fix(python_code_splitter): ensure secondary-split pieces retain qualified name in metadata

Author: GovindhKishore

haystack/components/preprocessors/python_code_splitter.py

@@ -556,11 +556,9 @@ def _secondary_split(self, unit: _CodeUnit, parent_doc: Document) -> list[Docume
             meta["secondary_split"] = True
             meta["secondary_split_index"] = idx
             meta["secondary_split_total"] = len(intermediate)
+            meta["qualified_name"] = qualified_name
+            results.append(Document(content=piece.content or "", meta=meta))
 
-            # Prepend the unit's qualified name as a comment to non-first pieces,
-            # since only the first piece naturally retains the def/class signature line.
-            content = piece.content or "" if idx == 0 else f"# {qualified_name}\n{piece.content or ''}"
-            results.append(Document(content=content, meta=meta))
         return results
 
     @component.output_types(documents=list[Document])

releasenotes/notes/python-code-splitter-secondary-split-context-8449dc47dcce0928.yaml

@@ -1,8 +1,9 @@
 ---
 fixes:
   - |
-    Fixed ``PythonCodeSplitter`` secondary-split pieces losing all identifying
-    context after the first piece when a function, method, or class body is
-    oversized and falls back to line-based splitting. Every piece after the
-    first is now prefixed with the unit's qualified name (e.g. ``# ClassName.method_name``)
-    so that retrieval can still associate the chunk with its source function.
+    Fixed ``PythonCodeSplitter`` losing identifying context for oversized
+    functions, methods, or classes. When a unit is too large and falls back
+    to line-based secondary splitting, only the first resulting piece
+    naturally retains the source ``def``/``class`` line; every piece now
+    includes a ``qualified_name`` field in ``meta`` identifying the function,
+    method, or class it came from.

test/components/preprocessors/test_python_code_splitter.py

@@ -836,6 +836,12 @@ def longer():
         assert chunks_with_short and chunks_with_long
         assert chunks_with_short[0] is not chunks_with_long[0]
 
+    def test_qualified_name_in_meta_for_all_pieces(self, oversized_function_source):
+        splitter = PythonCodeSplitter(min_effective_lines=2, max_effective_lines=5, oversized_factor=3)
+        result = splitter.run(documents=[Document(content=oversized_function_source)])
+        for piece in result["documents"]:
+            assert piece.meta.get("qualified_name") == "giant"
+
 
 class TestEdgeCases:
     def test_module_with_only_docstring(self):
@@ -872,26 +878,3 @@ def test_invalid_syntax_raises_syntax_error(self):
         doc = Document(content="class Broken(\n    pass\n")
         with pytest.raises(SyntaxError):
             splitter.run(documents=[doc])
-
-
-class TestSecondarySplitQualifiedName:
-    def test_first_piece_has_no_prefix(self, oversized_function_source):
-        splitter = PythonCodeSplitter(min_effective_lines=2, max_effective_lines=5, oversized_factor=3)
-        result = splitter.run(documents=[Document(content=oversized_function_source)])
-        secondary = sorted(
-            (c for c in result["documents"] if c.meta.get("secondary_split")),
-            key=lambda c: c.meta["secondary_split_index"],
-        )
-        assert secondary
-        assert not (secondary[0].content or "").startswith("# giant")
-
-    def test_non_first_pieces_are_prefixed_with_qualified_name(self, oversized_function_source):
-        splitter = PythonCodeSplitter(min_effective_lines=2, max_effective_lines=5, oversized_factor=3)
-        result = splitter.run(documents=[Document(content=oversized_function_source)])
-        secondary = sorted(
-            (c for c in result["documents"] if c.meta.get("secondary_split")),
-            key=lambda c: c.meta["secondary_split_index"],
-        )
-        assert len(secondary) >= 2
-        for piece in secondary[1:]:
-            assert (piece.content or "").startswith("# giant")