--- /dev/null Thu Jan 01 00:00:00 1970 +0000
+++ b/src/cm/utils/html.py Mon Nov 23 15:14:29 2009 +0100
@@ -0,0 +1,30 @@
+"""
+Package to manipulage html chunks
+"""
+
+from BeautifulSoup import BeautifulSoup, Comment
+
+def surrond_text_node(html_chunk, start_html, end_html):
+ """
+ Surround text nodes in html_chunk
+ """
+ soup = BeautifulSoup(html_chunk)
+ text_nodes = get_text_nodes(soup)
+ for text_node in text_nodes:
+ if text_node.string.strip():
+ text_node.replaceWith(start_html + text_node.string + end_html)
+ return unicode(soup)
+
+
+# utilities
+def get_text_nodes(soup):
+ return soup(text=lambda text:not isinstance(text, Comment))
+
+
+import re
+
+def cleanup_textarea(input):
+ """
+ Cleanup \r\n to standard \n
+ """
+ return re.sub('(\r\n)|(\n)|(\r)','\n',input)
\ No newline at end of file