2384: KinshipReport use hard to translate sentences (by Josip)

svn: r14094
This commit is contained in:
Jérôme Rapinat 2010-01-18 15:46:32 +00:00
parent d9ad7798bc
commit 659c51a011
3 changed files with 210 additions and 1 deletions

View File

@ -13,7 +13,8 @@ pkgdata_PYTHON = \
rel_es.py\ rel_es.py\
rel_fi.py\ rel_fi.py\
rel_fr.py\ rel_fr.py\
rel_hu.py\ rel_hu.py\
rel_hr.py\
rel_it.py\ rel_it.py\
rel_nl.py\ rel_nl.py\
rel_no.py\ rel_no.py\

195
src/plugins/rel/rel_hr.py Normal file
View File

@ -0,0 +1,195 @@
# -*- coding: utf-8 -*-
#
# Gramps - a GTK+/GNOME based genealogy program
#
# 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; either version 2 of the License, or
# (at your option) any later version.
#
# 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
#
#
# Copyright (C) 2010 Josip (josip at pisoj dot com)
"""
Croatian-specific classes for calculating relationships and kinship names.
"""
#-------------------------------------------------------------------------
#
# GRAMPS modules
#
#-------------------------------------------------------------------------
#import gen.lib
import Relationship
_PARENTS = ['', 'otac i majka', 'djedovi i bake', 'pradjedovi i prabake',
'%s pradjedovi i %s prabake']
_CHILDS = ['', 'djeca', 'unučad', 'praunučad', '%s praunučad']
_CHILDS_SP = ['', 'snahe i zetovi', 'prasnahe i prazetovi',
'supružnici praunučadi', 'supružnici %s praunučadi']
#-------------------------------------------------------------------------
#
# Class HrDeclination
#
#-------------------------------------------------------------------------
class HrDeclination(object):
"""
Croatian declination system
"""
numdcl = { 1: 'prv', 2: 'drug', 3: 'treć', 4: 'četvrt', 5: 'pet',
6: 'šest', 7: 'sedm', 8: 'osm', 9: 'devet', 10: 'deset',
11: 'jedanaest', 12: 'dvanaest', 13: 'trinaest', 14: 'četrnaest',
15: 'petnaest', 16: 'šestnaest', 17:'sedamnaest', 18: 'osamnaest',
19: 'devetnaest', 20: 'dvadeset', 30: 'trideset', 40: 'četrdeset',
50: 'pedeset', 60: 'šesdeset', 70: 'sedamdeset', 80: 'osamdeset',
90: 'devedeset',
(1, 1, 1): 'i', (1, 1, 5): 'i', (1, 2, 1): 'i', (1, 2, 5): 'i',
(1, 1, 2): 'og', (1, 1, 4): 'og', (2, 1, 2): 'og',
(1, 1, 3): 'om', (1, 1, 6): 'om', (0, 1, 7): 'om',
(2, 1, 3): 'om', (2, 1, 6): 'om', (2, 2, 7): 'om',
(1, 1, 7): 'im', (2, 1, 7): 'im', (1, 2, 3): 'im', (1, 2, 6): 'im',
(1, 2, 7): 'im', (0, 2, 3): 'im', (0, 2,6): 'im', (0, 2, 7): 'im',
(0, 1, 1): 'a', (0, 1, 5): 'a', (2, 2, 1): 'a', (2, 2, 5): 'a',
(0, 1, 2): 'e', (0, 1, 4): 'e', (1, 2, 4): 'e',
(0, 2, 1): 'e', (0, 2, 4): 'e', (0, 2, 5): 'e',
(2 , 1, 1): 'o', (2, 1, 4): 'o', (2, 1, 5): 'o',
(0, 1, 3): 'oj', (0, 1, 6): 'oj', (2, 2, 3): 'oj', (2, 2, 6): 'oj',
(0, 2, 2): 'ih', (1, 2, 2): 'ih', (2, 2, 2): 'ih',
(2, 2, 4): 'u' }
def get_ordnum(self, num, gender, number, case):
"""
Declination of ordinal numbers
gender: 0 = feminine, 1 = masculine, 2 = neuter
case: 1 = nominative, 2 = genitive, 3 = dative, 4 = accusative
5 = vocative, 6 = locative, 7 = instrumental
number: 1 = single, 2 = plural
:rtype: str
"""
if 0 < num < 100:
lres = []
#lres.append(self.numdcl[(num / 100) * 100])
if (num < 21) or (num % 10) == 0:
lres.append(self.numdcl[num])
else:
lres.append(self.numdcl[(num % 10) * 10])
lres.append(self.numdcl[(num % 10)])
sufix = self.numdcl[(gender, number, case)]
if num % 10 == 3:
if gender != 0 and sufix != 'oj':
sufix = sufix.replace('o','e')
lres[-1] += sufix
res = ' '.join(lres)
