Merge pull request #120 from mailgun/sergey/talon-1.3.3

bump talon version

setup.py

@@ -29,7 +29,7 @@ class InstallCommand(install):
 
 
 setup(name='talon',
-      version='1.3.0',
+      version='1.3.3',
       description=("Mailgun library "
                    "to extract message quotations and signatures."),
       long_description=open("README.rst").read(),

talon/quotations.py

@@ -172,6 +172,9 @@ MAX_HTML_LEN = 2794202
 QUOT_PATTERN = re.compile('^>+ ?')
 NO_QUOT_LINE = re.compile('^[^>].*[\S].*')
 
+# Regular expression to identify if a line is a header.
+RE_HEADER = re.compile(": ")
+
 
 def extract_from(msg_body, content_type='text/plain'):
     try:
@@ -386,9 +389,6 @@ def _extract_from_html(msg_body):
     then checking deleted checkpoints,
     then deleting necessary tags.
     """
-    if len(msg_body) > MAX_HTML_LEN:
-        return msg_body
-
     if msg_body.strip() == b'':
         return msg_body
 
@@ -453,6 +453,54 @@ def _extract_from_html(msg_body):
     return html.tostring(html_tree_copy)
 
 
+def split_emails(msg):
+    """
+    Given a message (which may consist of an email conversation thread with multiple emails), mark the lines to identify
+     split lines, content lines and empty lines.
+
+    Correct the split line markers inside header blocks. Header blocks are identified by the regular expression
+    RE_HEADER.
+
+    Return the corrected markers
+    """
+    delimiter = get_delimiter(msg)
+    msg_body = preprocess(msg, delimiter)
+    # don't process too long messages
+    lines = msg_body.splitlines()[:MAX_LINES_COUNT]
+    markers = mark_message_lines(lines)
+
+    # we don't want splitlines in header blocks
+    markers = _correct_splitlines_in_headers(markers, lines)
+
+    return markers
+
+
+def _correct_splitlines_in_headers(markers, lines):
+    """Corrects markers by removing splitlines deemed to be inside header blocks"""
+    updated_markers = ""
+    i = 0
+    in_header_block = False
+
+    for m in markers:
+        # Only set in_header_block flag true when we hit an 's' and the line is a header.
+        if m == 's':
+            if not in_header_block:
+                if bool(re.search(RE_HEADER, lines[i])):
+                    in_header_block = True
+            else:
+                m = 't'
+
+        # If the line is not a header line, set in_header_block false.
+        if not bool(re.search(RE_HEADER, lines[i])):
+            in_header_block = False
+
+        # Add the marker to the new updated markers string.
+        updated_markers += m
+        i += 1
+
+    return updated_markers
+
+
 def _readable_text_empty(html_tree):
     return not bool(html_tree_to_text(html_tree).strip())
 

talon/utils.py

@@ -178,7 +178,10 @@ def html_fromstring(s):
     """Parse html tree from string. Return None if the string can't be parsed.
     """
     try:
-        return html5parser.fromstring(s, parser=_HTML5LIB_PARSER)
+        if html_too_big(s):
+            return None
+
+        return html5parser.fromstring(s, parser=_html5lib_parser())
     except Exception:
         pass
 
@@ -187,7 +190,10 @@ def html_document_fromstring(s):
     """Parse html tree from string. Return None if the string can't be parsed.
     """
     try:
-        return html5parser.document_fromstring(s, parser=_HTML5LIB_PARSER)
+        if html_too_big(s):
+            return None
+
+        return html5parser.document_fromstring(s, parser=_html5lib_parser())
     except Exception:
         pass
 
@@ -196,6 +202,10 @@ def cssselect(expr, tree):
     return CSSSelector(expr)(tree)
 
 
+def html_too_big(s):
+    return s.count('<') > _MAX_TAGS_COUNT
+
+
 def _contains_charset_spec(s):
     """Return True if the first 4KB contain charset spec
     """
@@ -220,6 +230,21 @@ def _encode_utf8(s):
     return s.encode('utf-8') if isinstance(s, six.text_type) else s
 
