# -*- coding: utf-8 -*-
# vim: autoindent shiftwidth=4 expandtab textwidth=120 tabstop=4 softtabstop=4
###############################################################################
# OpenLP - Open Source Lyrics Projection #
# --------------------------------------------------------------------------- #
# Copyright (c) 2008-2017 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:`opspro` module provides the functionality for importing
a OPS Pro database into the OpenLP database.
"""
# WARNING: See https://docs.python.org/2/library/sqlite3.html for value substitution
# in SQL statements
import logging
import re
import pyodbc
import struct
from openlp.core.common import translate
from openlp.plugins.songs.lib.importers.songimport import SongImport
log = logging.getLogger(__name__)
[docs]class OPSProImport(SongImport):
"""
The :class:`OPSProImport` class provides the ability to import the
WorshipCenter Pro Access Database
"""
def __init__(self, manager, **kwargs):
"""
Initialise the WorshipCenter Pro importer.
"""
super(OPSProImport, self).__init__(manager, **kwargs)
[docs] def do_import(self):
"""
Receive a single file to import.
"""
password = self.extract_mdb_password()
try:
conn = pyodbc.connect('DRIVER={{Microsoft Access Driver (*.mdb)}};DBQ={source};'
'PWD={password}'.format(source=self.import_source, password=password))
except (pyodbc.DatabaseError, pyodbc.IntegrityError, pyodbc.InternalError, pyodbc.OperationalError) as e:
log.warning('Unable to connect the OPS Pro database {source}. {error}'.format(source=self.import_source,
error=str(e)))
# Unfortunately no specific exception type
self.log_error(self.import_source, translate('SongsPlugin.OPSProImport',
'Unable to connect the OPS Pro database.'))
return
cursor = conn.cursor()
cursor.execute('SELECT Song.ID, SongNumber, SongBookName, Title, CopyrightText, Version, Origin FROM Song '
'LEFT JOIN SongBook ON Song.SongBookID = SongBook.ID ORDER BY Title')
songs = cursor.fetchall()
self.import_wizard.progress_bar.setMaximum(len(songs))
for song in songs:
if self.stop_import_flag:
break
# Type means: 0=Original, 1=Projection, 2=Own
cursor.execute('SELECT Lyrics, Type, IsDualLanguage FROM Lyrics WHERE SongID = ? AND Type < 2 '
'ORDER BY Type DESC', float(song.ID))
lyrics = cursor.fetchone()
cursor.execute('SELECT CategoryName FROM Category INNER JOIN SongCategory '
'ON Category.ID = SongCategory.CategoryID WHERE SongCategory.SongID = ? '
'ORDER BY CategoryName', float(song.ID))
topics = cursor.fetchall()
try:
self.process_song(song, lyrics, topics)
except Exception as e:
self.log_error(self.import_source,
translate('SongsPlugin.OPSProImport',
'"{title}" could not be imported. {error}').format(title=song.Title, error=e))
[docs] def process_song(self, song, lyrics, topics):
"""
Create the song, i.e. title, verse etc.
The OPS Pro format is a fairly simple text format using tags and anchors/labels. Linebreaks are \r\n.
Double linebreaks are slide dividers. OPS Pro support dual language using tags.
Tags are in [], see the liste below:
[join] are used to separate verses that should be keept on the same slide.
