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'**** ') :
26 if line.startswith(u'-*') :
31 def testvar(line, variable) :
33 varmod = [val.split('_') for val in line[1:]]
34 vars = [var[0] for var in varmod]
36 return '_'.join([variable, varmod[vars.index(variable)][1]]).replace(u'*','')
40 def testmod(line, mods) :
44 return mod.replace(u'*','')
49 def __init__(self, parent, option) :
50 dial = ExtractDialog(parent, option)
52 res = dial.ShowModal()
54 parametres = dial.make_param()
55 if option == 'splitvar' :
56 SplitFromVar(parametres)
57 elif option == 'mods' :
58 ExtractMods(parametres)
59 elif option == 'them' :
60 SplitFromThem(parametres)
62 dial = wx.MessageDialog(parent, 'Done !', style = wx.OK)
67 def __init__(self, parametres) :
68 self.filein = parametres['filein']
69 self.var = parametres['var']
70 self.encodein = parametres['encodein']
71 self.encodeout = parametres['encodeout']
72 self.basepath = os.path.dirname(self.filein)
78 with codecs.open(self.filein, 'r', self.encodein) as fin :
81 varmod = testvar(line, self.var)
84 if varmod not in filedict :
85 filename = os.path.join(self.basepath, varmod + '.txt')
86 filedict[varmod] = open(filename, 'w')
87 fileout = filedict[varmod]
91 fileout.write(line.encode(self.encodeout))
96 def __init__(self, parametres) :
97 self.filein = parametres['filein']
98 self.them = parametres['them']
99 self.encodein = parametres['encodein']
100 self.encodeout = parametres['encodeout']
101 self.basepath = os.path.dirname(self.filein)
102 self.pathout = os.path.join(self.basepath, '_'.join([them.replace(u'-*','') for them in self.them]))
103 self.fileout = open(self.pathout, 'w')
111 with codecs.open(self.filein, 'r', self.encodein) as fin :
114 self.writetext(self.fileout, lastet, text)
118 l = line.strip().rstrip('\n\r')
125 self.writetext(self.fileout, lastet, text)
127 def writetext(self, fileout, lastet, text):
129 self.fileout.write(lastet.encode(self.encodeout) + text.encode(self.encodeout))
133 def __init__(self, parametres) :
134 self.onefile = parametres.get('onefile', False)
135 self.filein = parametres['filein']
136 self.mods = parametres['mods']
137 self.encodein = parametres['encodein']
138 self.encodeout = parametres['encodeout']
139 self.basepath = os.path.dirname(self.filein)
141 filename = os.path.join(self.basepath, '_'.join([mod.replace(u'*','') for mod in self.mods])+'.txt')
142 self.fileout = open(filename, 'w')
148 with codecs.open(self.filein, 'r', self.encodein) as fin :
151 modinline = testmod(line, self.mods)
154 if not self.onefile :
155 if modinline not in filedict :
156 filename = os.path.join(self.basepath, modinline + '.txt')
157 filedict[modinline] = open(filename, 'w')
158 fileout = filedict[modinline]
160 fileout = self.fileout
164 fileout.write(line.encode(self.encodeout))
165 if not self.onefile :
172 class SubCorpus(Corpus) :
173 def __init__(self, parent, corpus, sgts) :
174 Corpus.__init__(self, parent, corpus.parametres)
176 self.corpus = copycorpus(corpus)
177 self.corpus.make_lems(self.parametres['lem'])
178 textes = list(set([corpus.getucefromid(sgt).uci for sgt in sgts]))
179 self.ucis = [corpus.ucis[i] for i in textes]
180 for texte in self.ucis :
181 texte.uces = [uce for uce in texte.uces if uce.ident in self.sgts]
182 self.make_formes(corpus)
183 self.pathout = corpus.pathout
184 self.parametres['sub'] = self.sgts
186 def make_formes(self, corpus) :
188 for forme in self.corpus.formes :
189 sgtseff = self.corpus.getformeuceseff(forme)
190 sgts = set(self.sgts).intersection(sgtseff.keys())
192 self.formes[forme] = self.corpus.formes[forme]
193 self.formes[forme].freq = sum([sgtseff[sgt] for sgt in sgts])
195 def getlemuces(self, lem) :
196 return list(set(self.sgts).intersection(self.corpus.getlemuces(lem)))
198 def converttabletocorpus(table, fileout, enc='UTF8') :
200 var = var[0:len(var)-1]
202 et = [zip(var, line[0:len(line)-1]) for line in table]
203 et = ['**** ' + ' '.join(['*' + '_'.join(val) for val in line]) for line in et]
204 txt = ['\n'.join([et[i], line[-1]]) for i, line in enumerate(table)]
206 #with open(fileout, 'w') as f :
212 if __name__ == '__main__' :
213 #SplitFromVar(parametres)
214 ExtractMods(parametres, True)