Changeset d2ba33c

Timestamp:

Aug 16, 2017, 12:53:41 PM (7 years ago)

Author:

Jason Gross <jasongross9@gmail.com>

Branches:

master

Children:

5dee79a

Parents:

47225c9

git-author:

Jason Gross <jgross@mit.edu> (01/01/14 20:59:51)

git-committer:

Jason Gross <jasongross9@gmail.com> (08/16/17 12:53:41)

Message:

Use g_utf8_casefold and g_utf8_normalize

We define a convenience function compat_casefold in util.c for reuse in
filters.

Files:

• perl/lib/BarnOwl/Message/Zephyr.pm (modified) (3 diffs)
• perl/modules/Jabber/lib/BarnOwl/Message/Jabber.pm (modified) (2 diffs)
• perlglue.xs (modified) (1 diff)
• t/BarnOwl_Message_Zephyr.t (added)
• util.c (modified) (1 diff)

perl/lib/BarnOwl/Message/Zephyr.pm

@@ -9 +9 @@
 
 use base qw( BarnOwl::Message );
-use Unicode::Normalize qw( NFKC );
 
 sub strip_realm {
@@ -22 +21 @@
     my ($user, $realm) = split(/@/,$principal);
     return $realm;
+}
+
+sub casefold_principal {
+    my $principal = shift;
+    # split the principal right after the final @, without eating any
+    # characters; this way, we always get at least '@' in $user
+    my ($user, $realm) = split(/(?<=@)(?=[^@]+$)/, $principal);
+    return lc($user) . uc($realm);
 }
 
@@ -261 +268 @@
             my $realm = '';
             $realm .= '@' . $m->realm if $m->realm ne BarnOwl::zephyr_getrealm();
-            return (lc(NFKC($m->class)) . $realm);
+            return (BarnOwl::compat_casefold($m->class) . uc($realm));
         }
     } else {
         push @filenames, $m->recipient;
     }
-    return map { lc(NFKC(BarnOwl::zephyr_smartstrip_user(strip_realm($_)))) } @filenames;
+    return map { casefold_principal(BarnOwl::zephyr_smartstrip_user(strip_realm($_))) } @filenames;
 }
 

perl/modules/Jabber/lib/BarnOwl/Message/Jabber.pm

@@ -15 +15 @@
 
 use base qw( BarnOwl::Message );
-use Unicode::Normalize qw( NFKC );
 
 sub jtype { shift->{jtype} };
@@ -174 +173 @@
 
 sub log_filenames {
-    return map { lc(NFKC($_)) } BarnOwl::Message::log_filenames(@_);
+    return map { BarnOwl::compat_casefold($_) } BarnOwl::Message::log_filenames(@_);
 }
 

perlglue.xs

@@ -413 +413 @@
                 RETVAL
 
+const utf8 *
+compat_casefold(in)
+        const char * in
+        PREINIT:
+                char *rv;
+        CODE:
+                rv = owl_util_compat_casefold(in);
+                RETVAL = rv;
+        OUTPUT:
+                RETVAL
+        CLEANUP:
+                g_free(rv);
+
 
 MODULE = BarnOwl                PACKAGE = BarnOwl::Zephyr

util.c

@@ -643 +643 @@
 }
 
+CALLER_OWN char *owl_util_compat_casefold(const char *str)
+{
+  /*
+   * Quoting Anders Kaseorg at https://github.com/barnowl/barnowl/pull/54#issuecomment-31452543:
+   *
+   * The Unicode specification calls this compatibility caseless matching, and
+   * the correct transformation actually has five calls:
+   * NFKC(toCasefold(NFKD(toCasefold(NFD(string))))) Zephyr’s current
+   * implementation incorrectly omits the innermost NFD, but that difference
+   * only matters for characters including U+0345 ◌ͅ COMBINING GREEK
+   * YPOGEGRAMMENI. I think we should just write the correct version and get
+   * Zephyr fixed.
+   *
+   * Neither of these operations should be called toNFKC_Casefold, because that
+   * has slightly different behavior regarding Default_Ignorable_Code_Point. I
+   * propose compat_casefold. And I guess if Jabber wants it too, we should
+   * move it to util.c.
+   */
+  char *tmp0 = g_utf8_normalize(str, -1, G_NORMALIZE_NFD);
+  char *tmp1 = g_utf8_casefold(tmp0, -1);
+  char *tmp2 = g_utf8_normalize(tmp1, -1, G_NORMALIZE_NFKD);
+  char *tmp3 = g_utf8_casefold(tmp2, -1);
+  char *out = g_utf8_normalize(tmp3, -1, G_NORMALIZE_NFKC);
+  g_free(tmp0);
+  g_free(tmp1);
+  g_free(tmp2);
+  g_free(tmp3);
+
+  return out;
+}
+
 /* This is based on _extract() and _isCJ() from perl's Text::WrapI18N */
 int owl_util_can_break_after(gunichar c)