From 8d42f42ef207e3aff64a37bb41aa1d2e63e5ec21 Mon Sep 17 00:00:00 2001
From: Harshit Bansal <harshitbansal2015@gmail.com>
Date: Wed, 11 Oct 2017 20:41:19 +0000
Subject: [PATCH] notifications: Correctly convert relative narrow links to
 absolute URLs.

---
 zerver/lib/notifications.py        | 23 ++++++++++++++++++++---
 zerver/tests/test_notifications.py |  8 ++++++++
 2 files changed, 28 insertions(+), 3 deletions(-)

diff --git a/zerver/lib/notifications.py b/zerver/lib/notifications.py
index a80201f94a..5afe0e786d 100644
--- a/zerver/lib/notifications.py
+++ b/zerver/lib/notifications.py
@@ -71,9 +71,26 @@ def topic_narrow_url(realm, stream, topic):
 def relative_to_full_url(base_url, content):
     # type: (Text, Text) -> Text
     # Convert relative URLs to absolute URLs.
-    elem = lxml.html.fromstring(content)  # type: ignore # https://github.com/python/typeshed/issues/525
-    elem.make_links_absolute(base_url)
-    content = lxml.html.tostring(elem).decode("utf-8")  # type: ignore # https://github.com/python/typeshed/issues/525
+    fragment = lxml.html.fromstring(content)  # type: ignore # https://github.com/python/typeshed/issues/525
+
+    # We handle narrow URLs separately because of two reasons:
+    # 1: 'lxml' seems to be having an issue in dealing with URLs that begin
+    # `#` due to which it doesn't add a `/` before joining the base_url to
+    # the relative URL.
+    # 2: We also need to update the title attribute in the narrow links which
+    # is not possible with `make_links_absolute()`.
+    for link_info in fragment.iterlinks():
+        elem, attrib, link, pos = link_info
+        match = re.match("/?#narrow/", link)
+        if match is not None:
+            link = re.sub(r"^/?#narrow/", base_url + "/#narrow/", link)
+            elem.set(attrib, link)
+            # Only manually linked narrow URLs have title attribute set.
+            if elem.get('title') is not None:
+                elem.set('title', link)
+
+    fragment.make_links_absolute(base_url)
+    content = lxml.html.tostring(fragment).decode("utf-8")  # type: ignore # https://github.com/python/typeshed/issues/525
 
     # Inline images can't be displayed in the emails as the request
     # from the mail server can't be authenticated because it has no
diff --git a/zerver/tests/test_notifications.py b/zerver/tests/test_notifications.py
index 12f1a4b2da..69acf4bada 100644
--- a/zerver/tests/test_notifications.py
+++ b/zerver/tests/test_notifications.py
@@ -420,6 +420,14 @@ class TestMissedMessages(ZulipTestCase):
         expected_output = '<p>Set src="/avatar/username@example.com?s=30"</p>'
         self.assertEqual(actual_output, expected_output)
 
+        # A narrow URL which begins with a '#'.
+        test_data = '<p><a href="#narrow/stream/test/subject/test.20topic/near/142"' +  \
+                    'title="#narrow/stream/test/subject/test.20topic/near/142">Conversation</a></p>'
+        actual_output = relative_to_full_url("http://example.com", test_data)
+        expected_output = '<p><a href="http://example.com/#narrow/stream/test/subject/test.20topic/near/142" ' + \
+                          'title="http://example.com/#narrow/stream/test/subject/test.20topic/near/142">Conversation</a></p>'
+        self.assertEqual(actual_output, expected_output)
+
     def test_fix_emoji(self):
         # type: () -> None
         # An emoji.