fix(http-client-python): synthesize filename in multipart Content-Disposition for bare file inputs

.chronus/changes/fix-multipart-filename-2026-5-29-17-50-0.md

@@ -0,0 +1,7 @@
+---
+changeKind: fix
+packages:
+  - "@typespec/http-client-python"
+---
+
+Synthesize filename in multipart Content-Disposition for bare file inputs. When callers pass bare bytes/str/IO instead of a (filename, content) tuple for multipart file fields, the `prepare_multipart_form_data` helper now wraps them with a synthesized filename so servers that require `filename=` in the Content-Disposition header no longer reject the upload.

packages/http-client-python/generator/pygen/codegen/serializers/general_serializer.py

@@ -242,6 +242,7 @@ def need_utils_utils_file(self) -> str:
                 ImportType.LOCAL,
             )
             file_import.add_import("json", ImportType.STDLIB)
+            file_import.add_import("os", ImportType.STDLIB)
 
         return template.render(
             code_model=self.code_model,

packages/http-client-python/generator/pygen/codegen/templates/utils.py.jinja2

@@ -76,23 +76,60 @@ def serialize_multipart_data_entry(data_entry: Any) -> Any:
         return json.dumps(data_entry, cls=SdkJSONEncoder, exclude_readonly=True)
     return data_entry
 
+def _normalize_multipart_file_entry(field_name: str, entry: Any, index: int) -> Any:
+    """Ensure a multipart file entry carries a filename for Content-Disposition.
+
+    Servers distinguish file parts from plain form fields by the presence of
+    ``filename=`` in the ``Content-Disposition`` header.  When callers pass
+    bare bytes/str/IO the HTTP client omits the filename and the server may
+    reject the upload.  This helper wraps bare values into a (filename, content)
+    tuple, deriving the name from IO.name when available.
+
+    :param str field_name: The multipart field name used as a filename fallback.
+    :param entry: The user-provided file entry (tuple, bytes, str, or IO).
+    :type entry: any
+    :param int index: The positional index of the entry within the field, used
+        to disambiguate fallback filenames when multiple entries are provided.
+    :return: Either the original tuple entry, or a ``(filename, content)`` tuple
+        wrapping the bare value.
+    :rtype: any
+    """
+    if isinstance(entry, tuple):
+        return entry
+    filename: Optional[str] = None
+    name_attr = getattr(entry, "name", None)
+    if isinstance(name_attr, str) and name_attr:
+        filename = os.path.basename(name_attr)
+    if not filename:
+        filename = f"{field_name}_{index}" if index else field_name
+
+    # Return a 3-tuple with an explicit "application/octet-stream" content type.
+    # A 2-tuple (filename, content) would leave the part's Content-Type unset, and
+    # the sdk core library only defaults to "application/octet-stream" for bare
+    # (non-tuple) values - a tuple bypasses that default and falls back to the
+    # HTTP "text/plain" default instead. Setting it explicitly preserves the
+    # pre-existing behavior for bare bytes/IO across all transports.
+    return (filename, entry, "application/octet-stream")
+
 def prepare_multipart_form_data(
     body: Mapping[str, Any], multipart_fields: list[str], data_fields: list[str]
 ) -> list[FileType]:
     files: list[FileType] = []
-    for multipart_field in multipart_fields:
-        multipart_entry = body.get(multipart_field)
-        if isinstance(multipart_entry, list):
-            files.extend([(multipart_field, e) for e in multipart_entry ])
-        elif multipart_entry:
-            files.append((multipart_field, multipart_entry))
 
-    # if files is empty, sdk core library can't handle multipart/form-data correctly, so
-    # we put data fields into files with filename as None to avoid that scenario.
+    # Data fields first so streaming server-side parsers see metadata before
+    # binary file parts.
     for data_field in data_fields:
         data_entry = body.get(data_field)
         if data_entry:
             files.append((data_field, str(serialize_multipart_data_entry(data_entry))))
 
+    for multipart_field in multipart_fields:
+        multipart_entry = body.get(multipart_field)
+        if isinstance(multipart_entry, list):
+            for idx, e in enumerate(multipart_entry):
+                files.append((multipart_field, _normalize_multipart_file_entry(multipart_field, e, idx)))
+        elif multipart_entry is not None:
+            files.append((multipart_field, _normalize_multipart_file_entry(multipart_field, multipart_entry, 0)))
+
     return files
 {% endif %}

packages/http-client-python/tests/mock_api/shared/test_multipart_filetype_variants.py

