fix(classifiers): guard against content=None in DocumentLanguageClassifier (fixes #11418)

haystack/components/classifiers/document_language_classifier.py

@@ -111,6 +111,12 @@ def run(self, documents: list[Document]) -> dict[str, list[Document]]:
 
     def _detect_language(self, document: Document) -> str | None:
         language = None
+        if document.content is None:
+            logger.warning(
+                "Langdetect cannot detect the language of Document with id: {document_id} because its content is None",
+                document_id=document.id,
+            )
+            return language
         try:
             language = langdetect.detect(document.content)
         except langdetect.LangDetectException:

test/components/classifiers/test_document_language_classifier.py

@@ -50,3 +50,33 @@ def test_warning_if_no_language_detected(self, caplog):
             classifier = DocumentLanguageClassifier()
             classifier.run(documents=[Document(content=".")])
             assert "Langdetect cannot detect the language of Document with id" in caplog.text
+
+    def test_content_none_does_not_raise(self):
+        """Regression test for https://github.com/deepset-ai/haystack/issues/11418.
+
+        Documents with content=None (blob-only documents) must not raise TypeError.
+        They should be classified as 'unmatched' and a warning must be emitted.
+        """
+        classifier = DocumentLanguageClassifier()
+        # Should NOT raise TypeError
+        result = classifier.run(documents=[Document(content=None)])
+        assert len(result["documents"]) == 1
+        assert result["documents"][0].meta["language"] == "unmatched"
+
+    def test_content_none_emits_warning(self, caplog):
+        """Regression test: a warning is logged for documents with content=None."""
+        with caplog.at_level(logging.WARNING):
+            classifier = DocumentLanguageClassifier()
+            classifier.run(documents=[Document(content=None)])
+            assert "Langdetect cannot detect the language of Document with id" in caplog.text
+
+    def test_mixed_none_and_text_content(self):
+        """Documents with content=None and normal documents can coexist in the same batch."""
+        classifier = DocumentLanguageClassifier()
+        docs = [
+            Document(content="This is an english sentence."),
+            Document(content=None),
+        ]
+        result = classifier.run(documents=docs)
+        assert result["documents"][0].meta["language"] == "en"
+        assert result["documents"][1].meta["language"] == "unmatched"