bazarr/libs/subliminal/score.py

# -*- coding: utf-8 -*-
"""
This module provides the default implementation of the `compute_score` parameter in
:meth:`~subliminal.core.ProviderPool.download_best_subtitles` and :func:`~subliminal.core.download_best_subtitles`.

.. note::

    To avoid unnecessary dependency on `sympy <http://www.sympy.org/>`_ and boost subliminal's import time, the
    resulting scores are hardcoded here and manually updated when the set of equations change.

Available matches:

  * hash
  * title
  * year
  * series
  * season
  * episode
  * release_group
  * source
  * audio_codec
  * resolution
  * hearing_impaired
  * video_codec
  * series_imdb_id
  * imdb_id
  * tvdb_id

"""
from __future__ import division, print_function
from __future__ import absolute_import
import logging

from .video import Episode, Movie

logger = logging.getLogger(__name__)


#: Scores for episodes
episode_scores = {'hash': 359, 'series': 180, 'year': 90, 'season': 30, 'episode': 30, 'release_group': 15,
                  'source': 7, 'audio_codec': 3, 'resolution': 2, 'video_codec': 2, 'hearing_impaired': 1}

#: Scores for movies
movie_scores = {'hash': 119, 'title': 60, 'year': 30, 'release_group': 15,
                'source': 7, 'audio_codec': 3, 'resolution': 2, 'video_codec': 2, 'hearing_impaired': 1}

#: Equivalent release groups
equivalent_release_groups = ({'LOL', 'DIMENSION'}, {'ASAP', 'IMMERSE', 'FLEET'}, {'AVS', 'SVA'})


def get_equivalent_release_groups(release_group):
    """Get all the equivalents of the given release group.

    :param str release_group: the release group to get the equivalents of.
    :return: the equivalent release groups.
    :rtype: set

    """
    for equivalent_release_group in equivalent_release_groups:
        if release_group in equivalent_release_group:
            return equivalent_release_group

    return {release_group}


def get_scores(video):
    """Get the scores dict for the given `video`.

    This will return either :data:`episode_scores` or :data:`movie_scores` based on the type of the `video`.

    :param video: the video to compute the score against.
    :type video: :class:`~subliminal.video.Video`
    :return: the scores dict.
    :rtype: dict

    """
    if isinstance(video, Episode):
        return episode_scores
    elif isinstance(video, Movie):
        return movie_scores

    raise ValueError('video must be an instance of Episode or Movie')


def compute_score(subtitle, video, hearing_impaired=None):
    """Compute the score of the `subtitle` against the `video` with `hearing_impaired` preference.

    :func:`compute_score` uses the :meth:`Subtitle.get_matches <subliminal.subtitle.Subtitle.get_matches>` method and
    applies the scores (either from :data:`episode_scores` or :data:`movie_scores`) after some processing.

    :param subtitle: the subtitle to compute the score of.
    :type subtitle: :class:`~subliminal.subtitle.Subtitle`
    :param video: the video to compute the score against.
    :type video: :class:`~subliminal.video.Video`
    :param bool hearing_impaired: hearing impaired preference.
    :return: score of the subtitle.
    :rtype: int

    """
    logger.info('Computing score of %r for video %r with %r', subtitle, video, dict(hearing_impaired=hearing_impaired))

    # get the scores dict
    scores = get_scores(video)
    logger.debug('Using scores %r', scores)

    # get the matches
    matches = subtitle.get_matches(video)
    logger.debug('Found matches %r', matches)

    # on hash match, discard everything else
    if 'hash' in matches:
        logger.debug('Keeping only hash match')
        matches &= {'hash'}

