[BUG] Does saveToOpenSearch supports index mappings?

[pjpastrana]

What is the bug?

I have created an index with its specific settings and mappings for the data fields.
When I try to bulk insert my dataframe calling saveToOpenSearch, I get the following error:
Could not write all entries for bulk operation [1000/1000]. Error sample (first [5] error messages): org.opensearch.hadoop.rest.OpenSearchHadoopRemoteException: illegal_argument_exception: mapper [field.x] cannot be changed from type [keyword] to [text]...

Most likely because saveToOpenSearch is assuming a dynamic mapping and not the defined mapping from the index.

I can't find any reference in the docs of saveToOpenSearch supporting the mappings definitions for the index as an argument. Is this a bug? or am I missing something here?
I don't get this behavior when the index has been created without any mappings or if it does not exists and is automatically created by the saveToOpenSearch function call.

How can one reproduce the bug?

Create the index

index_body = {
  "settings": {
    "index": index_settings
  },
  "mappings": mappings
}
client = OpenSearch(
    hosts = [{'host': host, 'port': port}],
    use_ssl = True,
    verify_certs = True
)
client.indices.create(
  index_name, 
  body=index_body
)

Push data to OpenSearch

val df = spark.sql(s"select * from table")
df.saveToOpenSearch(index_name)

What is the expected behavior?

I expect the records to be added to the index similarly to saveToOpenSearch function automatically creates the index.

What is your host/environment?

Operating system, version.
Scala 2.12
Spark 3.4.1
OpenSearch 1.3.2
opensearch-hadoop:1.1.0

[dblock]

[Catch All Triage - 1, 2, 3, 4]