[translation] consider simplifying input to translation

[kristapratico]

Consult user data available and any findings from user studies to determine if we should simplify our input to create_translation_job to optimize for the use case of only one source and target.

This currently requires you to import and instantiate a few objects

    from azure.ai.translation.document import (
        DocumentTranslationClient,
        DocumentTranslationInput,
        TranslationTarget
    )

    client = DocumentTranslationClient(endpoint, AzureKeyCredential(key))

    job = client.create_translation_job(inputs=[
            DocumentTranslationInput(
                source_url=source_container_url,
                targets=[
                    TranslationTarget(
                        target_url=target_container_url,
                        language_code="es"
                    )
                ]
            )
        ]
    )

Some ideas - we could consider "overloading" the method itself or accepting additional kwargs in DocumentTranslationInput.

1. def create_translation_job(**kwargs):

• requires inputs. Must pass inputs if you need to pass more than one source and/or target.
• requires source_url, target_url, and language_code along with any other optional kwargs

2. DocumentTranslationInput takes kwargs target_url and language_code. Consider other optional kwargs. Must supply targets if planning on passing more than one target in.

[mshaban-msft]

@kristapratico
i don't quite get it!
correct me if i'm wrong ..

so for the 1st option, it should be something like this
create_translation_job(source_url, target_url, language_code, inputs=None)

and for the 2nd option
job = client.create_translation_job(inputs=[ DocumentTranslationInput( source_url=source_container_url, target_url=target_container_url, language_code="es" ) ] )

[kristapratico]

@mohamedshabanofficial These are just ideas... nothing concrete yet.

For the first option - might be easiest to look at this example: https://github.com/Azure/azure-sdk-for-python/blob/master/sdk/metricsadvisor/azure-ai-metricsadvisor/azure/ai/metricsadvisor/_metrics_advisor_client.py#L441

L441 defines the method that the user will call. For typing, we define "overloads" that show the different sets of required params on L391 and L419. In the method we do client side validation to ensure the correct subset of required parameters are passed for either "overload". So it would be like what you wrote, but minus the inputs:

create_translation_job(source_url, target_url, language_code)

or

create_translation_job(inputs)

Second option - yes, that would be the idea. This would align closer with what .NET has too.

[kristapratico]

Option 1 seems like the way to go.

In this case, we can probably do an overload based on number of parameters.
Three parameters, source, target, language.
Single parameter that is list-like (or enumerable), we assume that it is input jobs. (maybe call actions to align with TA?)