Batching in `upload_documents()` does not work

[cecheta]

• Package Name: azure-search-documents
• Package Version: 11.6.0b10
• Operating System: Windows (WSL2)
• Python Version: 3.12.9

Describe the bug
I noticed that in the source code of SearchClient.upload_documents(), more specifically the _index_documents_actions() function, there is meant to be a batch split and retry if the batch is too large and a 413: RequestEntityTooLargeError error is produced. However, the batch splitting doesn't work. Instead, the following error is observed:

Traceback (most recent call last):
  File "/home/user/Code/playground/.venv/lib/python3.12/site-packages/azure/search/documents/_search_client.py", line 703, in _index_documents_actions
    batch_response = self._client.documents.index(batch=batch, error_map=error_map, **kwargs)
                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/user/Code/playground/.venv/lib/python3.12/site-packages/azure/core/tracing/decorator.py", line 105, in wrapper_use_tracer
    return func(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^
  File "/home/user/Code/playground/.venv/lib/python3.12/site-packages/azure/search/documents/_generated/operations/_documents_operations.py", line 1232, in index
    map_error(status_code=response.status_code, response=response, error_map=error_map)
  File "/home/user/Code/playground/.venv/lib/python3.12/site-packages/azure/core/exceptions.py", line 163, in map_error
    raise error
azure.search.documents._search_documents_error.RequestEntityTooLargeError: Operation returned an invalid status 'Request Entity Too Large'
Content: The page was not displayed because the request entity is too large.

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/home/user/Code/playground/main.py", line 37, in <module>
    client.get_search_client("my-index").upload_documents(documents=documents)
  File "/home/user/Code/playground/.venv/lib/python3.12/site-packages/azure/search/documents/_search_client.py", line 596, in upload_documents
    results = self.index_documents(batch, **kwargs)
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/user/Code/playground/.venv/lib/python3.12/site-packages/azure/core/tracing/decorator.py", line 105, in wrapper_use_tracer
    return func(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^
  File "/home/user/Code/playground/.venv/lib/python3.12/site-packages/azure/search/documents/_search_client.py", line 695, in index_documents
    return self._index_documents_actions(actions=batch.actions, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/user/Code/playground/.venv/lib/python3.12/site-packages/azure/search/documents/_search_client.py", line 709, in _index_documents_actions
    batch_response_first_half = self._index_documents_actions(
                                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/user/Code/playground/.venv/lib/python3.12/site-packages/azure/search/documents/_search_client.py", line 703, in _index_documents_actions
    batch_response = self._client.documents.index(batch=batch, error_map=error_map, **kwargs)
                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
TypeError: azure.search.documents._generated.operations._documents_operations.DocumentsOperations.index() got multiple values for keyword argument 'error_map'

To Reproduce
Steps to reproduce the behavior:

1. Create AI Search resource
2. Set the following env vars: AZURE_SEARCH_ENDPOINT, AZURE_SEARCH_API_KEY
3. Run the following code:

import os

from azure.core.credentials import AzureKeyCredential
from azure.search.documents.indexes import SearchIndexClient
from azure.search.documents.indexes.models import (
    SearchFieldDataType,
    SearchIndex,
    SimpleField,
)

client = SearchIndexClient(
    os.getenv("AZURE_SEARCH_ENDPOINT"),
    AzureKeyCredential(os.getenv("AZURE_SEARCH_API_KEY")),
)

index_name = "my-index"

client.create_or_update_index(
    SearchIndex(
        name=index_name,
        fields=[
            SimpleField(name="id", type=SearchFieldDataType.String, key=True),
            SimpleField(name="content", type=SearchFieldDataType.String),
        ],
    )
)

documents = [{"id": str(i), "content": " " * 100000} for i in range(10000)]

client.get_search_client(index_name).upload_documents(documents=documents)

Expected behavior
After the first batch fails, it should split into two smaller batches and retry both.

Additional context
From what I've seen, removing error_map=error_map here and here seems to fix it.

[github-actions]

Hello @cecheta. I'm an AI assistant for the azure-sdk-for-python repository. I have some suggestions that you can try out while the team gets back to you.

• The error is caused by passing the error_map keyword argument twice during recursive batch splitting, which results in a TypeError.
• A potential workaround is to remove error_map from the recursive calls as you noted in the source at line 710 and line 717.
• Consider opening a PR or filing an issue with these observations so that the batch-splitting logic can be updated to correctly handle the error_map parameter during retries.
• Verify if this behavior occurs consistently with large documents triggering a 413: RequestEntityTooLargeError to confirm the fix.

The team will get back to you shortly, hopefully this helps in the meantime.

[cecheta]

Somewhat related, I noticed that if you try to upload more than 32000 documents, that also produces an error of too many documents, but because it's not a 413 RequestEntityTooLargeError, that doesn't get caught.

Perhaps the retry could be applied to that scenario as well?

[xiangyan99]

Could you try the latest version and see if it works?

https://pypi.org/project/azure-search-documents/11.6.0b11/

[github-actions]

Hi @cecheta. Thank you for opening this issue and giving us the opportunity to assist. To help our team better understand your issue and the details of your scenario please provide a response to the question asked above or the information requested above. This will help us more accurately address your issue.

[cecheta]

Hi @xiangyan99 , I just tried again but get this error:

Traceback (most recent call last):
  File "/home/user/Code/playground/app.py", line 30, in <module>
    client.get_search_client(index_name).upload_documents(documents=documents)
  File "/home/user/Code/playground/.venv/lib/python3.12/site-packages/azure/search/documents/_search_client.py", line 596, in upload_documents
    results = self.index_documents(batch, **kwargs)
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/user/Code/playground/.venv/lib/python3.12/site-packages/azure/core/tracing/decorator.py", line 105, in wrapper_use_tracer
    return func(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^
  File "/home/user/Code/playground/.venv/lib/python3.12/site-packages/azure/search/documents/_search_client.py", line 695, in index_documents
    return self._index_documents_actions(actions=batch.actions, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/user/Code/playground/.venv/lib/python3.12/site-packages/azure/search/documents/_search_client.py", line 703, in _index_documents_actions
    batch_response = self._client.documents.index(batch=batch, **kwargs)
                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/user/Code/playground/.venv/lib/python3.12/site-packages/azure/core/tracing/decorator.py", line 105, in wrapper_use_tracer
    return func(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^
  File "/home/user/Code/playground/.venv/lib/python3.12/site-packages/azure/search/documents/_generated/operations/_documents_operations.py", line 1234, in index
    raise HttpResponseError(response=response, model=error)
azure.core.exceptions.HttpResponseError: Operation returned an invalid status 'Request Entity Too Large'
Content: The page was not displayed because the request entity is too large.

It looks like error_map=error_map was removed from the self._client.documents.index() call, when it should have been removed from the two recursive self._index_documents_actions() calls.