From 75c9f6d4ae440df3c00625f79ea20079628a8d5f Mon Sep 17 00:00:00 2001
From: Adam Williamson <awilliam@redhat.com>
Date: Feb 22 2018 09:20:15 +0000
Subject: Remove some funky bits from inline pattern regexes


The bits removed here are *very* weird. They are inverted sets
which will match anything but a literal pipe or any character in
the \w class. The following bit, `(?<!\w)`, is a negative
lookbehind assertion which means 'match unless the previous
character is in the \w class'. So these two actually apply to the
same character and are almost entirely redundant.

I think the weird set was *meant* to be something like (^|w),
and the approximate idea here was to match 'start of string
or any non-word character followed by a #'. If so, then in fact
just removing the wacky inverted set is all we need to do, as
negative lookbehind assertions are allowed to match at the start
of the string. (There is actually a whole hidden complication
here where markdown, behind the scenes, adds some more bits to
the pattern we feed it to form a complete regex, but it happens
that everything works OK with that). The tests should suffice to
demonstrate that these regexes still behave as we expect. These
regexes originally came from @ralph, who says he's OK with this
change.

The changes to `handleMatch()` are all to do with the details of
what markdown actually does with these pattern classes. In short
markdown expects that the pattern's `handleMatch()` method spits
out something that corresponds to all the characters the class'
regex pattern consumes. The weird sets we remove in this commit
actually consume a character immediately before the `#`, and the
space in these `text` values is intended to replace that
character. The modified regexes no longer consume that character,
so `handleMatch()` no longer needs to replace it.

This actually fixes a subtle bug, because the character matched
by the set was *not* in fact always a space. This meant that if
you wrote something like 'Issue.#23' in a comment (for some
reason...), it would be turned into 'Issue #23', with the '#23'
linkified - the '.' was turned into a space. With this change,
that bug no longer happens.

Signed-off-by: Adam Williamson <awilliam@redhat.com>

---

diff --git a/pagure/pfmarkdown.py b/pagure/pfmarkdown.py
index 4c569c7..363e94b 100644
--- a/pagure/pfmarkdown.py
+++ b/pagure/pfmarkdown.py
@@ -51,8 +51,8 @@ COMMIT_LINK_RE = \
     '([a-zA-Z0-9_-]*?/)?'\
     '([a-zA-Z0-9_-]+)'\
     '#(?P<id>[\w]{40})'
-IMPLICIT_ISSUE_RE = r'[^|\w](?<!\w)#([0-9]+)'
-IMPLICIT_PR_RE = r'[^|\w](?<!\w)PR#([0-9]+)'
+IMPLICIT_ISSUE_RE = r'(?<!\w)#([0-9]+)'
+IMPLICIT_PR_RE = r'(?<!\w)PR#([0-9]+)'
 IMPLICIT_COMMIT_RE = r'(?<![<\w#])([a-f0-9]{7,40})'
 STRIKE_THROUGH_RE = r'~~(.*?)~~'
 
@@ -157,7 +157,7 @@ class ImplicitIssuePattern(markdown.inlinepatterns.Pattern):
     def handleMatch(self, m):
         """ When the pattern matches, update the text. """
         idx = markdown.util.AtomicString(m.group(2))
-        text = ' #%s' % idx
+        text = '#%s' % idx
         try:
             idx = int(idx)
         except (ValueError, TypeError):
@@ -185,7 +185,7 @@ class ImplicitPRPattern(markdown.inlinepatterns.Pattern):
     def handleMatch(self, m):
         """ When the pattern matches, update the text. """
         idx = markdown.util.AtomicString(m.group(2))
-        text = ' PR#%s' % idx
+        text = 'PR#%s' % idx
         try:
             idx = int(idx)
         except (ValueError, TypeError):