[split] or [splits] can be used to split a verse over several slides, while still being the same verse
Dual language tags:
[trans off] or [vertaal uit] turns dual language mode off for the following text
[trans on] or [vertaal aan] turns dual language mode on for the following text
[taal a] means the following lines are language a
[taal b] means the following lines are language b
"""
self.set_defaults()
self.title = song.Title
if song.CopyrightText:
for line in song.CopyrightText.splitlines():
if line.startswith('©') or line.lower().startswith('copyright'):
self.add_copyright(line)
else:
self.parse_author(line)
if song.Origin:
self.comments = song.Origin
if song.SongBookName:
self.song_book_name = song.SongBookName
if song.SongNumber:
self.song_number = song.SongNumber
for topic in topics:
self.topics.append(topic.CategoryName)
# Try to split lyrics based on various rules
if lyrics:
lyrics_text = lyrics.Lyrics
verses = re.split('\r\n\s*?\r\n', lyrics_text)
verse_tag_defs = {}
verse_tag_texts = {}
for verse_text in verses:
if verse_text.strip() == '':
continue
verse_def = 'v'
# Detect verse number
verse_number = re.match('^(\d+)\r\n', verse_text)
if verse_number:
verse_text = re.sub('^\d+\r\n', '', verse_text)
verse_def = 'v' + verse_number.group(1)
# Detect verse tags
elif re.match('^.+?\:\r\n', verse_text):
tag_match = re.match('^(.+?)\:\r\n(.*)', verse_text, flags=re.DOTALL)
tag = tag_match.group(1).lower()
tag = tag.split(' ')[0]
verse_text = tag_match.group(2)
if 'refrein' in tag or 'chorus' in tag:
verse_def = 'c'
elif 'bridge' in tag:
verse_def = 'b'
verse_tag_defs[tag] = verse_def
verse_tag_texts[tag] = verse_text
# Detect tag reference
elif re.match('^\(.*?\)$', verse_text):
tag_match = re.match('^\((.*?)\)$', verse_text)
tag = tag_match.group(1).lower()
if tag in verse_tag_defs:
verse_text = verse_tag_texts[tag]
verse_def = verse_tag_defs[tag]
# Detect end tag
elif re.match('^\[slot\]\r\n', verse_text, re.IGNORECASE):
verse_def = 'e'
verse_text = re.sub('^\[slot\]\r\n', '', verse_text, flags=re.IGNORECASE)
# Replace the join tag with line breaks
verse_text = verse_text.replace('[join]', '')
# Replace the split tag with line breaks and an optional split
verse_text = re.sub('\[splits?\]', '\r\n[---]', verse_text)
# Handle translations
if lyrics.IsDualLanguage:
verse_text = self.handle_translation(verse_text)
# Remove comments
verse_text = re.sub('\(.*?\)\r\n', '', verse_text, flags=re.IGNORECASE)
self.add_verse(verse_text, verse_def)
self.finish()
[docs] def handle_translation(self, verse_text):
"""
Replace OPS Pro translation tags with a {translation} tag
:param verse_text: the verse text
:return: the verse text with replaced tags
"""
language = None
translation = True
translation_verse_text = ''
start_tag = '{translation}'
end_tag = '{/translation}'
verse_text_lines = verse_text.splitlines()
idx = 0
while idx < len(verse_text_lines):
# Detect if translation is turned on or off
if verse_text_lines[idx] in ['[trans off]', '[vertaal uit]']:
translation = False
idx += 1
elif verse_text_lines[idx] in ['[trans on]', '[vertaal aan]']:
translation = True
idx += 1
elif verse_text_lines[idx] == '[taal a]':
language = 'a'
idx += 1
elif verse_text_lines[idx] == '[taal b]':
language = 'b'
idx += 1
if not idx < len(verse_text_lines):
break
# Handle the text based on whether translation is off or on
if language:
if language == 'b':
translation_verse_text += start_tag
while idx < len(verse_text_lines) and not verse_text_lines[idx].startswith('['):
translation_verse_text += verse_text_lines[idx] + '\r\n'
idx += 1
if language == 'b':
translation_verse_text += end_tag
language = None
elif translation:
translation_verse_text += verse_text_lines[idx] + '\r\n'
idx += 1
if idx < len(verse_text_lines) and not verse_text_lines[idx].startswith('['):
translation_verse_text += start_tag + verse_text_lines[idx] + end_tag + '\r\n'
idx += 1
else:
translation_verse_text += verse_text_lines[idx] + '\r\n'
idx += 1
while idx < len(verse_text_lines) and not verse_text_lines[idx].startswith('['):
translation_verse_text += verse_text_lines[idx] + '\r\n'
idx += 1
return translation_verse_text