlayer):
        subclass_fields = [
            f
            for f in inspect_field_names(Model)
            if f not in inspect_field_names(BasePlayer)
            and f not in ['id', 'group_id', 'subsession_id']
        ]

        if for_export:
            return ['id_in_group'] + subclass_fields + ['payoff']
        else:
            return ['id_in_group', 'role'] + subclass_fields + ['payoff']

    if issubclass(Model, BaseGroup):
        subclass_fields = [
            f
            for f in inspect_field_names(Model)
            if f not in inspect_field_names(BaseGroup)
            and f not in ['id', 'subsession_id']
        ]

        return ['id_in_subsession'] + subclass_fields

    if issubclass(Model, BaseSubsession):
        subclass_fields = [
            f
            for f in inspect_field_names(Model)
            if f not in inspect_field_names(BaseGroup) and f != 'id'
        ]

        return ['round_number'] + subclass_fields


def sanitize_for_csv(value) -> str:
    if value is None:
        return ''
    if value is True:
        return 1
    if value is False:
        return 0
    if isinstance(value, (Currency, RealWorldCurrency)):
        # FIXME: django 1.11 upgrade: make sure querying currency values
        # doesn't slow it down too much (behavior of .values() changed)
        # Alexander has an idea, contact him if it's slow
        return Decimal(value)
    if isinstance(value, numbers.Number):
        return value
    value = force_text(value)
    return value.replace('\n', ' ').replace('\r', ' ')


def sanitize_for_live_update(value):
    # force_text is necessary e.g. for CountryField, which is otherwise
    # not Json serializable
    value = str(sanitize_for_csv(value))
    MAX_LENGTH = 30
    if len(value) > MAX_LENGTH:
        return value[:MAX_LENGTH] + '...'
    return value


def get_payoff_plus_participation_fee(session, participant_values_dict):
    payoff = Currency(participant_values_dict['payoff'])
    return session._get_payoff_plus_participation_fee(payoff)


def get_rows_for_wide_csv():

    sessions = Session.objects.order_by('id')
    session_cache = {row.id: row for row in sessions}

    session_config_fields = set()
    for session in sessions:
        for field_name in SessionConfig(session.config).editable_fields():
            session_config_fields.add(field_name)
    session_config_fields = list(session_config_fields)

    participants = Participant.objects.order_by('id').values()
    if not participants:
        # 1 empty row
        return [[]]

    session_fields = get_field_names_for_csv(Session)
    participant_fields = get_field_names_for_csv(Participant)
    participant_fields.append('payoff_plus_participation_fee')
    header_row = ['participant.{}'.format(fname) for fname in participant_fields]
    header_row += ['session.{}'.format(fname) for fname in session_fields]
    header_row += ['session.config.{}'.format(fname) for fname in session_config_fields]
    rows = [header_row]
    for participant in participants:
        session = session_cache[participant['session_id']]
        participant[
            'payoff_plus_participation_fee'
        ] = get_payoff_plus_participation_fee(session, participant)
        row = [sanitize_for_csv(participant[fname]) for fname in participant_fields]

        row += [sanitize_for_csv(getattr(session, fname)) for fname in session_fields]
        row += [
            sanitize_for_csv(session.config.get(fname))
            for fname in session_config_fields
        ]
        rows.append(row)

    # heuristic to get the most relevant order of apps
    app_sequences = collections.Counter()
    for session in sessions:
        # we loaded the config earlier
        app_sequence = session.config['app_sequence']
        app_sequences[tuple(app_sequence)] += session.num_participants
    most_common_app_sequence = app_sequences.most_common(1)[0][0]

    # can't use settings.INSTALLED_OTREE_APPS, because maybe the app
    # was removed from SESSION_CONFIGS.
    app_names_with_data = set()
    for session in sessions:
        for app_name in session.config['app_sequence']:
            app_names_with_data.add(app_name)

    apps_not_in_popular_sequence = [
        app for app in app_names_with_data if app not in most_common_app_sequence
    ]

    order_of_apps = list(most_common_app_sequence) + apps_not_in_popular_sequence

    rounds_per_app = OrderedDict()
    for app_name in order_of_apps:
        try:
            models_module = get_models_module(app_name)
        except ModuleNotFoundError:
            # this should only happen with devserver because on production server,
            # you would need to resetdb after renaming an app.
            logger.warning(
                f'Cannot export data for app {app_name}, which existed when the session was run '
                f'but no longer exists.'
            )
            continue
        agg_dict = models_module.Subsession.objects.all().aggregate(Max('round_number'))
        highest_round_number = agg_dict['round_number__max']

        if highest_round_number is not None:
            rounds_per_app[app_name] = highest_round_number
    for app_name in rounds_per_app:
        for round_number in range(1, rounds_per_app[app_name] + 1):
            new_rows = get_rows_for_wide_csv_round(app_name, round_number, sessions)
            for i in range(len(rows)):
                rows[i].extend(new_rows[i])
    return rows


def get_rows_for_wide_csv_round(app_name, round_number, sessions):

    models_module = otree.common.get_models_module(app_name)
    Player = models_module.Player
    Group = models_module.Group
    Subsession = models_module.Subsession

    rows = []

    group_cache = {row['id']: row for row in Group.objects.values()}

    columns_for_models = {
        Model.__name__.lower(): get_field_names_for_csv(Model)
        for Model in [Player, Group, Subsession]
    }

    model_order = ['player', 'group', 'subsession']

    header_row = []
    for model_name in model_order:
        for colname in columns_for_models[model_name]:
            header_row.append(
                '{}.{}.{}.{}'.format(app_name, round_number, model_name, colname)
            )

    rows.append(header_row)
    empty_row = ['' for _ in range(len(header_row))]

