Add HTML output format.

pixelocr/formatting.py

@@ -70,3 +70,21 @@ class TextFormat(OutputFormat):
 
     def format_tag(self, tag, text):
         return text
+
+
+class HTMLFormat(OutputFormat):
+    def assign_glyph_tag(self, glyph_data):
+        style = glyph_data.style
+        if style.bold:
+            return 'b'
+        elif style.italic:
+            return 'i'
+        else:
+            return None
+
+    def format_tag(self, tag, text):
+        from xml.sax.saxutils import escape
+        if tag:
+            return '<{tag}>{text}</{tag}>'.format(tag=tag, text=escape(text))
+        else:
+            return text

pixelocr/gui/ocrengine.py

@@ -67,7 +67,7 @@ class OCREngine(QThread):
 
     def recognize_page(self, page):
         glyph_data_seq = itertools.chain(*(self.recognize_line(line) for line in page.lines))
-        output_format = formatting.TextFormat()
+        output_format = formatting.HTMLFormat()
         return ''.join(output_format.format(glyph_data_seq))
 
     def recognize_line(self, line):