@@ -0,0 +1,151 @@
+# -------------------------------------------------------------------------
+# Copyright (c) Microsoft Corporation. All rights reserved.
+# Licensed under the MIT License. See License.txt in the project root for
+# license information.
+# --------------------------------------------------------------------------
+"""Offline unit tests for ``prepare_multipart_form_data``.
+
+Verify that every concrete variant of the ``FileType`` union produces a
+multipart-equivalent normalized entry — i.e. the same field name, filename,
+and content payload.  These tests run entirely offline (no network, no mock
+server) and operate directly on the generated helper.
+"""
+
+import io
+from pathlib import Path
+
+import pytest
+
+from payload.multipart._utils.utils import prepare_multipart_form_data
+
+FILENAME = "image.jpg"
+CONTENT = b"\xff\xd8\xff\xe0 fake jpeg"
+FIELD = "profileImage"
+
+
+def _read(value):
+    """Return raw bytes regardless of whether *value* is bytes or IO."""
+    if hasattr(value, "read"):
+        try:
+            value.seek(0)
+        except Exception:  # pylint: disable=broad-except
+            pass
+        return value.read()
+    return value
+
+
+def _canonicalize(prepared, field=FIELD):
+    """Extract the first entry for *field* as (field, filename, bytes)."""
+    for f, entry in prepared:
+        if f == field:
+            assert isinstance(entry, tuple), f"helper must wrap entry as a tuple, got {entry!r}"
+            filename = entry[0]
+            content = _read(entry[1])
+            return (f, filename, content)
+    raise AssertionError(f"field {field!r} not found in {prepared!r}")
+
+
+# ── Variant helpers ──────────────────────────────────────────────────────
+
+
+def _io_from_disk(tmp_path):
+    p = tmp_path / FILENAME
+    p.write_bytes(CONTENT)
+    return p.open("rb")
+
+
+# ── Tests ────────────────────────────────────────────────────────────────
+
+
+class TestNormalizeBareInputs:
+    """Bare bytes / IO must be wrapped with a synthesized filename."""
+
+    def test_bare_io_gets_filename_from_name_attr(self, tmp_path):
+        """IO objects with a .name attribute use basename as filename."""
+        body = {FIELD: _io_from_disk(tmp_path)}
+        result = prepare_multipart_form_data(body, [FIELD], [])
+        field, filename, content = _canonicalize(result)
+        assert field == FIELD
+        assert filename == FILENAME
+        assert content == CONTENT
+
+    def test_bare_bytes_gets_field_name_as_filename(self):
+        """Bare bytes without .name fall back to the field name."""
+        body = {FIELD: CONTENT}
+        result = prepare_multipart_form_data(body, [FIELD], [])
+        field, filename, content = _canonicalize(result)
+        assert field == FIELD
+        assert filename == FIELD  # fallback
+        assert content == CONTENT
+
+    def test_bare_bytes_io_gets_field_name_as_filename(self):
+        """BytesIO without .name falls back to the field name."""
+        body = {FIELD: io.BytesIO(CONTENT)}
+        result = prepare_multipart_form_data(body, [FIELD], [])
+        field, filename, content = _canonicalize(result)
+        assert field == FIELD
+        assert filename == FIELD  # BytesIO.name is not a real path
+        assert content == CONTENT
+
+
+class TestTuplePassthrough:
+    """Tuple variants of FileType must pass through unchanged."""
+
+    def test_two_tuple(self):
+        body = {FIELD: (FILENAME, CONTENT)}
+        result = prepare_multipart_form_data(body, [FIELD], [])
+        _, entry = result[0]
+        assert entry == (FILENAME, CONTENT)
+
+    def test_three_tuple(self):
+        body = {FIELD: (FILENAME, CONTENT, "image/jpeg")}
+        result = prepare_multipart_form_data(body, [FIELD], [])
+        _, entry = result[0]
+        assert entry == (FILENAME, CONTENT, "image/jpeg")
+
+
+class TestListEntries:
+    """List-valued file fields normalize each element independently."""
+
+    def test_list_of_bare_bytes(self):
+        body = {FIELD: [b"file0", b"file1"]}
+        result = prepare_multipart_form_data(body, [FIELD], [])
+        assert len(result) == 2
+        _, entry0 = result[0]
+        _, entry1 = result[1]
+        # index 0 → field name (no suffix), index 1+ → field_N
+        assert entry0[0] == FIELD
+        assert entry1[0] == f"{FIELD}_1"
+
+    def test_list_of_tuples(self):
+        body = {FIELD: [("a.jpg", b"a"), ("b.jpg", b"b")]}
+        result = prepare_multipart_form_data(body, [FIELD], [])
+        assert len(result) == 2
+        _, entry0 = result[0]
+        _, entry1 = result[1]
+        assert entry0 == ("a.jpg", b"a")
+        assert entry1 == ("b.jpg", b"b")
+
+
+class TestDataFieldOrdering:
+    """Data fields must appear before file fields."""
+
+    def test_data_precedes_files(self, tmp_path):
+        body = {"id": "123", FIELD: _io_from_disk(tmp_path)}
+        result = prepare_multipart_form_data(body, [FIELD], ["id"])
+        fields = [f for f, _ in result]
+        assert fields == ["id", FIELD]
+
+
+class TestEdgeCases:
+    """Edge cases: None values, empty content."""
+
+    def test_none_value_skipped(self):
+        body = {FIELD: None}
+        result = prepare_multipart_form_data(body, [FIELD], [])
+        assert len(result) == 0
+
+    def test_missing_field_skipped(self):
+        body = {}
+        result = prepare_multipart_form_data(body, [FIELD], [])
+        assert len(result) == 0