gramps/gramps2/src/DateParser.py

237 lines
8.1 KiB
Python
Raw Normal View History

#
# Gramps - a GTK+/GNOME based genealogy program
#
# Copyright (C) 2004 Donald N. Allingham
#
# 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
#
"""
U.S. English date parsing class. Serves as the base class for any localized
date parsing class.
"""
__author__ = "Donald N. Allingham"
__version__ = "$Revision$"
import re
import time
import locale
import Date
class DateParser:
"""
Converts a text string into a Date object. If the date cannot be
converted, the text string is assigned.
"""
# determine the code set returned by nl_langinfo
_codeset = locale.nl_langinfo(locale.CODESET)
month_to_int = {
unicode(locale.nl_langinfo(locale.MON_1),_codeset).lower() : 1,
unicode(locale.nl_langinfo(locale.ABMON_1),_codeset).lower() : 1,
unicode(locale.nl_langinfo(locale.MON_2),_codeset).lower() : 2,
unicode(locale.nl_langinfo(locale.ABMON_2),_codeset).lower() : 2,
unicode(locale.nl_langinfo(locale.MON_3),_codeset).lower() : 3,
unicode(locale.nl_langinfo(locale.ABMON_3),_codeset).lower() : 3,
unicode(locale.nl_langinfo(locale.MON_4),_codeset).lower() : 4,
unicode(locale.nl_langinfo(locale.ABMON_4),_codeset).lower() : 4,
unicode(locale.nl_langinfo(locale.MON_5),_codeset).lower() : 5,
unicode(locale.nl_langinfo(locale.ABMON_5),_codeset).lower() : 5,
unicode(locale.nl_langinfo(locale.MON_6),_codeset).lower() : 6,
unicode(locale.nl_langinfo(locale.ABMON_6),_codeset).lower() : 6,
unicode(locale.nl_langinfo(locale.MON_7),_codeset).lower() : 7,
unicode(locale.nl_langinfo(locale.ABMON_7),_codeset).lower() : 7,
unicode(locale.nl_langinfo(locale.MON_8),_codeset).lower() : 8,
unicode(locale.nl_langinfo(locale.ABMON_8),_codeset).lower() : 8,
unicode(locale.nl_langinfo(locale.MON_9),_codeset).lower() : 9,
unicode(locale.nl_langinfo(locale.ABMON_9),_codeset).lower() : 9,
unicode(locale.nl_langinfo(locale.MON_10),_codeset).lower() : 10,
unicode(locale.nl_langinfo(locale.ABMON_10),_codeset).lower(): 10,
unicode(locale.nl_langinfo(locale.MON_11),_codeset).lower() : 11,
unicode(locale.nl_langinfo(locale.ABMON_11),_codeset).lower(): 11,
unicode(locale.nl_langinfo(locale.MON_12),_codeset).lower() : 12,
unicode(locale.nl_langinfo(locale.ABMON_12),_codeset).lower(): 12,
}
modifier_to_int = {
'before' : Date.MOD_BEFORE,
'bef' : Date.MOD_BEFORE,
'bef.' : Date.MOD_BEFORE,
'after' : Date.MOD_AFTER,
'aft' : Date.MOD_AFTER,
'aft.' : Date.MOD_AFTER,
'about' : Date.MOD_ABOUT,
'abt.' : Date.MOD_ABOUT,
'abt' : Date.MOD_ABOUT,
'circa' : Date.MOD_ABOUT,
'c.' : Date.MOD_ABOUT,
'around' : Date.MOD_ABOUT,
}
quality_to_int = {
'estimated' : Date.QUAL_ESTIMATED,
'est.' : Date.QUAL_ESTIMATED,
'est' : Date.QUAL_ESTIMATED,
'calc.' : Date.QUAL_CALCULATED,
'calc' : Date.QUAL_CALCULATED,
'calculated' : Date.QUAL_CALCULATED,
}
_qual_str = '(' + '|'.join(
[ key.replace('.','\.') for key in quality_to_int.keys() ]
) + ')'
_mod_str = '(' + '|'.join(
[ key.replace('.','\.') for key in modifier_to_int.keys() ]
) + ')'
