2 # -*- coding: utf-8 -*-
3 #Author: Pierre Ratinaud
4 #Copyright (c) 2008-2010 Pierre Ratinaud
8 import wx.lib.sized_controls as sc
9 import wx.lib.filebrowsebutton as filebrowse
17 def ParseDocument(filename) :
19 with codecs.open(filename, 'r', 'utf-8') as f :
21 content = content.replace('<hlt>', ' ').replace('</hlt>', ' ')
22 dom = xml.dom.minidom.parseString(content.encode("utf-8"))
24 articles = dom.getElementsByTagName("article")
26 for article in articles :
27 headline = article.getElementsByTagName("headline")
29 para_headline = headline[0].getElementsByTagName("paragraph")
30 val_headline = [val.firstChild.nodeValue.replace('\n', ' ') for val in para_headline]
33 leadParagraph = article.getElementsByTagName("leadParagraph")
34 if leadParagraph != [] :
35 para_leadParagraph = leadParagraph[0].getElementsByTagName("paragraph")
36 val_leadParagraph = [val.firstChild.nodeValue.replace('\n', ' ') for val in para_leadParagraph]
38 val_leadParagraph = []
39 publicationDate = article.getElementsByTagName("publicationDate")
40 if publicationDate != [] :
41 para_publicationDate = publicationDate[0].getElementsByTagName("date")
42 if para_publicationDate == [] :
43 para_publicationDate = publicationDate[0].getElementsByTagName("dateTime")
44 val_publicationDate = [val.firstChild.nodeValue.replace('\n', ' ') for val in para_publicationDate]
46 val_publicationDate = []
47 sourceName = article.getElementsByTagName("sourceName")
49 val_sourceName = sourceName[0].firstChild.nodeValue.replace('\n', ' ')
51 val_sourceName = INCONNU
52 tailParagraphs = article.getElementsByTagName("tailParagraphs")
53 if tailParagraphs != [] :
54 para_tailParagraphs = tailParagraphs[0].getElementsByTagName("paragraph")
55 val_tailParagraphs = [val.firstChild.nodeValue.replace('\n', ' ') for val in para_tailParagraphs]
57 val_tailParagraphs = []
58 inter = [' '.join(val_headline), val_sourceName,' '.join(val_publicationDate), ' '.join(val_leadParagraph), ' '.join(val_tailParagraphs)]
59 inter = [re.sub(ur'[ "\n\r]+', ' ', val).replace('"',' ').replace('\n', ' ').replace('\r', ' ') for val in inter]
60 #inter = ['"' + val +'"' for val in inter]
64 def getcorpus_from_xml(xmldir, corpus_out):
65 files = os.listdir(xmldir)
66 files = [os.path.join(xmldir,file) for file in files if os.path.splitext(file)[1] == '.xml']
70 fileout = codecs.open(corpus_out, 'w', 'utf-8')
72 rs = ParseDocument(file)
73 #dates = [row[2].split('-') for row in rs]
74 #dates = [[date[0],date[1],date[2].split('T')[0]] for date in dates]
75 #txt = '\n'.join(['\n'.join([' '.join([u'****', '*%s' % row[1].replace(' ','_').replace('\'','_'), '*%s' % row[2].replace('-','_')]), row[3], row[4]]) for row in rs])
76 #avec la date decompose
77 txt = '\n'.join(['\n'.join([' '.join([u'****', '*s_%s' % row[1].replace(' ','').replace('\'',''), '*annee_%s' % row[2].split('-')[0], '*mois_%s' % row[2].split('-')[1], '*jour_%s' % row[2].split('-')[2].split('T')[0]]), row[3], row[4]]) for row in rs])
