ICU-23313 UnicodeSet: binary query negation with NOT EQUAL TO

Author: eggrobin

icu4c/source/common/uniset_props.cpp

@@ -693,6 +693,7 @@ class UnicodeSet::Lexer {
         std::optional<int32_t> queryOperatorPosition;
         int32_t queryExpressionStart = parsePosition_.getIndex();
         bool exteriorlyNegated = false;
+        bool interiorlyNegated = false;
         UBool unusedEscaped;
         // Do not skip whitespace so we can recognize unspaced :].  Lex escapes and
         // named-element: while ICU does not support string-valued properties and thus has no
@@ -742,7 +743,14 @@ class UnicodeSet::Lexer {
                 // Neither a named-element nor an escaped-element can be part of a closing :].
                 lastUnescaped = -1;
             } else if (!queryOperatorPosition.has_value() && lastUnescaped == u'=') {
-                // TODO(egg): Propose and add support for ≠.
+                queryOperatorPosition = parsePosition_.getIndex() - 1;
+            } else if (!queryOperatorPosition.has_value() && lastUnescaped == u'≠') {
+                if (exteriorlyNegated) {
+                    // Reject doubly negated property queries.
+                    errorCode = U_ILLEGAL_ARGUMENT_ERROR;
+                    return {};
+                }
+                interiorlyNegated = true;
                 queryOperatorPosition = parsePosition_.getIndex() - 1;
             } else if ((first == u'[' && penultimateUnescaped == u':' && lastUnescaped == u']') ||
                        (first == u'\\' && lastUnescaped == u'}')) {
@@ -772,7 +780,7 @@ class UnicodeSet::Lexer {
                     pattern_.tempSubStringBetween(queryExpressionStart,
                                                   queryOperatorPosition.value_or(queryExpressionLimit)),
                     propertyPredicate, errorCode);
-                if (exteriorlyNegated) {
+                if (exteriorlyNegated != interiorlyNegated) {
                     result.complement().removeAllStrings();
                 }
                 result.setPattern(pattern_.tempSubStringBetween(queryStart, parsePosition_.getIndex()));

icu4c/source/test/intltest/usettest.cpp

@@ -1118,6 +1118,34 @@ void UnicodeSetTest::TestPropertySet() {
         expectContainment(UnicodeString(DATA[i], -1, US_INV), CharsToUnicodeString(DATA[i+1]),
                           CharsToUnicodeString(DATA[i+2]));
     }
+    {
+        UErrorCode status = U_ZERO_ERROR;
+        UnicodeSet s1(u"[:Noncharacter_Code_Point≠No:]", status);
+        UnicodeSet s2(u"[:Noncharacter_Code_Point:]", status);
+        TEST_ASSERT_SUCCESS(status);
+        TEST_ASSERT(s1 == s2);
+    }
+    {
+        UErrorCode status = U_ZERO_ERROR;
+        UnicodeSet s1(uR"(\p{Noncharacter_Code_Point≠No})", status);
+        UnicodeSet s2(uR"(\p{Noncharacter_Code_Point})", status);
+        TEST_ASSERT_SUCCESS(status);
+        TEST_ASSERT(s1 == s2);
+    }
+    {
+        UErrorCode status = U_ZERO_ERROR;
+        UnicodeSet s1(uR"(\p{dt≠can})", status);
+        UnicodeSet s2(uR"(\P{dt=can})", status);
+        TEST_ASSERT_SUCCESS(status);
+        TEST_ASSERT(s1 == s2);
+    }
+    {
+        UErrorCode status = U_ZERO_ERROR;
+        UnicodeSet s1(uR"([:dt≠can:])", status);
+        UnicodeSet s2(uR"([:^dt=can:])", status);
+        TEST_ASSERT_SUCCESS(status);
+        TEST_ASSERT(s1 == s2);
+    }
 }
 
 /**
@@ -4778,6 +4806,9 @@ void UnicodeSetTest::TestParseErrors() {
             uR"(\p{Uppercase_Letter=})",
             // Well-formed in ICU 78 and earlier, disallowed by ICU-23306.
             uR"([: ^general category = punctuation :])",
+            // Doubly negated property queries.
+            uR"(\P{Decomposition_Type≠compat})",
+            u"[:^Noncharacter_Code_Point≠No:]",
         }) {
         UErrorCode errorCode = U_ZERO_ERROR;
         const UnicodeSet set(expression, errorCode);