ICU-22890 Add lookahead to parser

This fixes tests that were broken by the original fix in this PR.
unicode-org · Sep 14, 2024 · 8699988 · 8699988
1 parent 2a9ea63
commit 8699988
Show file tree

Hide file tree

Showing 2 changed files with 26 additions and 22 deletions.
diff --git a/icu4c/source/i18n/messageformat2_parser.cpp b/icu4c/source/i18n/messageformat2_parser.cpp
@@ -507,6 +507,17 @@ void Parser::parseTokenWithWhitespace(UChar32 c, UErrorCode& errorCode) {
     CHECK_BOUNDS(source, index, parseError, errorCode);
 }
 
+// Terrible hack to get around the ambiguity between unsupported keywords
+// and supported keywords
+bool Parser::nextIs(const std::u16string_view &keyword) const {
+    for(int32_t i = 0; i < (int32_t) keyword.length(); i++) {
+        if (!inBounds(source, index + i) || source[index + i] != keyword[i]) {
+            return false;
+        }
+    }
+    return true;
+}
+
 /*
   Consumes a non-empty sequence of `name-char`s, the first of which is
   also a `name-start`.
@@ -1653,6 +1664,10 @@ void Parser::parseUnsupportedStatement(UErrorCode& status) {
     UnicodeString keyword(PERIOD);
     normalizedInput += UnicodeString(PERIOD);
     index++;
+
+    // A message just consisting of '.' is a parse error
+    CHECK_BOUNDS(source, index, parseError, status);
+
     keyword += parseName(status);
     builder.setKeyword(keyword);
 
@@ -1718,15 +1733,6 @@ void Parser::parseUnsupportedStatement(UErrorCode& status) {
     dataModel.addUnsupportedStatement(builder.build(status), status);
 }
 
-// Terrible hack to get around the ambiguity between `matcher` and `reserved-statement`
-bool Parser::nextIsMatch() const {
-    for(int32_t i = 0; i < 6; i++) {
-        if (!inBounds(source, index + i) || source[index + i] != ID_MATCH[i]) {
-            return false;
-        }
-    }
-    return true;
-}
 /*
   Consume a possibly-empty sequence of declarations separated by whitespace;
   each declaration matches the `declaration` nonterminal in the grammar
@@ -1739,15 +1745,14 @@ void Parser::parseDeclarations(UErrorCode& status) {
     CHECK_BOUNDS(source, index, parseError, status);
 
     while (source[index] == PERIOD) {
-        CHECK_BOUNDS(source, index + 1, parseError, status);
         if (source[index + 1] == ID_LOCAL[1]) {
             parseLocalDeclaration(status);
         } else if (source[index + 1] == ID_INPUT[1]) {
             parseInputDeclaration(status);
         } else {
             // Unsupported statement
             // Lookahead is needed to disambiguate this from a `match`
-            if (!nextIsMatch()) {
+            if (!nextIs(u".match")) {
                 parseUnsupportedStatement(status);
             } else {
                 // Done parsing declarations
@@ -1786,8 +1791,10 @@ UnicodeString Parser::parseText(UErrorCode& status) {
     }
 
     if (!(isTextChar(source[index]) || source[index] == BACKSLASH)) {
-        // Error -- text is expected here
-        ERROR(parseError, status, index);
+        if (source[index] != RIGHT_CURLY_BRACE) {
+            // Error -- text is expected here
+            ERROR(parseError, status, index);
+        }
         return str;
     }
 
@@ -2088,8 +2095,6 @@ Pattern Parser::parseSimpleMessage(UErrorCode& status) {
 void Parser::parseSelectors(UErrorCode& status) {
     CHECK_ERROR(status);
 
-    U_ASSERT(inBounds(source, index));
-
     parseToken(ID_MATCH, status);
 
     bool empty = true;
@@ -2213,12 +2218,11 @@ void Parser::parseBody(UErrorCode& status) {
         dataModel.setPattern(parseQuotedPattern(status));
         break;
     }
-    case ID_MATCH[0]: {
-        // Selectors
-        parseSelectors(status);
-        return;
-    }
     default: {
+        if (nextIs(u".match")) {
+            parseSelectors(status);
+            return;
+        }
         ERROR(parseError, status, index);
         errorPattern(status);
         return;
@@ -2240,7 +2244,7 @@ void Parser::parse(UParseError &parseErrorResult, UErrorCode& status) {
             || (index < static_cast<uint32_t>(source.length()) + 1
                 && source[index] == LEFT_CURLY_BRACE
                 && source[index + 1] == LEFT_CURLY_BRACE)) {
-            // A complex message begins with a '.' or '{'
+            // A complex message begins with a keyword or '{'
             parseDeclarations(status);
             parseBody(status);
             simple = false;

diff --git a/icu4c/source/i18n/messageformat2_parser.h b/icu4c/source/i18n/messageformat2_parser.h
@@ -115,7 +115,7 @@ namespace message2 {
 	void parseToken(const UChar32 (&)[N], UErrorCode&);
 	template <int32_t N>
 	void parseTokenWithWhitespace(const UChar32 (&)[N], UErrorCode&);
-        bool nextIsMatch() const;
+        bool nextIs(const std::u16string_view&) const;
 	UnicodeString parseName(UErrorCode&);
         UnicodeString parseIdentifier(UErrorCode&);
         UnicodeString parseDigits(UErrorCode&);