openlp/openlp/core/ui/media/vendor/mediainfoWrapper.py

130 lines
5.0 KiB
Python

# -*- coding: utf-8 -*-
# vim: autoindent shiftwidth=4 expandtab textwidth=120 tabstop=4 softtabstop=4
###############################################################################
# OpenLP - Open Source Lyrics Projection #
# --------------------------------------------------------------------------- #
# Copyright (c) 2008-2018 OpenLP Developers #
# --------------------------------------------------------------------------- #
# 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; version 2 of the License. #
# #
# 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 #
###############################################################################
"""
The :mod:`~openlp.core.ui.media.mediainfo` module contains code to run mediainfo on a media file and obtain
information related to the rwquested media.
"""
import json
import os
from subprocess import check_output
from bs4 import BeautifulSoup, NavigableString
ENV_DICT = os.environ
class Track(object):
def __getattribute__(self, name):
try:
return object.__getattribute__(self, name)
except Exception:
pass
return None
def __init__(self, xml_dom_fragment):
self.xml_dom_fragment = xml_dom_fragment
self.track_type = xml_dom_fragment.attrs['type']
for el in self.xml_dom_fragment.children:
if not isinstance(el, NavigableString):
node_name = el.name.lower().strip().strip('_')
if node_name == 'id':
node_name = 'track_id'
node_value = el.string
other_node_name = "other_%s" % node_name
if getattr(self, node_name) is None:
setattr(self, node_name, node_value)
else:
if getattr(self, other_node_name) is None:
setattr(self, other_node_name, [node_value, ])
else:
getattr(self, other_node_name).append(node_value)
for o in [d for d in self.__dict__.keys() if d.startswith('other_')]:
try:
primary = o.replace('other_', '')
setattr(self, primary, int(getattr(self, primary)))
except Exception:
for v in getattr(self, o):
try:
current = getattr(self, primary)
setattr(self, primary, int(v))
getattr(self, o).append(current)
break
except Exception:
pass
def __repr__(self):
return "<Track track_id='{0}', track_type='{1}'>".format(self.track_id, self.track_type)
def to_data(self):
data = {}
for k, v in self.__dict__.items():
if k != 'xml_dom_fragment':
data[k] = v
return data
class MediaInfoWrapper(object):
def __init__(self, xml):
self.xml_dom = xml
xml_types = (str,) # no unicode type in python3
if isinstance(xml, xml_types):
self.xml_dom = MediaInfoWrapper.parse_xml_data_into_dom(xml)
@staticmethod
def parse_xml_data_into_dom(xml_data):
return BeautifulSoup(xml_data, "xml")
@staticmethod
def parse(filename, environment=ENV_DICT):
xml = check_output(['mediainfo', '-f', '--Output=XML', '--Inform=OLDXML', filename])
if not xml.startswith(b'<?xml'):
xml = check_output(['mediainfo', '-f', '--Output=XML', filename])
xml_dom = MediaInfoWrapper.parse_xml_data_into_dom(xml)
return MediaInfoWrapper(xml_dom)
def _populate_tracks(self):
if self.xml_dom is None:
return
for xml_track in self.xml_dom.Mediainfo.File.find_all("track"):
self._tracks.append(Track(xml_track))
@property
def tracks(self):
if not hasattr(self, "_tracks"):
self._tracks = []
if len(self._tracks) == 0:
self._populate_tracks()
return self._tracks
def to_data(self):
data = {'tracks': []}
for track in self.tracks:
data['tracks'].append(track.to_data())
return data
def to_json(self):
return json.dumps(self.to_data())