Split PostgreSQL hash operators from adjacent identifiers

Author: minleejae

src/main/jjtree/net/sf/jsqlparser/parser/JSqlParserCC.jjt

@@ -1723,6 +1723,14 @@ TOKEN_MGR_DECLS : {
         token.endColumn = input_stream.getEndColumn();
     }
 
+    /** Splits a longest-match identifier without losing token positions. */
+    private void truncateIdentifierToken(Token token, int length) {
+        input_stream.backup(token.image.length() - length);
+        token.image = token.image.substring(0, length);
+        token.endLine = input_stream.getEndLine();
+        token.endColumn = input_stream.getEndColumn();
+    }
+
     /**
      * Consumes the body of a block comment after the opening delimiter has been matched,
      * honouring nesting, up to and including the outermost closing delimiter. Then backs
@@ -2462,8 +2470,19 @@ TOKEN:
             && Boolean.TRUE.equals(configuration.getValue(Feature.allowHashLineComments))) {
         int hashIndex = matchedToken.image.indexOf('#');
         if (hashIndex > 0) {
-            input_stream.backup(matchedToken.image.length() - hashIndex);
-            matchedToken.image = matchedToken.image.substring(0, hashIndex);
+            truncateIdentifierToken(matchedToken, hashIndex);
+        }
+    }
+    // PostgreSQL does not allow # in unquoted identifiers. Re-lex it as an
+    // operator, including #> and #>>, even when it touches the left operand.
+    if (matchedToken.kind == S_IDENTIFIER
+            && AbstractJSqlParser.Dialect.POSTGRESQL.name().equals(configuration.getValue(Feature.dialect))) {
+        int hashIndex = matchedToken.image.indexOf('#');
+        if (hashIndex > 0) {
+            truncateIdentifierToken(matchedToken, hashIndex);
+        } else if (hashIndex == 0) {
+            truncateIdentifierToken(matchedToken, 1);
+            matchedToken.kind = S_HASH_OPERATOR;
         }
     }
 }
@@ -2484,8 +2503,7 @@ TOKEN:
             && Boolean.TRUE.equals(configuration.getValue(Feature.allowHashLineComments))) {
         int hashIndex = matchedToken.image.indexOf('#');
         if (hashIndex > 0) {
-            input_stream.backup(matchedToken.image.length() - hashIndex);
-            matchedToken.image = matchedToken.image.substring(0, hashIndex);
+            truncateIdentifierToken(matchedToken, hashIndex);
         }
     }
 }

src/site/sphinx/usage.rst

@@ -716,6 +716,11 @@ One grammar covers every supported RDBMS, but a few pieces of syntax mean differ
 
 Features set explicitly *after* the preset win over it.
 
+With ``Dialect.POSTGRESQL``, ``#`` terminates an unquoted identifier, so JSON
+operators such as ``js#>>'{a}'`` and ``js#>'{a}'`` work without surrounding
+spaces. Quote identifiers containing ``#``, for example ``"js#"``. Other
+dialects retain their existing identifier and hash-comment rules.
+
 With ``Dialect.SQLSERVER``, ``PRIMARY KEY NONCLUSTERED (id)`` and
 ``UNIQUE CLUSTERED (id)`` store their clustering option in ``Index.getClustering()``
 for both ``CREATE TABLE`` and ``ALTER TABLE``. Without that dialect, these words