_mon_str = '(' + '|'.join(month_to_int.keys()) + ')'
_qual = re.compile("%s\s+(.*)" % _qual_str,re.IGNORECASE)
_span = re.compile("from\s+(.*)\s+to\s+(.*)",re.IGNORECASE)
_range = re.compile("(bet.|between)\s+(.*)\s+and\s+(.*)",re.IGNORECASE)
_modifier = re.compile('%s\s+(.*)' % _mod_str,re.IGNORECASE)
_text = re.compile('%s\s+(\d+)?\s*,?\s*((\d+)(/\d+)?)?' % _mon_str,re.IGNORECASE)
_text2 = re.compile('(\d+)?\s+?%s\s*((\d+)(/\d+)?)?' % _mon_str,re.IGNORECASE)
_numeric = re.compile("((\d+)[/\.])?((\d+)[/\.])?(\d+)")
_iso = re.compile("(\d+)-(\d+)-(\d+)")
def _get_int(self,val):
"""
Converts the string to an integer if the value is not None. If the
value is None, a zero is returned
"""
if val == None:
return 0
else:
return int(val)
def _parse_subdate(self,text):
"""
Converts only the date portion of a date.
"""
try:
value = time.strptime(text)
return (value[2],value[1],value[0],False)
except ValueError:
pass
match = self._text.match(text)
if match:
groups = match.groups()
if groups[0] == None:
m = 0
else:
m = self.month_to_int[groups[0].lower()]
d = self._get_int(groups[1])
if groups[2] == None:
y = 0
s = None
else:
y = int(groups[3])
s = groups[4] != None
return (d,m,y,s)
match = self._text2.match(text)
if match:
groups = match.groups()
if groups[1] == None:
m = 0
else:
m = self.month_to_int[groups[1].lower()]
d = self._get_int(groups[0])
if groups[2] == None:
y = 0
s = None
else:
y = int(groups[3])
s = groups[4] != None
return (d,m,y,s)
match = self._iso.match(text)
if match:
groups = match.groups()
y = self._get_int(groups[0])
m = self._get_int(groups[1])
d = self._get_int(groups[2])
return (d,m,y,False)
match = self._numeric.match(text)
if match:
groups = match.groups()
m = self._get_int(groups[1])
d = self._get_int(groups[3])
y = self._get_int(groups[4])
return (d,m,y,False)
return Date.EMPTY
def set_date(self,date,text):
"""
Parses the text and sets the date according to the parsing.
"""
date.set_text_value(text)
qual = Date.QUAL_NONE
match = self._qual.match(text)
if match:
grps = match.groups()
qual = self.quality_to_int[grps[0].lower()]
text = grps[1]
match = self._span.match(text)
if match:
grps = match.groups()
start = self._parse_subdate(grps[0])
stop = self._parse_subdate(grps[1])
date.set(qual,Date.MOD_SPAN,Date.CAL_GREGORIAN,start + stop)
return
match = self._range.match(text)
if match:
grps = match.groups()
start = self._parse_subdate(grps[1])
stop = self._parse_subdate(grps[2])
date.set(qual,Date.MOD_RANGE,Date.CAL_GREGORIAN,start + stop)
return
match = self._modifier.match(text)
if match:
grps = match.groups()
start = self._parse_subdate(grps[1])
mod = self.modifier_to_int.get(grps[0].lower(),Date.MOD_NONE)
date.set(qual,mod,Date.CAL_GREGORIAN,start)
return
subdate = self._parse_subdate(text)
if subdate == Date.EMPTY:
date.set_as_text(text)
else:
date.set(qual,Date.MOD_NONE,Date.CAL_GREGORIAN,subdate)
def parse(self,text):
"""
Parses the text, returning a Date object.
"""
new_date = Date.Date()
self.set_date(new_date,text)
return new_date