Incorrect code generation for API marked as long-running with response type "file"

[KupoCat]

I am trying to generate a python SDK from openapi swagger files and some of my APIs are marked with "x-ms-long-running-operation". For most of the cases this works fine, and the generated code works, but one of the APIs has a response type of "file":

"responses": {
          "200": {
            "description": "...",
            "schema": {
              "type": "file"
            }
          },
...

In this case, the code is generated incorrectly and calling it results in the following error:

deserialized = response.iter_bytes()
                   ^^^^^^^^
NameError: name 'response' is not defined

Looking at the generated code, it seems there is a slight naming problem in the generated deserialization callback:

def get_long_running_output(pipeline_response):
    deserialized = response.iter_bytes()
    if cls:
        return cls(pipeline_response, deserialized, {})  # type: ignore
    return deserialized

I believe in the above code, response should be pipeline_response, and when I manually change this after the generation the code works and method returns the correct value.

[msyyc]

Hi @KupoCat, could you share the swagger link and your command to generated sdk? We need more info to reproduce this issue.

[KupoCat]

Hi, these are our swagger files: https://github.com/microsoft/fabric-rest-api-specs

We are trying to add a new API definition that looks like this:

"/workspaces/{workspaceId}/dataflows/{dataflowId}/executeQuery": {
      "post": {
        "summary": "Executes a query against a dataflow and returns the result.",
        "description": "Executes a specified query against a published dataflow and streams the results in Apache Arrow format. Supports using custom mashup documents for advanced scenarios.\n\nThis API supports [long running operations (LRO)](/rest/api/fabric/articles/long-running-operation).\n\n## Permissions\n\n The caller must have *execute* permissions for the dataflow.\n\n## Required Delegated Scopes\n\nDataflow.Execute.All or Item.Execute.All.\n\n## Microsoft Entra supported identities\n\nThis API supports the Microsoft [identities](/rest/api/fabric/articles/identity-support) listed in this section.\n\n| Identity | Support |\n|-|-|\n| User | Yes |\n| [Service principal](/entra/identity-platform/app-objects-and-service-principals#service-principal-object) and [Managed identities](/entra/identity/managed-identities-azure-resources/overview) | Yes |\n\n## Interface",
        "tags": [
          "QueryExecution"
        ],
        "operationId": "QueryExecution_ExecuteQuery",
        "consumes": [
          "application/json"
        ],
        "produces": [
          "application/json"
        ],
        "x-ms-fabric-sdk-long-running-operation": true,
        "parameters": [
          {
            "in": "path",
            "name": "workspaceId",
            "description": "The workspace ID.",
            "type": "string",
            "format": "uuid",
            "required": true
          },
          {
            "in": "path",
            "name": "dataflowId",
            "description": "The Dataflow ID.",
            "type": "string",
            "format": "uuid",
            "required": true
          },
          {
            "in": "body",
            "name": "executeQueryRequest",
            "description": "Execute query request payload.",
            "schema": {
              "$ref": "./definitions.json#/definitions/ExecuteQueryRequest"
            },
            "required": true
          }
        ],
        "responses": {
          "200": {
            "description": "Query result was succesfully written. Results are streamed in Apache Arrow format. Refer to the [Arrow documentation](https://arrow.apache.org/docs/python/ipc.html#reading-from-stream-and-file-format-for-pandas) on how to read the stream in Python and other languages. \n\nErrors encountered during query execution or during streaming will be in an additional column at the end named 'PQ Arrow Metadata'. \n\nMedia Type: \"application/vnd.apache.arrow.stream\"",
            "schema": {
              "type": "file"
            }
          },
          "202": {
            "description": "Request accepted, query execution in progress.",
            "headers": {
              "Location": {
                "description": "The URL of the operation status, which can be used to track the operation state.",
                "type": "string"
              },
              "x-ms-operation-id": {
                "description": "The operation ID which can be used with long running operations (LRO) APIs to track the operation state and get the result.",
                "type": "string",
                "format": "uuid"
              },
              "Retry-After": {
                "type": "integer",
                "description": "The number of seconds to wait before retrying the request."
              }
            }
          },
          "default": {
            "description": "Common error codes:\n\n* DataflowExecuteQueryError - Query execution failed. Some possible reasons include: the specified query name is invalid or empty, the custom mashup document is invalid, or the specified query name was not found in the dataflow (or in the custom mashup document if provided).",
            "schema": {
              "$ref": "../common/definitions.json#/definitions/ErrorResponse"
            }
          }
        }
      }

To generate the code we run the autorest dataflows.md command, where dataflows.md is the following file:

input-file: "..\\..\\..\\..\\swaggers\\v1\\Dataflow\\swagger.json"
namespace: microsoft.fabric.api.dataflow
package-name: fabric
python: true
models-mode: msrest
output-folder: "..\\..\\..\\..\\SDKs\\python\\src\\microsoft_fabric_api\\generated\\dataflow"
clear-output-folder: true
override-client-name: FabricDataflowClient
add-credential: true
no-namespace-folders: true
credential-scopes: https://api.fabric.microsoft.com/.default
use-extension:
  "@autorest/python": "6.15.0"
  "@autorest/modelerfour": "4.26.0"

directive:
    # https://github.com/Azure/autorest/blob/main/docs/extensions/readme.md#x-ms-long-running-operation
  - reason: Mark LRO
    from: swagger-document
    where: $['paths'][*][*]
    transform: |
      if ($['x-ms-fabric-sdk-long-running-operation']) {
        $['x-ms-long-running-operation'] = true
        delete ['x-ms-fabric-sdk-long-running-operation']
      }

[msyyc]

I could reproduce this error and it is a bug. We will discuss inner team about how to fix it.