Merge pull request #18628 from vespa-engine/arnej/workaround-libyell-bug

don't call accentDrop at all for empty input

2 files changed, 28 insertions, 2 deletions

diff --git a/indexinglanguage/src/main/java/com/yahoo/vespa/indexinglanguage/expressions/NormalizeExpression.java b/indexinglanguage/src/main/java/com/yahoo/vespa/indexinglanguage/expressions/NormalizeExpression.java
index 78a7a858aba..fa68d45d110 100644
--- a/indexinglanguage/src/main/java/com/yahoo/vespa/indexinglanguage/expressions/NormalizeExpression.java
+++ b/indexinglanguage/src/main/java/com/yahoo/vespa/indexinglanguage/expressions/NormalizeExpression.java
@@ -43,6 +43,9 @@ public final class NormalizeExpression extends Expression {
     protected void doExecute(ExecutionContext context) {
         Transformer transformer = linguistics.getTransformer();
         var orig = String.valueOf(context.getValue());
+        if (orig.isEmpty()) {
+            return; // must be a no-op for all linguistics/language combinations
+        }
         var lang = context.resolveLanguage(linguistics);
         var transformed = transformer.accentDrop(orig, lang);
         try {
diff --git a/indexinglanguage/src/test/java/com/yahoo/vespa/indexinglanguage/expressions/NormalizeTestCase.java b/indexinglanguage/src/test/java/com/yahoo/vespa/indexinglanguage/expressions/NormalizeTestCase.java
index 3ad1b129b4d..9f6ba4a1554 100644
--- a/indexinglanguage/src/test/java/com/yahoo/vespa/indexinglanguage/expressions/NormalizeTestCase.java
+++ b/indexinglanguage/src/test/java/com/yahoo/vespa/indexinglanguage/expressions/NormalizeTestCase.java
@@ -60,18 +60,24 @@ public class NormalizeTestCase {
     }
 
     class MyMockTransformer implements Transformer {
-        boolean first = true;
+        public boolean first = true;
         @Override
         public String accentDrop(String input, Language language) {
             if (first) {
                 first = false;
-                return input.replace(' ', '\u0008');
+                return "\u0008";
             } else {
                 return input.replace(' ', '/');
             }
         }
     }        
 
+    boolean getFirst(Transformer t) {
+        assertTrue(t instanceof MyMockTransformer);
+        var mmt = (MyMockTransformer)t;
+        return mmt.first;
+    }
+
     class MyMockLinguistics extends SimpleLinguistics {
         private Transformer transformer = new MyMockTransformer();
         @Override
@@ -86,9 +92,26 @@ public class NormalizeTestCase {
         ctx.setLanguage(Language.ENGLISH);
         ctx.setValue(new StringFieldValue("bad norm"));
         var linguistics = new MyMockLinguistics();
+        assertTrue(getFirst(linguistics.getTransformer()));
         new NormalizeExpression(linguistics).execute(ctx);
         FieldValue val = ctx.getValue();
         assertTrue(val instanceof StringFieldValue);
         assertEquals("bad/norm", ((StringFieldValue)val).getString());
+        assertFalse(getFirst(linguistics.getTransformer()));
     }
+
+    @Test
+    public void requireThatEmptyIsNop() {
+        ExecutionContext ctx = new ExecutionContext(new SimpleTestAdapter());
+        ctx.setLanguage(Language.ENGLISH);
+        var orig = new StringFieldValue("");
+        ctx.setValue(orig);
+        var linguistics = new MyMockLinguistics();
+        assertTrue(getFirst(linguistics.getTransformer()));
+        new NormalizeExpression(linguistics).execute(ctx);
+        FieldValue val = ctx.getValue();
+        assertTrue(val == orig);
+        assertTrue(getFirst(linguistics.getTransformer()));
+    }
+
 }