else:
res = str(num) + '. '
return res
HRD = HrDeclination()
HRDN = HRD.get_ordnum
def _get_childs(level, inlaw):
""" children, grandchildren et.
"""
if level < len(_CHILDS)-1:
if not inlaw:
return _CHILDS[level]
else: return _CHILDS_SP[level]
else:
if not inlaw:
return '%s' % _CHILDS[4] % (HRDN(level-2, 0, 1, 1))
elif inlaw:
return '%s' % _CHILDS_SP[4] % (HRDN(level-2, 0, 2, 1))
def _get_parents(level):
""" parents, grandparents
"""
if level < 4:
return _PARENTS[level]
else:
return u'%s' % _PARENTS[4] % (HRDN(level-2, 1, 1, 1),
HRDN(level-2, 0, 2, 1))
def _get_uncles(gen, inlaw):
""" in general: uncles, ants---
"""
if gen == 2:
if not inlaw:
return 'stričevi, ujaci i tetke'
else:
return 'strine, ujne i tetci'
elif gen == 3:
if not inlaw:
return 'prastričevi, praujaci i pratetke'
else:
return 'prastrine, praujne i pratetci'
else:
if not inlaw:
return '%s prastričevi, %s praujaci i %s pratetke' % (HRDN(gen-2,
1, 2, 1), HRDN(gen-2, 1, 2, 1), HRDN(gen-2, 0, 2, 1))
else:
return 'prastrine, praujne i pratetci'
#-------------------------------------------------------------------------
#
# CroatianRelationshipCalculator
#
#-------------------------------------------------------------------------
class RelationshipCalculator(Relationship.RelationshipCalculator):
"""
RelationshipCalculator Class
"""
def __init__(self):
Relationship.RelationshipCalculator.__init__(self)
def get_plural_relationship_string(self, gena, genb, reltocommon_a='',
reltocommon_b='', only_birth=True, in_law_a=False, in_law_b=False):
"""returns kinship terms for groups of persons
"""
if not in_law_b:
rel_str = u'potomci %s generacije predaka u %s koljenu: '\
% (HRDN(gena, 0, 1, 2), HRDN(genb, 2, 1, 6))
elif in_law_b:
rel_str = u'supružnici potomaka %s generacije predaka u %s '\
'koljenu: ' % (HRDN(gena, 0, 1, 2), HRDN(genb, 2, 1, 6))
if genb == 0:
#These are ancestors
rel_str = u'%s generacija predaka: %s' % (HRDN(gena,
0, 1, 1),_get_parents(gena))
elif gena == 0:
# These are descendants
if not in_law_b:
rel_str = u'%s generacija potomaka: %s' % (HRDN(genb,
0, 1, 1), _get_childs(genb, in_law_b))
else: rel_str = u'supružnici %s generacije potomaka: %s' \
% (HRDN(genb, 0, 2, 1), _get_childs(genb, in_law_b))
elif gena == 1 == genb:
# These are siblings
if not in_law_b:
rel_str += u'braća i sestre'
else: rel_str = u'supružnici braće i sestara'
elif genb == 1:
# These are aunts/uncles
rel_str += _get_uncles(gena, in_law_b)
return rel_str
if __name__ == "__main__":
# Test function. Call it as follows from the command line (so as to find
# imported modules):
# export PYTHONPATH=/path/to/gramps/src
# python src/plugins/rel/rel_ru.py
from Relationship import test
RC = RelationshipCalculator()
test(RC, True)

View File

@ -111,6 +111,19 @@ plg.lang_list = ["fr", "FR", "fr_FR", "fr_CA", "francais",
"fr_FR.UTF-8", "fr_FR.utf-8", "fr_FR.utf8", "fr_FR.UTF-8", "fr_FR.utf-8", "fr_FR.utf8",
"fr_CA.UTF-8"] "fr_CA.UTF-8"]
# hr
plg = newplugin()
plg.id = 'relcalc_hr'
plg.name = _("Croatian Relationship Calculator")
plg.description = _("Calculates relationships between people")
plg.version = '1.0'
plg.status = STABLE
plg.fname = 'rel_hr.py'
plg.ptype = RELCALC
plg.relcalcclass = 'RelationshipCalculator'
plg.lang_list = ["hrvatski", "Hrvatski", "croatian", "Croatian",
"hr", "HR", "hr_HR", "hr_HR.UTF-8", "hr_HR.utf-8"]
# hu # hu
plg = newplugin() plg = newplugin()
plg.id = 'relcalc_hu' plg.id = 'relcalc_hu'