src/test/java/net/sf/jsqlparser/parser/PostgreSqlHashOperatorTest.java

@@ -0,0 +1,99 @@
+/*-
+ * #%L
+ * JSQLParser library
+ * %%
+ * Copyright (C) 2004 - 2026 JSQLParser
+ * %%
+ * Dual licensed under GNU LGPL 2.1 or Apache License 2.0
+ * #L%
+ */
+package net.sf.jsqlparser.parser;
+
+import static org.junit.jupiter.api.Assertions.*;
+
+import net.sf.jsqlparser.expression.JsonExpression;
+import net.sf.jsqlparser.expression.operators.arithmetic.BitwiseRightShift;
+import net.sf.jsqlparser.expression.operators.relational.Intersects;
+import net.sf.jsqlparser.expression.operators.relational.NotEqualsTo;
+import net.sf.jsqlparser.parser.AbstractJSqlParser.Dialect;
+import net.sf.jsqlparser.schema.Column;
+import net.sf.jsqlparser.statement.select.PlainSelect;
+import net.sf.jsqlparser.util.deparser.StatementDeParser;
+import org.junit.jupiter.api.Test;
+import org.junit.jupiter.params.ParameterizedTest;
+import org.junit.jupiter.params.provider.ValueSource;
+
+class PostgreSqlHashOperatorTest {
+    private static PlainSelect parse(String sql) throws Exception {
+        return (PlainSelect) CCJSqlParserUtil.parse(sql, p -> p.withDialect(Dialect.POSTGRESQL));
+    }
+
+    @ParameterizedTest
+    @ValueSource(strings = {"js#>>'{a,b,1}'", "js #>> '{a,b,1}'", "js#>> '{a,b,1}'",
+            "t.js#>>'{a,b,1}'", "\"js#\"#>>'{a,b,1}'", "js/*comment*/#>>'{a,b,1}'"})
+    void preservesJsonPrecedenceWithOrWithoutSpacesIssue2163(String expression) throws Exception {
+        PlainSelect select = parse("SELECT * FROM t WHERE " + expression + " <> 'bar'");
+        NotEqualsTo condition = (NotEqualsTo) select.getWhere();
+        assertInstanceOf(JsonExpression.class, condition.getLeftExpression());
+        assertTrue(select.toString().contains(" #>> "));
+        StringBuilder output = new StringBuilder();
+        select.accept(new StatementDeParser(output));
+        assertEquals(select.toString(), output.toString());
+        assertEquals(select.toString(), parse(output.toString()).toString());
+    }
+
+    @ParameterizedTest
+    @ValueSource(strings = {"js#>'{a}'", "js#>'{a}'#>>'{b}'"})
+    void recognizesOtherHashJsonOperators(String expression) throws Exception {
+        PlainSelect select = parse("SELECT " + expression + " FROM t");
+        assertInstanceOf(JsonExpression.class, select.getSelectItem(0).getExpression());
+        assertEquals(select.toString(), parse(select.toString()).toString());
+    }
+
+    @Test
+    void splitsHashXorAndPreservesQuotedIdentifiersAndLiterals() throws Exception {
+        PlainSelect select = parse("SELECT a#b, a#2, \"a#b\", '#>>', $$a#b$$ FROM t");
+        assertInstanceOf(Intersects.class, select.getSelectItem(0).getExpression());
+        assertInstanceOf(Intersects.class, select.getSelectItem(1).getExpression());
+        assertEquals("\"a#b\"", ((Column) select.getSelectItem(2).getExpression()).getColumnName());
+        assertEquals(select.toString(), parse(select.toString()).toString());
+    }
+
+    @Test
+    void leavesLegacyAndSqlServerNamesAndMysqlCommentsIntact() throws Exception {
+        for (Dialect dialect : Dialect.values()) {
+            if (dialect != Dialect.POSTGRESQL && dialect != Dialect.MYSQL
+                    && dialect != Dialect.MARIADB && dialect != Dialect.BIGQUERY) {
+                PlainSelect select = (PlainSelect) CCJSqlParserUtil.parse("SELECT a#b FROM #temp",
+                        p -> p.withDialect(dialect));
+                assertEquals("a#b",
+                        ((Column) select.getSelectItem(0).getExpression()).getColumnName());
+            }
+        }
+        PlainSelect legacy = (PlainSelect) CCJSqlParserUtil.parse("SELECT js#>>'{a}' FROM t");
+        assertInstanceOf(BitwiseRightShift.class, legacy.getSelectItem(0).getExpression());
+        PlainSelect mysql = (PlainSelect) CCJSqlParserUtil.parse("SELECT a#comment\nFROM t",
+                p -> p.withDialect(Dialect.MYSQL));
+        assertEquals("SELECT a FROM t", mysql.toString());
+    }
+
+    @Test
+    void keepsTokenOffsetsAfterSplittingAndAcrossStatements() throws Exception {
+        CCJSqlParser parser = CCJSqlParserUtil.newParser("js#>>'{}'")
+                .withDialect(Dialect.POSTGRESQL);
+        Token name = parser.getNextToken();
+        Token operator = parser.getNextToken();
+        assertEquals("js", name.image);
+        assertEquals(1, name.beginColumn);
+        assertEquals(2, name.endColumn);
+        assertEquals(1, name.absoluteBegin);
+        assertEquals(3, name.absoluteEnd);
+        assertEquals("#>>", operator.image);
+        assertEquals(3, operator.beginColumn);
+        assertEquals(5, operator.endColumn);
+        assertEquals(3, operator.absoluteBegin);
+        assertEquals(6, operator.absoluteEnd);
+        assertEquals(2, CCJSqlParserUtil.parseStatements("SELECT js#>>'{}' FROM t; SELECT 1;",
+                p -> p.withDialect(Dialect.POSTGRESQL)).size());
+    }
+}