antlr/antlr4 · error · ArgumentException

unterminated tag in pattern: {pattern}

Error message

unterminated tag in pattern: {pattern}

What it means

While splitting a pattern into chunks, the tag scanner found more start delimiters than stop delimiters — a '<' (or custom start) with no matching '>' before the end of the pattern. The scanner also honors the escape character (default '\'), so an escaped '>' does not count as a stop.

Source

Thrown at runtime/CSharp/src/Tree/Pattern/ParseTreePatternMatcher.cs:634

                            if (p == pattern.IndexOf(stop, p))
                            {
                                stops.Add(p);
                                p += stop.Length;
                            }
                            else
                            {
                                p++;
                            }
                        }
                    }
                }
            }
            //		System.out.println("");
            //		System.out.println(starts);
            //		System.out.println(stops);
            if (starts.Count > stops.Count)
            {
                throw new ArgumentException("unterminated tag in pattern: " + pattern);
            }
            if (starts.Count < stops.Count)
            {
                throw new ArgumentException("missing start tag in pattern: " + pattern);
            }
            int ntags = starts.Count;
            for (int i = 0; i < ntags; i++)
            {
                if (starts[i] >= stops[i])
                {
                    throw new ArgumentException("tag delimiters out of order in pattern: " + pattern);
                }
            }
            // collect into chunks now
            if (ntags == 0)
            {
                string text = Sharpen.Runtime.Substring(pattern, 0, n);
                chunks.Add(new TextChunk(text));

View on GitHub (pinned to 7d5770395b)

Solutions

1. Balance the delimiters: every '<' needs a matching '>' in the pattern.
2. Escape literal angle brackets that are text, not tags: use \< and \> so the scanner ignores them.
3. If angle brackets are common in your target text, choose different delimiters with SetDelimiters('[', ']', '\\').
4. Validate user-supplied patterns by counting unescaped starts/stops before compiling.

Example fix

// before
var p = parser.CompileParseTreePattern("x < y", RULE_expr, null);

// after
var p = parser.CompileParseTreePattern("x \< y", RULE_expr, null);

// Count unescaped delimiters before compiling
int CountUnescaped(string s, char c) { int n = 0; for (int i = 0; i < s.Length; i++) { if (s[i] == '\\') i++; else if (s[i] == c) n++; } return n; }
if (CountUnescaped(pattern, '<') > CountUnescaped(pattern, '>')) throw new ArgumentException("unterminated tag");

try { matcher.Compile(pattern); } catch (ArgumentException ex) when (ex.Message.Contains("unterminated tag")) { /* prompt user to close or escape the '<' */ }

When it happens

Trigger: Pattern strings like "<ID = <expr>" or "a < b" (unescaped '<' used as text) passed to Compile; a custom escape via SetDelimiters that no longer matches what the pattern author used.

Common situations: Patterns intended to match source text containing literal angle brackets (HTML/XML/comparison operators) without escaping; user-supplied patterns from config; truncation of long patterns cutting off a closing '>'.

Related errors

• missing start tag in pattern: {pattern}
• tag delimiters out of order in pattern: {pattern}
• pattern cannot be null
• start cannot be null or empty
• stop cannot be null or empty

AI-assisted analysis of antlr/antlr4@7d5770395b (2026-08-14). Data as JSON: /api/errors/06fcf4e20435eeaf. Report an issue: GitHub.