78 fileout.write(txt+'\n\n')
82 class PrefImport(wx.Dialog):
83 def __init__(self, parent, size=wx.DefaultSize, pos=wx.DefaultPosition, style=wx.DEFAULT_DIALOG_STYLE):
85 pre.SetExtraStyle(wx.DIALOG_EX_CONTEXTHELP)
86 pre.Create(parent, -1, '', pos, size, style)
90 self.txt1 = wx.StaticText(self, -1, u"Répertoire des fichiers xml")
91 self.dbb = filebrowse.DirBrowseButton(self, -1, size=(450, -1), changeCallback = self.fbbCallback)
93 self.txt2 = wx.StaticText(self, -1, u"Fichier en sortie")
94 self.fbb = filebrowse.FileBrowseButton(self, -1, size=(450, -1), fileMode = 2)
97 self.btnsizer = wx.StdDialogButtonSizer()
98 btn_ok = wx.Button(self, wx.ID_OK)
99 btn = wx.Button(self, wx.ID_CANCEL)
100 self.btnsizer.AddButton(btn_ok)
101 self.btnsizer.AddButton(btn)
102 self.btnsizer.Realize()
105 self.Bind(wx.EVT_BUTTON, self.checkfile, btn_ok)
107 #self.SetButtonSizer(self.CreateStdDialogButtonSizer(wx.OK | wx.CANCEL))
108 self.Bind(wx.EVT_BUTTON, self.checkfile)
112 self.SetMinSize(self.GetSize())
114 def __do_layout(self):
115 sizer = wx.BoxSizer(wx.VERTICAL)
116 grid_sizer_1 = wx.BoxSizer(wx.HORIZONTAL)
117 grid_sizer_2 = wx.BoxSizer(wx.HORIZONTAL)
118 grid_sizer_1.Add(self.txt1, 0, wx.ALIGN_LEFT | wx.ALIGN_CENTER_VERTICAL, 0)
119 grid_sizer_1.Add(self.dbb, 2, wx.ALIGN_LEFT | wx.ALIGN_CENTER_VERTICAL, 0)
120 grid_sizer_2.Add(self.txt2, 0, wx.ALIGN_LEFT | wx.ALIGN_CENTER_VERTICAL, 0)
121 grid_sizer_2.Add(self.fbb, 2, wx.ALIGN_LEFT | wx.ALIGN_CENTER_VERTICAL, 0)
122 sizer.Add(grid_sizer_1, 0, wx.EXPAND, 0)
123 sizer.Add(grid_sizer_2, 0, wx.EXPAND, 0)
124 sizer.Add(self.btnsizer, 0, wx.EXPAND, 0)
130 def fbbCallback(self, evt):
131 if self.fbb.GetValue() == "" :
132 self.fbb.SetValue(os.path.join(self.dbb.GetValue(), 'corpus.txt'))
133 #self.log.write('FileBrowseButton: %s\n' % evt.GetString())
135 def checkfile(self, evt) :
136 if evt.GetId() == wx.ID_OK :
137 if self.dbb.GetValue() != "" :
138 if os.path.exists(self.fbb.GetValue()):
139 dlg = wx.MessageDialog(self,
140 u"%s\nCe fichier existe, continuer quand même ?" % self.fbb.GetValue(), 'ATTENTION', wx.NO | wx.YES | wx.ICON_WARNING)
142 if dlg.ShowModal() not in [wx.ID_NO, wx.ID_CANCEL]:
143 self.EndModal(wx.ID_OK)
145 self.EndModal(wx.ID_OK)
147 dlg = wx.MessageDialog(self, u"Vous devez choisir le répertoire contenant le ou les fichier(s) xml", 'ATTENTION', wx.OK | wx.ICON_WARNING)
152 self.EndModal(wx.ID_CANCEL)
156 class ImportFactiva():
157 def __init__(self,parent):
158 self.dial = PrefImport(parent)
159 self.dial.CenterOnParent()
160 val = self.dial.ShowModal()
162 xmldir = self.dial.dbb.GetValue()
163 corp_out = self.dial.fbb.GetValue()
164 res = getcorpus_from_xml(xmldir, corp_out)
166 dlg = wx.MessageDialog(parent, u"Pas de fichier \'.xml\' dans %s" % xmldir, 'ATTENTION', wx.OK | wx.NO_DEFAULT | wx.ICON_WARNING)
170 parent.filename = corp_out