2 # -*- coding: utf-8 -*-
3 #Author: Pierre Ratinaud
4 #Copyright (c) 2008-2013, Pierre Ratinaud
9 from dialog import ExtractDialog
10 from corpus import Corpus, copycorpus
14 parametres = {'filein' : 'corpus/lru2.txt',
17 'mods' : [u'*annee_2010', u'*annee_2011']}
20 if line.startswith(u'**** ') :
25 def testvar(line, variable) :
27 varmod = [val.split('_') for val in line[1:]]
28 vars = [var[0] for var in varmod]
30 return '_'.join([variable, varmod[vars.index(variable)][1]]).replace(u'*','')
34 def testmod(line, mods) :
38 return mod.replace(u'*','')
43 def __init__(self, parent, option) :
44 dial = ExtractDialog(parent, option)
46 res = dial.ShowModal()
48 parametres = dial.make_param()
49 if option == 'splitvar' :
50 SplitFromVar(parametres)
52 ExtractMods(parametres)
55 def __init__(self, parametres) :
56 self.filein = parametres['filein']
57 self.var = parametres['var']
58 self.encodein = parametres['encodein']
59 self.encodeout = parametres['encodeout']
60 self.basepath = os.path.dirname(self.filein)
66 with codecs.open(self.filein, 'r', self.encodein) as fin :
69 varmod = testvar(line, self.var)
72 if varmod not in filedict :
73 filename = os.path.join(self.basepath, varmod + '.txt')
74 filedict[varmod] = open(filename, 'w')
75 fileout = filedict[varmod]
79 fileout.write(line.encode(self.encodeout))
84 def __init__(self, parametres) :
85 self.onefile = parametres.get('onefile', False)
86 self.filein = parametres['filein']
87 self.mods = parametres['mods']
88 self.encodein = parametres['encodein']
89 self.encodeout = parametres['encodeout']
90 self.basepath = os.path.dirname(self.filein)
92 filename = os.path.join(self.basepath, '_'.join([mod.replace(u'*','') for mod in self.mods])+'.txt')
93 self.fileout = open(filename, 'w')
99 with codecs.open(self.filein, 'r', self.encodein) as fin :
102 modinline = testmod(line, self.mods)
105 if not self.onefile :
106 if modinline not in filedict :
107 filename = os.path.join(self.basepath, modinline + '.txt')
108 filedict[modinline] = open(filename, 'w')
109 fileout = filedict[modinline]
111 fileout = self.fileout
115 fileout.write(line.encode(self.encodeout))
116 if not self.onefile :
123 class SubCorpus(Corpus) :
124 def __init__(self, parent, corpus, sgts) :
125 Corpus.__init__(self, parent, corpus.parametres)
127 self.corpus = copycorpus(corpus)
128 self.corpus.make_lems(self.parametres['lem'])
129 textes = list(set([corpus.getucefromid(sgt).uci for sgt in sgts]))
130 self.ucis = [corpus.ucis[i] for i in textes]
131 for texte in self.ucis :
132 texte.uces = [uce for uce in texte.uces if uce.ident in self.sgts]
133 self.make_formes(corpus)
134 self.pathout = corpus.pathout
135 self.parametres['sub'] = self.sgts
137 def make_formes(self, corpus) :
139 for forme in self.corpus.formes :
140 sgtseff = self.corpus.getformeuceseff(forme)
141 sgts = set(self.sgts).intersection(sgtseff.keys())
143 self.formes[forme] = self.corpus.formes[forme]
144 self.formes[forme].freq = sum([sgtseff[sgt] for sgt in sgts])
146 def getlemuces(self, lem) :
147 return list(set(self.sgts).intersection(self.corpus.getlemuces(lem)))
154 if __name__ == '__main__' :
155 #SplitFromVar(parametres)
156 ExtractMods(parametres, True)