    # handle equivalent matches
    if isinstance(video, Episode):
        if 'title' in matches:
            logger.debug('Adding title match equivalent')
            matches.add('episode')
        if 'series_imdb_id' in matches:
            logger.debug('Adding series_imdb_id match equivalent')
            matches |= {'series', 'year'}
        if 'imdb_id' in matches:
            logger.debug('Adding imdb_id match equivalents')
            matches |= {'series', 'year', 'season', 'episode'}
        if 'tvdb_id' in matches:
            logger.debug('Adding tvdb_id match equivalents')
            matches |= {'series', 'year', 'season', 'episode'}
        if 'series_tvdb_id' in matches:
            logger.debug('Adding series_tvdb_id match equivalents')
            matches |= {'series', 'year'}
    elif isinstance(video, Movie):
        if 'imdb_id' in matches:
            logger.debug('Adding imdb_id match equivalents')
            matches |= {'title', 'year'}

    # handle hearing impaired
    if hearing_impaired is not None and subtitle.hearing_impaired == hearing_impaired:
        logger.debug('Matched hearing_impaired')
        matches.add('hearing_impaired')

    # compute the score
    score = sum((scores.get(match, 0) for match in matches))
    logger.info('Computed score %r with final matches %r', score, matches)

    # ensure score is within valid bounds
    assert 0 <= score <= scores['hash'] + scores['hearing_impaired']

    return score


def solve_episode_equations():
    from sympy import Eq, solve, symbols

    hash, series, year, season, episode, release_group = symbols('hash series year season episode release_group')
    source, audio_codec, resolution, video_codec = symbols('source audio_codec resolution video_codec')
    hearing_impaired = symbols('hearing_impaired')

    equations = [
        # hash is best
        Eq(hash, series + year + season + episode + release_group + source + audio_codec + resolution + video_codec),

        # series counts for the most part in the total score
        Eq(series, year + season + episode + release_group + source + audio_codec + resolution + video_codec + 1),

        # year is the second most important part
        Eq(year, season + episode + release_group + source + audio_codec + resolution + video_codec + 1),

        # season is important too
        Eq(season, release_group + source + audio_codec + resolution + video_codec + 1),

        # episode is equally important to season
        Eq(episode, season),

        # release group is the next most wanted match
        Eq(release_group, source + audio_codec + resolution + video_codec + 1),

        # source counts as much as audio_codec, resolution and video_codec
        Eq(source, audio_codec + resolution + video_codec),

        # audio_codec is more valuable than video_codec
        Eq(audio_codec, video_codec + 1),

        # resolution counts as much as video_codec
        Eq(resolution, video_codec),

        # video_codec is the least valuable match but counts more than the sum of all scoring increasing matches
        Eq(video_codec, hearing_impaired + 1),

        # hearing impaired is only used for score increasing, so put it to 1
        Eq(hearing_impaired, 1),
    ]

    return solve(equations, [hash, series, year, season, episode, release_group, source, audio_codec, resolution,
                             hearing_impaired, video_codec])


def solve_movie_equations():
    from sympy import Eq, solve, symbols

    hash, title, year, release_group = symbols('hash title year release_group')
    source, audio_codec, resolution, video_codec = symbols('source audio_codec resolution video_codec')
    hearing_impaired = symbols('hearing_impaired')

    equations = [
        # hash is best
        Eq(hash, title + year + release_group + source + audio_codec + resolution + video_codec),

        # title counts for the most part in the total score
        Eq(title, year + release_group + source + audio_codec + resolution + video_codec + 1),

        # year is the second most important part
        Eq(year, release_group + source + audio_codec + resolution + video_codec + 1),

        # release group is the next most wanted match
        Eq(release_group, source + audio_codec + resolution + video_codec + 1),

        # source counts as much as audio_codec, resolution and video_codec
        Eq(source, audio_codec + resolution + video_codec),

        # audio_codec is more valuable than video_codec
        Eq(audio_codec, video_codec + 1),

        # resolution counts as much as video_codec
        Eq(resolution, video_codec),

        # video_codec is the least valuable match but counts more than the sum of all scoring increasing matches
        Eq(video_codec, hearing_impaired + 1),

        # hearing impaired is only used for score increasing, so put it to 1
        Eq(hearing_impaired, 1),
    ]

    return solve(equations, [hash, title, year, release_group, source, audio_codec, resolution, hearing_impaired,
                             video_codec])