 
+def _html5lib_parser():
+    """
+    html5lib is a pure-python library that conforms to the WHATWG HTML spec
+    and is not vulnarable to certain attacks common for XML libraries
+    """
+    return html5lib.HTMLParser(
+        # build lxml tree
+        html5lib.treebuilders.getTreeBuilder("lxml"),
+        # remove namespace value from inside lxml.html.html5paser element tag
+        # otherwise it yields something like "{http://www.w3.org/1999/xhtml}div"
+        # instead of "div", throwing the algo off
+        namespaceHTMLElements=False
+    )
+
+
 _UTF8_DECLARATION = (b'<meta http-equiv="Content-Type" content="text/html;'
                      b'charset=utf-8">')
 
@@ -229,13 +254,6 @@ _HARDBREAKS = ['br', 'hr', 'tr']
 
 _RE_EXCESSIVE_NEWLINES = re.compile("\n{2,10}")
 
-# html5lib is a pure-python library that conforms to the WHATWG HTML spec
-# and is not vulnarable to certain attacks common for XML libraries
-_HTML5LIB_PARSER = html5lib.HTMLParser(
-    # build lxml tree
-    html5lib.treebuilders.getTreeBuilder("lxml"),
-    # remove namespace value from inside lxml.html.html5paser element tag
-    # otherwise it yields something like "{http://www.w3.org/1999/xhtml}div"
-    # instead of "div", throwing the algo off
-    namespaceHTMLElements=False
-)
+# an extensive research shows that exceeding this limit
+# might lead to excessive processing time
+_MAX_TAGS_COUNT = 419

tests/html_quotations_test.py

@@ -385,7 +385,7 @@ def test_gmail_forwarded_msg():
     eq_(RE_WHITESPACE.sub('', msg_body), RE_WHITESPACE.sub('', extracted))
 
 
-@patch.object(quotations, 'MAX_HTML_LEN', 1)
+@patch.object(u, '_MAX_TAGS_COUNT', 4)
 def test_too_large_html():
     msg_body = 'Reply' \
                '<div class="gmail_quote">' \

tests/text_quotations_test.py

@@ -696,3 +696,27 @@ def test_standard_replies():
                 "'%(reply)s' != %(stripped)s for %(fn)s" % \
                 {'reply': reply_text, 'stripped': stripped_text,
                  'fn': filename}
+
+
+def test_split_email():
+    msg = """From: Mr. X
+Date: 24 February 2016
+To: Mr. Y
+Subject: Hi
+Attachments: none
+Goodbye.
+From: Mr. Y
+To: Mr. X
+Date: 24 February 2016
+Subject: Hi
+Attachments: none
+
+Hello.
+
+-- Original Message --
+On 24th February 2016 at 09.32am Conal Wrote:
+Hey!
+"""
+    expected_markers = "stttttsttttetestt"
+    markers = quotations.split_emails(msg)
+    eq_(markers, expected_markers)

tests/utils_test.py

@@ -120,6 +120,12 @@ def test_comment_no_parent():
 def test_html_fromstring_exception():
     eq_(None, u.html_fromstring("<html></html>"))
 
+@patch.object(u, 'html_too_big', Mock())
+@patch.object(u.html5parser, 'fromstring')
+def test_html_fromstring_too_big(fromstring):
+    eq_(None, u.html_fromstring("<html></html>"))
+    assert_false(fromstring.called)
+
 
 @patch.object(u.html5parser, 'document_fromstring')
 def test_html_document_fromstring_exception(document_fromstring):
@@ -127,7 +133,26 @@ def test_html_document_fromstring_exception(document_fromstring):
     eq_(None, u.html_document_fromstring("<html></html>"))
 
 
+@patch.object(u, 'html_too_big', Mock())
+@patch.object(u.html5parser, 'document_fromstring')
+def test_html_document_fromstring_too_big(document_fromstring):
+    eq_(None, u.html_document_fromstring("<html></html>"))
+    assert_false(document_fromstring.called)
+
+
 @patch.object(u, 'html_fromstring', Mock(return_value=None))
 def test_bad_html_to_text():
     bad_html = "one<br>two<br>three"
     eq_(None, u.html_to_text(bad_html))
+
+
+@patch.object(u, '_MAX_TAGS_COUNT', 3)
+def test_html_too_big():
+    eq_(False, u.html_too_big("<div></div>"))
+    eq_(True, u.html_too_big("<div><span>Hi</span></div>"))
+
+
+@patch.object(u, '_MAX_TAGS_COUNT', 3)
+def test_html_to_text():
+    eq_("Hello", u.html_to_text("<div>Hello</div>"))
+    eq_(None, u.html_to_text("<div><span>Hi</span></div>"))