Replace bleach.linkify with cmark-gfm autolink

Yet another issue with Bleach 3.0's linkification: when used via the
filter as part of sanitization (which is necessary right now due to
*another* issue where it escapes valid HTML tags), it doesn't properly
linkify urls that contain an ampersand.

As a (temporary?) workaround, this stops using Bleach's linkification
entirely and switches to cmark-gfm's "autolink" extension. These aren't
perfectly equivalent, and the switch results in two other issues that I
consider more minor than links including ampersands not working:

- autolink will initially create links for ftp:// urls and email
  addresses. The final sanitization will remove these links due to the
  protocol whitelist, but it will leave behind a bare <a> tag. So the
  text will *appear* linked but not actually link to anything. If I
  decide to stick with autolink, it should be pretty straightforward to
  fix this by stripping all bare <a> tags from the final HTML.
- autolink doesn't create links for bare domains. For example, writing
  "example.com" won't result in a link, it's necessary to write
  "www.example.com" or include a protocol like "http://example.com".

tildes/tests/test_markdown.py

@@ -130,14 +130,6 @@ def test_https_link_linkified():
     assert '<a href="https://example.com">' in processed
 
 
-def test_bare_domain_linkified():
-    """Ensure that a bare domain results in a link."""
-    markdown = "I can just write example.com too."
-    processed = convert_markdown_to_safe_html(markdown)
-
-    assert '<a href="http://example.com">' in processed
-
-
 def test_link_with_path_linkified():
     """Ensure a link with a path results in a link."""
     markdown = "So http://example.com/a/b_c_d/e too?"
@@ -148,18 +140,10 @@ def test_link_with_path_linkified():
 
 def test_link_with_query_string_linkified():
     """Ensure a link with a query string results in a link."""
-    markdown = "Also http://example.com?something=true works?"
+    markdown = "Also http://example.com?something=true&test=yes works?"
     processed = convert_markdown_to_safe_html(markdown)
 
-    assert '<a href="http://example.com?something=true">' in processed
-
-
-def test_email_address_not_linkified():
-    """Ensure that an email address does not get linkified."""
-    markdown = "Please contact somebody@example.com about that."
-    processed = convert_markdown_to_safe_html(markdown)
-
-    assert "<a" not in processed
+    assert '<a href="http://example.com?something=true&amp;test=yes">' in processed
 
 
 def test_other_protocol_urls_not_linkified():
@@ -170,7 +154,7 @@ def test_other_protocol_urls_not_linkified():
         markdown = f"Testing {protocol}://example.com for linking"
         processed = convert_markdown_to_safe_html(markdown)
 
-        assert "<a" not in processed
+        assert "href" not in processed
 
 
 def test_html_attr_whitelist_violation():

tildes/tildes/lib/cmark.py

@@ -13,7 +13,7 @@ CMARK_EXT_DLL = CDLL("/usr/local/lib/libcmark-gfm-extensions.so")
 # CMARK_OPT_UNSAFE (1 << 17)
 CMARK_OPTS = (1 << 2) | (1 << 17)
 
-CMARK_EXTENSIONS = (b"strikethrough", b"table")
+CMARK_EXTENSIONS = (b"autolink", b"strikethrough", b"table")
 
 cmark_parser_new = CMARK_DLL.cmark_parser_new
 cmark_parser_new.restype = c_void_p

tildes/tildes/lib/markdown.py

@@ -5,19 +5,7 @@
 
 from functools import partial
 import re
-from typing import (
-    Any,
-    Callable,
-    Dict,
-    Iterator,
-    List,
-    Match,
-    Optional,
-    Pattern,
-    Tuple,
-    Union,
-)
-from urllib.parse import urlparse
+from typing import Any, Callable, Iterator, List, Match, Optional, Pattern, Tuple
 
 from bs4 import BeautifulSoup
 import bleach
@@ -117,29 +105,6 @@ BAD_ORDERED_LIST_REGEX = re.compile(
     r"\.\s"  # Followed by a period and a space
 )
 
-# Type alias for the "namespaced attr dict" used inside bleach.linkify callbacks. This
-# looks pretty ridiculous, but it's a dict where the keys are namespaced attr names,
-# like `(None, 'href')`, and there's also a `_text` key for getting the innerText of the
-# <a> tag.
-NamespacedAttrDict = Dict[Union[Tuple[Optional[str], str], str], str]
-
-
-def linkify_protocol_whitelist(
-    attrs: NamespacedAttrDict, new: bool = False
-) -> Optional[NamespacedAttrDict]:
-    """bleach.linkify callback: prevent links to non-whitelisted protocols."""
-    # pylint: disable=unused-argument
-    href = attrs.get((None, "href"))
-    if not href:
-        return attrs
-
-    parsed = urlparse(href)
-
-    if parsed.scheme not in PROTOCOL_WHITELIST:
-        return None
-
-    return attrs
-
 
 @histogram_timer("markdown_processing")
 def convert_markdown_to_safe_html(markdown: str) -> str:
@@ -444,17 +409,12 @@ def linkify_and_sanitize_html(html: str) -> str:
     # list of tag names to exclude from linkification
     linkify_skipped_tags = ["code", "pre"]
 
-    bleach_linkifier = partial(
-        bleach.linkifier.LinkifyFilter,
-        callbacks=[linkify_protocol_whitelist],
-        skip_tags=linkify_skipped_tags,
-    )
     tildes_linkifier = partial(LinkifyFilter, skip_tags=linkify_skipped_tags)
 
     cleaner = bleach.Cleaner(
         tags=HTML_TAG_WHITELIST,
         attributes=HTML_ATTRIBUTE_WHITELIST,
         protocols=PROTOCOL_WHITELIST,
-        filters=[bleach_linkifier, tildes_linkifier],
+        filters=[tildes_linkifier],
     )
     return cleaner.clean(html)