europress parser

[iramuteq] / parse_factiva_xml.py

diff --git a/parse_factiva_xml.py b/parse_factiva_xml.py
index 8d9b8a7..a7ed81d 100644 (file)
--- a/parse_factiva_xml.py
+++ b/parse_factiva_xml.py
@@ -2,28 +2,27 @@
 # -*- coding: utf-8 -*-
 #Author: Pierre Ratinaud
 #Copyright (c) 2008-2010 Pierre Ratinaud
-#Lisense: GNU/GPL
+#License: GNU/GPL
 
 import xml.dom.minidom
-import wx.lib.sized_controls as sc
 import wx.lib.filebrowsebutton as filebrowse
 import os
 import codecs
-import sys
 import re
 import wx
 from parse_factiva_mail import ParseFactivaMail
 from parse_factiva_txt import ParseFactivaPaste
+from parse_europress import ParseEuropress
+from import_txm import TXM2IRA
 
 def ParseDocument(filename) :
     print filename
     with codecs.open(filename, 'r', 'utf-8') as f :
-       content = f.read()
+        content = f.read()
     content = content.replace('<hlt>', ' ').replace('</hlt>', ' ')
     dom = xml.dom.minidom.parseString(content.encode("utf-8"))
     result = []
     articles = dom.getElementsByTagName("article")
-    i = 0
     for article in articles :
         headline = article.getElementsByTagName("headline")
         if headline != [] :
@@ -49,7 +48,7 @@ def ParseDocument(filename) :
         if sourceName != [] :
             val_sourceName = sourceName[0].firstChild.nodeValue.replace('\n', ' ')
         else :
-            val_sourceName = INCONNU
+            val_sourceName = 'INCONNU'
         tailParagraphs = article.getElementsByTagName("tailParagraphs")
         if tailParagraphs != [] :
             para_tailParagraphs = tailParagraphs[0].getElementsByTagName("paragraph")
@@ -64,13 +63,12 @@ def ParseDocument(filename) :
     
 def getcorpus_from_xml(xmldir, corpus_out):
     files = os.listdir(xmldir)
-    files = [os.path.join(xmldir,file) for file in files if os.path.splitext(file)[1] == '.xml']
+    files = [os.path.join(xmldir,f) for f in files if os.path.splitext(f)[1] == '.xml']
     if len(files) == 0 :
         return 'nofile'
-    result = []
     fileout = codecs.open(corpus_out, 'w', 'utf-8')
-    for file in files :
-        rs = ParseDocument(file)
+    for f in files :
+        rs = ParseDocument(f)
         #dates = [row[2].split('-') for row in rs]
         #dates = [[date[0],date[1],date[2].split('T')[0]] for date in dates]
         #txt = '\n'.join(['\n'.join([' '.join([u'****', '*%s' % row[1].replace(' ','_').replace('\'','_'), '*%s' % row[2].replace('-','_')]), row[3], row[4]]) for row in rs])
@@ -86,15 +84,15 @@ class PrefImport(wx.Dialog):
         pre.SetExtraStyle(wx.DIALOG_EX_CONTEXTHELP)
         pre.Create(parent, -1, '', pos, size, style)
         self.PostCreate(pre)
-        if methode == 'xml' :
-            txt = _(u'Select a directory of xml files')
+        if methode in ['xml', 'txm'] :
+            txt = _(u'Select a directory of xml files').decode('utf8')
         else :
-            txt = _(u'Select a directory of txt files')
+            txt = _(u'Select a directory of txt files').decode('utf8')
         self.parent = parent
         self.txt1 = wx.StaticText(self, -1, txt.encode('utf8'))
         self.dbb = filebrowse.DirBrowseButton(self, -1, size=(450, -1), changeCallback = self.fbbCallback)
         self.dbb.SetLabel("")
-        self.txt2 = wx.StaticText(self, -1, _(u'Output file').encode('utf8'))
+        self.txt2 = wx.StaticText(self, -1, _(u'Output file').decode('utf8'))
         self.fbb = filebrowse.FileBrowseButton(self, -1, size=(450, -1), fileMode = 2)
         self.fbb.SetLabel("")
 
@@ -133,7 +131,7 @@ class PrefImport(wx.Dialog):
    
     def fbbCallback(self, evt):
         if self.fbb.GetValue() == "" :
-             self.fbb.SetValue(os.path.join(self.dbb.GetValue(), 'corpus.txt'))
+            self.fbb.SetValue(os.path.join(self.dbb.GetValue(), 'corpus.txt'))
         #self.log.write('FileBrowseButton: %s\n' % evt.GetString())
 
     def checkfile(self, evt) :
@@ -159,19 +157,23 @@ class PrefImport(wx.Dialog):
 
 class ImportFactiva():
     def __init__(self, parent, methode):
-       self.dial =  PrefImport(parent, methode=methode)
-       self.dial.CenterOnParent()
-       val = self.dial.ShowModal()
-       if val == wx.ID_OK :
-           xmldir = self.dial.dbb.GetValue()
-           corp_out = self.dial.fbb.GetValue()
-           if methode == 'xml' :
+        self.dial =  PrefImport(parent, methode=methode)
+        self.dial.CenterOnParent()
+        val = self.dial.ShowModal()
+        if val == wx.ID_OK :
+            xmldir = self.dial.dbb.GetValue()
+            corp_out = self.dial.fbb.GetValue()
+            if methode == 'xml' :
                 res = getcorpus_from_xml(xmldir, corp_out)
-           elif methode == 'mail' :
-               res = ParseFactivaMail(xmldir, corp_out, 'utf8', parent.syscoding)
-           elif methode == 'txt' :
-               res = ParseFactivaPaste(xmldir, corp_out, 'utf8', parent.syscoding)
-           if res == 'nofile' :
+            elif methode == 'mail' :
+                res = ParseFactivaMail(xmldir, corp_out, 'utf8', parent.syscoding)
+            elif methode == 'txt' :
+                res = ParseFactivaPaste(xmldir, corp_out, 'utf8', parent.syscoding)
+            elif methode == 'txm' :
+                res = TXM2IRA(xmldir, corp_out, 'utf8', parent.syscoding)
+            elif methode == 'euro' :
+                res = ParseEuropress(xmldir, corp_out, 'utf8', 'utf8')
+            if res == 'nofile' :
                 dlg = wx.MessageDialog(parent, u"Pas de fichier \'.xml\' dans %s" % xmldir, 'ATTENTION', wx.OK | wx.NO_DEFAULT | wx.ICON_WARNING)
                 dlg.CenterOnParent()
                 dlg.ShowModal()