mimeviewer:tab_width in config is not work for browser

[Shun-ichi Goto <gotoh@…>]

On browsing source file, tab-width 8 is always used although specifying [mimeviewer] tab_width. I made a patch, but it may not correct fix.

api.py

@@ -216 +216 @@
                 candidates.append((qr, renderer))
         candidates.sort(lambda x,y: cmp(y[0], x[0]))
 
+        tab_width = int(self.config.get('mimeviewer', 'tab_width'))
+        
+        if mimetype.lower().startswith('text/'):
+            content = content.expandtabs(tab_width)
+
         for qr, renderer in candidates:
             try:
                 self.log.debug('Trying to render HTML preview using %s'
@@ -226 +231 @@
                 elif isinstance(result, (str, unicode)):
                     return result
                 elif annotations:
-                    return self._annotate(result, annotations)
+                    return self._annotate(result, annotations, tab_width)
                 else:
                     buf = StringIO()
                     buf.write('<div class="code-block"><pre>')
-                    tab_width = int(self.config.get('mimeviewer', 'tab_width'))
                     for line in result:
                         buf.write(line.expandtabs(tab_width) + '\n')
                     buf.write('</pre></div>')
@@ -239 +243 @@
                 self.log.warning('HTML preview using %s failed (%s)'
                                  % (renderer, e), exc_info=True)
 
-    def _annotate(self, lines, annotations):
+    def _annotate(self, lines, annotations, tab_width):
         buf = StringIO()
         buf.write('<table class="code-block listing"><thead><tr>')
         annotators = []
@@ -259 +263 @@
                 div, mod = divmod(len(m), 2)
                 return div * '&nbsp; ' + mod * '&nbsp;'
             return (match.group('tag') or '') + '&nbsp;'
-        tab_width = int(self.config.get('mimeviewer', 'tab_width'))
 
         for num, line in enum(_html_splitlines(lines)):
             cells = []

I'm using r2017. There's relating ticket:1655.

[Christopher Lenz]

The patch looks okay.

[Christopher Lenz]

Actually, no. The patch expands tabs before and after letting the renderer doing its job. Currently, the expansion is only done after the processing.

Is the pre-processing needed because the mimeviewer in question (which one is it?) expands tabs itself? If that's the case, we'll run into similar problems as those raised in #1920 about binary vs textual content. Simply checking for subtypes of text/ is not enough.

[Matthew Good]

cmlenz I'm gonna highjack this one.

[Matthew Good]

Fixed in [2156]