refactor(translator): Optimize LIKE queries to not always use =~.

When possible, use `CONTAINS`, `STARTS WITH` or `ENDS WITH`. Signed-off-by: Michael Simons <[email protected]>
neo4j · Jan 20, 2025 · b1e3435 · b1e3435
1 parent 03bc391
commit b1e3435
Show file tree

Hide file tree

Showing 3 changed files with 61 additions and 9 deletions.
diff --git a/neo4j-jdbc-translator/impl/src/main/java/org/neo4j/jdbc/translator/impl/SqlToCypher.java b/neo4j-jdbc-translator/impl/src/main/java/org/neo4j/jdbc/translator/impl/SqlToCypher.java
@@ -33,6 +33,7 @@
 import java.util.concurrent.atomic.AtomicInteger;
 import java.util.function.BiFunction;
 import java.util.function.Function;
+import java.util.function.LongSupplier;
 import java.util.function.Predicate;
 import java.util.function.Supplier;
 import java.util.logging.Level;
@@ -104,6 +105,7 @@ final class SqlToCypher implements Translator {
 	static final Pattern ELEMENT_ID_PATTERN = Pattern.compile("(?i)v\\$(?:(?<prefix>.+?)_)?id");
 	static final String ELEMENT_ID_FUNCTION_NAME = "elementId";
 	static final String ELEMENT_ID_ALIAS = "v$id";
+	static final Pattern PERCENT_OR_UNDERSCORE = Pattern.compile("[%_]");
 
 	static {
 		Logger.getLogger("org.jooq.Constants").setLevel(Level.WARNING);
@@ -1186,14 +1188,7 @@ else if (c instanceof QOM.RowIsNotNull e) {
 						.orElseThrow();
 				}
 				else if (c instanceof QOM.Like like) {
-					Expression rhs;
-					if (like.$arg2() instanceof Param<?> p && p.$inline() && p.getValue() instanceof String s) {
-						rhs = Cypher.literalOf(s.replace("%", ".*").replace("_", "."));
-					}
-					else {
-						rhs = expression(like.$arg2());
-					}
-					return expression(like.$arg1()).matches(rhs);
+					return like(like);
 				}
 				else if (c instanceof QOM.FieldCondition fc && fc.$field() instanceof Param<Boolean> param) {
 					return (Boolean.TRUE.equals(param.getValue()) ? Cypher.literalTrue() : Cypher.literalFalse())
@@ -1212,6 +1207,41 @@ else if (c instanceof QOM.InList<?> il) {
 			}
 		}
 
+		private Condition like(QOM.Like like) {
+			Expression rhs;
+			Expression lhs = expression(like.$arg1());
+			if (like.$arg2() instanceof Param<?> p && p.$inline() && p.getValue() instanceof String s) {
+				var sw = s.startsWith("%");
+				var ew = s.endsWith("%");
+				var length = s.length();
+				var cnt = new LongSupplier() {
+					Long value = null;
+
+					@Override
+					public long getAsLong() {
+						if (this.value == null) {
+							this.value = PERCENT_OR_UNDERSCORE.matcher(s).results().count();
+						}
+						return this.value;
+					}
+				};
+				if (sw && ew && length > 2 && cnt.getAsLong() == 2) {
+					return lhs.contains(Cypher.literalOf(s.substring(1, length - 1)));
+				}
+				else if (sw && length > 1 && cnt.getAsLong() == 1) {
+					return lhs.endsWith(Cypher.literalOf(s.substring(1)));
+				}
+				else if (ew && length > 1 && cnt.getAsLong() == 1) {
+					return lhs.startsWith(Cypher.literalOf(s.substring(0, length - 1)));
+				}
+				rhs = Cypher.literalOf(s.replaceAll("%+", ".*").replace("_", "."));
+			}
+			else {
+				rhs = expression(like.$arg2());
+			}
+			return lhs.matches(rhs);
+		}
+
 		private Condition rowCondition(Row r1, Row r2,
 				BiFunction<? super Expression, ? super Expression, ? extends Condition> comp,
 				BiFunction<? super Expression, ? super Expression, ? extends Condition> last) {

diff --git a/...j-jdbc-translator/impl/src/test/java/org/neo4j/jdbc/translator/impl/SqlToCypherTests.java b/...j-jdbc-translator/impl/src/test/java/org/neo4j/jdbc/translator/impl/SqlToCypherTests.java
@@ -407,6 +407,28 @@ void escapingShouldWork(Boolean prettyPrint, Boolean alwaysEscapeNames, String e
 		assertThat(cypher).isEqualTo(expected.replace("$", cfg.isPrettyPrint() ? System.lineSeparator() : " "));
 	}
 
+	@ParameterizedTest
+	@CsvSource(
+			textBlock = """
+					SELECT * FROM blub b WHERE name like '%Test%', MATCH (b:blub) WHERE b.name CONTAINS 'Test' RETURN *
+					SELECT * FROM blub b WHERE name like '%Test', MATCH (b:blub) WHERE b.name ENDS WITH 'Test' RETURN *
+					SELECT * FROM blub b WHERE name like 'Test%', MATCH (b:blub) WHERE b.name STARTS WITH 'Test' RETURN *
+					SELECT * FROM blub b WHERE name like 'This is _ %Test%', MATCH (b:blub) WHERE b.name =~ 'This is . .*Test.*' RETURN *
+					SELECT * FROM blub b WHERE name like '%', MATCH (b:blub) WHERE b.name =~ '.*' RETURN *
+					SELECT * FROM blub b WHERE name like '%%', MATCH (b:blub) WHERE b.name =~ '.*' RETURN *
+					SELECT * FROM blub b WHERE name like '%%%', MATCH (b:blub) WHERE b.name =~ '.*' RETURN *
+					SELECT * FROM blub b WHERE name like '_', MATCH (b:blub) WHERE b.name =~ '.' RETURN *
+					SELECT * FROM blub b WHERE name like '__', MATCH (b:blub) WHERE b.name =~ '..' RETURN *
+					SELECT * FROM blub b WHERE name like '___', MATCH (b:blub) WHERE b.name =~ '...' RETURN *
+					SELECT * FROM blub b WHERE name like '%_%', MATCH (b:blub) WHERE b.name =~ '.*..*' RETURN *
+					SELECT * FROM blub b WHERE name like '%ein%schöner%Name%', MATCH (b:blub) WHERE b.name =~ '.*ein.*schöner.*Name.*' RETURN *
+					""")
+	void likeShouldBeHandledNicely(String sql, String expected) {
+
+		var translator = SqlToCypher.defaultTranslator();
+		assertThat(translator.translate(sql)).isEqualTo(expected);
+	}
+
 	private static class TestDataExtractor extends Treeprocessor {
 
 		private final List<TestData> testData = new ArrayList<>();

diff --git a/neo4j-jdbc-translator/impl/src/test/resources/predicates.adoc b/neo4j-jdbc-translator/impl/src/test/resources/predicates.adoc
@@ -186,6 +186,6 @@ will be translated into a regular expressions, replacing the `%` with `.*`:
 
 [source,cypher,id=p5_0_expected]
 ----
-MATCH (m:`movies`) WHERE m.title =~ '.*Matrix.*' OR m.title =~ 'M.trix'
+MATCH (m:`movies`) WHERE m.title CONTAINS 'Matrix' OR m.title =~ 'M.trix'
 RETURN *
 ----