Add support for positional args in pipeline.get_config()

[tstadel]

Proposed changes:

• support positional args in pipeline.config() (and thus YAML creation)

reader = FARMReader("my-model") # no model_name_or_path="my-model" required anymore
...
pipeline.add_node(name="reader", component=reader, inputs=["retriever"])

# YAML will contain reader's model_name_or_path param
pipeline.save_to_yaml(...)

Status (please check what you already did):

• First draft (up for discussions & feedback)
• Final code
• Added tests

[julian-risch]

@tstadel Is there a specific reason why we would like to support positional arguments? Throughout Haystack we are using named arguments and I think one important reason for that is that our frequent code changes could easily break things otherwise. Users should not rely on that the position of an argument will be the same in Haystack v1.5 etc. It's better to use the name explicitly no?

[tstadel]

@julian-risch generally I agree, that keyword args are to be preferred. But this is more like a coding style topic and we should not enforce it in a strict sense for our users as long as positional args are commonly supported in python. Additionally we already had the problem within DC that not supporting positional args is confusing and we can expect users to expect them to work like in any other python lib.

TL;DR: the current implementation introduces more confusion than value and given that the support is cheap too we should opt for it.

[julian-risch]

LGTM! 👍 Thank you for explaining the motivation behind this PR. Looking into https://github.com/deepset-ai/haystack-hub-api/issues/611#issuecomment-1113275490 made it even easier for me to understand.