    for session in sessions:
        subsession = Subsession.objects.filter(
            session_id=session.id, round_number=round_number
        ).values()
        if not subsession:
            subsession_rows = [empty_row for _ in range(session.num_participants)]
        else:
            subsession = subsession[0]
            subsession_id = subsession['id']
            players = (
                Player.objects.filter(subsession_id=subsession_id)
                .order_by('id')
                .values()
            )

            if len(players) != session.num_participants:
                msg = (
                    "Session {} has {} participants, but round {} of app '{}' "
                    "has {} players. The number of players in the subsession "
                    "should always match the number of players in the session. "
                    "Reset the database and examine your code."
                ).format(
                    session.code,
                    session.num_participants,
                    round_number,
                    app_name,
                    len(players),
                )
                raise AssertionError(msg)

            subsession_rows = []

            for player in players:
                # because player.payoff is a property
                player['payoff'] = player['_payoff']
                row = []
                all_objects = {
                    'player': player,
                    'group': group_cache[player['group_id']],
                    'subsession': subsession,
                }

                for model_name in model_order:
                    for colname in columns_for_models[model_name]:
                        value = all_objects[model_name][colname]
                        row.append(sanitize_for_csv(value))
                subsession_rows.append(row)
        rows.extend(subsession_rows)
    return rows


def get_rows_for_csv(app_name):
    # need to use app_name and not app_label because the app might have been
    # removed from SESSION_CONFIGS
    models_module = otree.common.get_models_module(app_name)
    Player = models_module.Player
    Group = models_module.Group
    Subsession = models_module.Subsession

    columns_for_models = {
        Model.__name__.lower(): get_field_names_for_csv(Model)
        for Model in [Player, Group, Subsession, Participant, Session]
    }

    participant_ids = Player.objects.values_list('participant_id', flat=True)
    session_ids = Subsession.objects.values_list('session_id', flat=True)

    players = Player.objects.order_by('id').values()

    value_dicts = {
        'group': {row['id']: row for row in Group.objects.values()},
        'subsession': {row['id']: row for row in Subsession.objects.values()},
        'participant': {
            row['id']: row
            for row in Participant.objects.filter(id__in=participant_ids).values()
        },
        'session': {
            row['id']: row
            for row in Session.objects.filter(id__in=session_ids).values()
        },
    }

    model_order = ['participant', 'player', 'group', 'subsession', 'session']

    # header row
    rows = [
        [
            '{}.{}'.format(model_name, colname)
            for model_name in model_order
            for colname in columns_for_models[model_name]
        ]
    ]

    for player in players:
        # because player.payoff is a property
        player['payoff'] = player['_payoff']
        row = []
        all_objects = {'player': player}
        for model_name in value_dicts:
            obj_id = player['{}_id'.format(model_name)]
            all_objects[model_name] = value_dicts[model_name][obj_id]

        for model_name in model_order:
            for colname in columns_for_models[model_name]:
                value = all_objects[model_name][colname]
                row.append(sanitize_for_csv(value))
        rows.append(row)

    return rows


def get_rows_for_live_update(subsession: BaseSubsession):

    models_module = import_module(subsession.__module__)
    Player = models_module.Player
    Group = models_module.Group
    Subsession = models_module.Subsession

    columns_for_models = {
        Model.__name__.lower(): get_field_names_for_live_update(Model)
        for Model in [Player, Group, Subsession]
    }

    # we had a strange result on one person's heroku instance
    # where Meta.ordering on the Player was being ingnored
    # when you use a filter. So we add one explicitly.
    players = (
        Player.objects.filter(subsession_id=subsession.pk)
        .select_related('group', 'subsession')
        .order_by('pk')
    )

    model_order = ['player', 'group', 'subsession']

    rows = []
    for player in players:
        row = []
        for model_name in model_order:
            if model_name == 'player':
                model_instance = player
            else:
                model_instance = getattr(player, model_name)

            for colname in columns_for_models[model_name]:

                attr = getattr(model_instance, colname, '')
                if callable(attr):
                    try:
                        attr = attr()
                    except Exception:
                        attr = ""
                row.append(sanitize_for_live_update(attr))
        rows.append(row)

    return columns_for_models, rows


def export_wide(fp, file_extension='csv'):
    rows = get_rows_for_wide_csv()
    if file_extension == 'xlsx':
        _export_xlsx(fp, rows)
    else:
        _export_csv(fp, rows)


def export_app(app_name, fp, file_extension='csv'):
    rows = get_rows_for_csv(app_name)
    if file_extension == 'xlsx':
        _export_xlsx(fp, rows)
    else:
        _export_csv(fp, rows)


def _export_csv(fp, rows):
    writer = csv.writer(fp)
    writer.writerows(rows)


def _export_xlsx(fp, rows):
    '''
    CSV often does not open properly in Excel, e.g. unicode
    '''
    workbook = xlsxwriter.Workbook(fp, {'in_memory': True})
    worksheet = workbook.add_worksheet()

    for row_num, row in enumerate(rows):
        for col_num, cell_value in enumerate(row):
            worksheet.write(row_num, col_num, cell_value)
    workbook.close()


def export_time_spent(fp):
    """Write the data of the timespent on each_page as csv into the file-like
    object
    """

    column_names = [
        'session_id',
        'participant__id_in_session',
        'participant__code',
        'page_index',
        'app_name',
        'page_name',
        'epoch_time',
        'seconds_on_page',
        'subsession_pk',
        'auto_submitted',
    ]

    rows = PageCompletion.objects.order_by(
        'session', 'participant', 'page_index'
    ).values_list(*column_names)
    writer = csv.writer(fp)
    writer.writerows([column_names])
    writer.writerows(rows)