multisplit

[iramuteq] / import_txm.py

diff --git a/import_txm.py b/import_txm.py
index 6ef0c2b..37ffa2e 100644 (file)
--- a/import_txm.py
+++ b/import_txm.py
@@ -1,20 +1,16 @@
-#!/bin/env python
 # -*- coding: utf-8 -*-
 #Author: Pierre Ratinaud
-#Copyright (c) 2013 Pierre Ratinaud
-#Lisense: GNU/GPL
-
+#Copyright (c) 2008-2020 Pierre Ratinaud
+#modification pour python 3 : Laurent Mérat, 6x7 - mai 2020
+#License: GNU/GPL
 
+#------------------------------------
+# import des modules python
+#------------------------------------
 import os
 import xml.sax
 import glob
 
-
-
-#infiledir = '/home/pierre/TXM/corpora/voeux-bin/txm/VOEUX/'
-#fileout = 'VOEUXExportfromTXM.txt'
-
-
 class TXMParser(xml.sax.ContentHandler) :
     def __init__(self, fileout, encodage_out) :
         self.fileout = fileout
@@ -35,7 +31,7 @@ class TXMParser(xml.sax.ContentHandler) :
             pass
 
     def endElement(self, name) :
-        if name == 's' :
+        if name == 's' or name == 'w' :
             self.printsent()
         if name == 'p' :
             self.printsent()
@@ -43,36 +39,34 @@ class TXMParser(xml.sax.ContentHandler) :
           
     def characters(self, content) :
         if self.name == 'txm:form' :
-            if content not in [u'', u' ', u'\n', '\r'] :
+            if content not in ['', ' ', '\n', '\r'] :
                 self.sent.append(content.rstrip('\n\r'))
             #self.fileout.write(content.encode('utf8'))
 
     def text2stars(self, attrs) :
-        stars = ['_'.join(val).replace(' ', '_').replace("'", '_') for val in attrs.items()]
-        stars = [''.join([u'*', val]) for val in stars]
-        stars = u'**** ' + ' '.join(stars)
-        self.fileout.write(stars.encode(self.encodage_out))
+        stars = ['_'.join(val).replace(' ', '_').replace("'", '_').replace('/','').replace('.','').replace(';', '').replace(':', '').replace('·','') for val in list(attrs.items())]
+        stars = [''.join(['*', val]) for val in stars]
+        stars = '**** ' + ' '.join(stars)
+        self.fileout.write(stars)
         self.fileout.write('\n')
 
     def printsent(self) :
         if self.sent != [] :
             sent = ' ' + ' '.join(self.sent)
-            for val in [u' .', u' ,', u' ;', u' :', u' ?', u' !', u' -'] :
+            for val in [' .', ' ,', ' ;', ' :', ' ?', ' !', ' -'] :
                 sent = sent.replace(val, val.strip())
             sent = sent.replace("' ", "'")
-            self.fileout.write(sent.encode(self.encodage_out))
+            self.fileout.write(sent)
         self.sent = []
 
-
-
-class TXM2IRA :
-    def __init__(self, pathin, fileout, encodage_in, encodage_out) :
+def TXM2IRA(pathin, fileout, encodage_in, encodage_out) :
         parser = xml.sax.make_parser()
         files = glob.glob(os.path.join(pathin,'*.xml'))
+        if len(files) == 0 :
+            return 'nofile'
         with open(fileout, 'w') as fout :
             parser.setContentHandler(TXMParser(fout, encodage_out))
             for f in files :
                 parser.parse(open(f, 'r'))
                 fout.write('\n\n')
-
-#TXM2IRA(infiledir, fileout)
+        return None