fix: count parentheses in links (#5515)

CHANGELOG.md

@@ -22,7 +22,7 @@
 - Minor: Introduce HTTP API for plugins. (#5383, #5492, #5494)
 - Minor: Support more Firefox variants for incognito link opening. (#5503)
 - Minor: Replying to a message will now display the message being replied to. (#4350, #5519)
-- Minor: Links can now have prefixes and suffixes such as parentheses. (#5486)
+- Minor: Links can now have prefixes and suffixes such as parentheses. (#5486, #5515)
 - Bugfix: Fixed tab move animation occasionally failing to start after closing a tab. (#5426)
 - Bugfix: If a network request errors with 200 OK, Qt's error code is now reported instead of the HTTP status. (#5378)
 - Bugfix: Fixed restricted users usernames not being clickable. (#5405)

src/common/LinkParser.cpp

@@ -118,10 +118,12 @@ bool startsWithPort(QStringView string)
 /// As per https://github.github.com/gfm/#autolinks-extension-:
 ///
 /// '<', '*', '_', '~', and '(' are ignored at the beginning
-/// '>', '?', '!', '.', ',', ':', '*', '~', and ')' are ignored at the end
+/// '>', '?', '!', '.', ',', ':', '*', '~', and ')' are ignored at the end.
 ///
-/// A difference to GFM is that the source isn't scanned for parentheses and '_'
+/// A difference to GFM is that '_' isn't a valid suffix.
-/// isn't a valid suffix.
+///
+/// This might remove more than desired (e.g. "(a.com/(foo))" -> "a.com/(foo").
+/// Parentheses are counted after recognizing a valid IP/host.
 void strip(QStringView &source)
 {
     while (!source.isEmpty())
@@ -259,6 +261,43 @@ std::optional<Parsed> parse(const QString &source) noexcept
     if ((nDots == 3 && isValidIpv4(host)) ||
         isValidTld(host.mid(lastDotPos + 1)))
     {
+        // scan for parentheses (only if there were characters excluded)
+        if (link.end() != source.end() && !rest.empty())
+        {
+            size_t nestingLevel = 0;
+            // position after the last closing brace (i.e. the minimum characters to include)
+            const auto *lastClose = link.end();
+
+            // scan source from rest until the end:
+            //                            lastClose
+            //                                v
+            // (example.com/foo/bar/#baz_(qox)),
+            //             ▏╌╌rest (before)╌ ▏
+            //  ▏╌╌╌╌╌╌╌link (before)╌╌╌╌╌╌╌ ▏
+            //             ▏╌╌rest (after)╌╌╌ ▏
+            //  ▏╌╌╌╌╌╌╌link (after)╌╌╌╌╌╌╌╌╌ ▏
+            // ▏╌╌╌╌╌╌╌╌╌╌╌╌╌source╌╌╌╌╌╌╌╌╌╌╌╌ ▏
+            //             ▏╌╌╌╌╌╌╌search╌╌╌╌╌╌ ▏
+            for (const auto *it = rest.begin(); it < source.end(); it++)
+            {
+                if (it->unicode() == u'(')
+                {
+                    nestingLevel++;
+                    continue;
+                }
+
+                if (nestingLevel != 0 && it->unicode() == u')')
+                {
+                    nestingLevel--;
+                    if (nestingLevel == 0)
+                    {
+                        lastClose = it + 1;
+                    }
+                }
+            }
+            link = QStringView{link.begin(), std::max(link.end(), lastClose)};
+            rest = QStringView{rest.begin(), std::max(rest.end(), lastClose)};
+        }
         result = Parsed{
             .protocol = protocol,
             .host = host,

src/common/LinkParser.hpp

@@ -16,8 +16,9 @@ namespace chatterino::linkparser {
 ///
 /// Prefixes and suffixes are almost identical to the ones in GitHub Flavored
 /// Markdown (GFM - https://github.github.com/gfm/#autolinks-extension-).
-/// The main differences are that '_' isn't a valid suffix and parentheses
+/// The main difference is that '_' isn't a valid suffix.
-/// aren't counted (e.g. "(a.com/(foo)! would result in "a.com/(foo").
+/// Parentheses are counted inside the @a rest: parsing "(a.com/(foo))" would
+/// result in the link "a.com/(foo)".
 /// Matching is done case insensitive (e.g. "HTTp://a.com" would be valid).
 ///
 /// A @a protocol can either be empty, "http://", or "https://".

tests/src/LinkParser.cpp

@@ -23,8 +23,8 @@ struct Case {
             "", "_", "__", "<", "<<", "<_<", "(((", "<*_~(", "**", "~~",
         };
         QStringList suffixes{
-            "",  ">", "?",  "!",   ".",      ",",  ":",
+            "",   ">",   "?",      "!",  ".",  ",",  ":",  "*",    "~",
-            "*", "~", ">>", "?!.", "~~,*!?", "**",
+            ">>", "?!.", "~~,*!?", "**", ").", "),", ",)", ")),.", ")?",
         };
 
         for (const auto &prefix : prefixes)
@@ -89,14 +89,21 @@ TEST(LinkParser, parseDomainLinks)
         {"", "https.cat"},
         {"", "httpsd.cat"},
         {"", "http.cat", "/200"},
-        {"", "http.cat", "/200("},
+        {"", "http.cat", "/200()"},
-        {"", "a.com", "?("},
+        {"", "a.com", "?()"},
-        {"", "a.com", "#("},
+        {"", "a.com", "#()"},
         {"", "a.com", "/__my_user__"},
         {"", "a.b.c.-._.1.com", ""},
         {"", "0123456789.com", ""},
         {"", "ABCDEFGHIJKLMNOPQRSTUVWXYZ.com", ""},
         {"", "abcdefghijklmnopqrstuvwxyz.com", ""},
+        {"", "example.com", "/foo(bar)"},
+        {"", "example.com", "/foo((bar))"},
+        {"", "example.com", "/(f)(o)(o)(b)(a)r"},
+        {"", "example.com", "/foobar()()"},
+        {"", "example.com", "/foobar()(())baz"},
+        {"", "example.com", "/(foo)"},
+        {"", "example.com", "/()"},
         // non-ASCII characters are allowed
         {"", u"köln.de"_s, ""},
         {"", u"ü.com"_s, ""},
@@ -246,6 +253,8 @@ TEST(LinkParser, doesntParseInvalidLinks)
         "@@@.com",
         "%%%.com",
         "*.com",
+        "example.com(foo)",
+        "example.com()",
     };
 
     for (const auto &input : inputs)