fix: treat backslash as escape character only inside quoted strings, matching Jinja2 behaviour (fixes #1304)

src/main/java/com/hubspot/jinjava/LegacyOverrides.java

@@ -21,6 +21,7 @@ public interface LegacyOverrides extends WithLegacyOverrides {
     .withAllowAdjacentTextNodes(true)
     .withUseTrimmingForNotesAndExpressions(true)
     .withKeepNullableLoopValues(true)
+    .withHandleBackslashInQuotesOnly(true)
     .build();
   LegacyOverrides ALL = new Builder()
     .withEvaluateMapKeys(true)
@@ -32,6 +33,7 @@ public interface LegacyOverrides extends WithLegacyOverrides {
     .withAllowAdjacentTextNodes(true)
     .withUseTrimmingForNotesAndExpressions(true)
     .withKeepNullableLoopValues(true)
+    .withHandleBackslashInQuotesOnly(true)
     .build();
 
   @Value.Default
@@ -79,6 +81,23 @@ default boolean isKeepNullableLoopValues() {
     return false;
   }
 
+  /**
+   * When {@code true}, the token scanner treats backslash as an escape character
+   * only inside quoted string literals, leaving bare backslashes outside quotes
+   * untouched for the expression parser (JUEL) to handle. This matches the
+   * behaviour of Python's Jinja2, where the template scanner is not responsible
+   * for backslash interpretation at all.
+   *
+   * <p>When {@code false} (the default), the scanner consumes a backslash and
+   * the following character unconditionally, regardless of quote context. This
+   * is the legacy Jinjava behaviour, which prevents closing delimiters from
+   * being recognized after a backslash but diverges from Jinja2.
+   */
+  @Value.Default
+  default boolean isHandleBackslashInQuotesOnly() {
+    return false;
+  }
+
   class Builder extends ImmutableLegacyOverrides.Builder {}
 
   static Builder newBuilder() {

src/main/java/com/hubspot/jinjava/tree/TreeParser.java

@@ -32,13 +32,15 @@
 import com.hubspot.jinjava.lib.tag.FlexibleTag;
 import com.hubspot.jinjava.lib.tag.Tag;
 import com.hubspot.jinjava.tree.parse.ExpressionToken;
+import com.hubspot.jinjava.tree.parse.StringTokenScanner;
 import com.hubspot.jinjava.tree.parse.TagToken;
 import com.hubspot.jinjava.tree.parse.TextToken;
 import com.hubspot.jinjava.tree.parse.Token;
 import com.hubspot.jinjava.tree.parse.TokenScanner;
 import com.hubspot.jinjava.tree.parse.TokenScannerSymbols;
 import com.hubspot.jinjava.tree.parse.UnclosedToken;
 import com.hubspot.jinjava.tree.parse.WhitespaceControlParser;
+import java.util.Iterator;
 import org.apache.commons.lang3.StringUtils;
 
 public class TreeParser {
@@ -52,7 +54,7 @@ public class TreeParser {
 
   public TreeParser(JinjavaInterpreter interpreter, String input) {
     this.scanner =
-      Iterators.peekingIterator(new TokenScanner(input, interpreter.getConfig()));
+      Iterators.peekingIterator(createScanner(input, interpreter.getConfig()));
     this.interpreter = interpreter;
     this.symbols = interpreter.getConfig().getTokenScannerSymbols();
     this.whitespaceControlParser =
@@ -104,6 +106,13 @@ public Node buildTree() {
     return root;
   }
 
+  private static Iterator<Token> createScanner(String input, JinjavaConfig config) {
+    if (config.getTokenScannerSymbols().isStringBased()) {
+      return new StringTokenScanner(input, config);
+    }
+    return new TokenScanner(input, config);
+  }
+
   /**
    * @return null if EOF or error
    */

src/main/java/com/hubspot/jinjava/tree/parse/ExpressionToken.java

@@ -54,7 +54,14 @@ public int getType() {
 
   @Override
   protected void parse() {
-    this.expr = WhitespaceUtils.unwrap(image, "{{", "}}");
+    // Use the symbols-derived delimiter strings instead of the hardcoded "{{" / "}}"
+    // so that custom delimiters (e.g. "\VAR{" / "}") are stripped correctly.
+    this.expr =
+      WhitespaceUtils.unwrap(
+        image,
+        getSymbols().getExpressionStart(),
+        getSymbols().getExpressionEnd()
+      );
     this.expr = handleTrim(expr);
     this.expr = StringUtils.trimToEmpty(this.expr);
   }

src/main/java/com/hubspot/jinjava/tree/parse/NoteToken.java

@@ -48,8 +48,11 @@ public int getType() {
    */
   @Override
   protected void parse() {
-    if (image.length() > 4) { // {# #}
-      handleTrim(image.substring(2, image.length() - 2));
+    int startLen = getSymbols().getCommentStartLength();
+    int endLen = getSymbols().getCommentEndLength();
+
+    if (image.length() > startLen + endLen) {
+      handleTrim(image.substring(startLen, image.length() - endLen));
     }
     content = "";
   }