Make `EventDescriptor.name` required

[evvaaaa]

We should make stream name required.

It seems that since it was added in the days of yore, name hasn't been required, though the bundler and declare_stream demand it.

We please make this required in the schema too - e.g remove = "" here.

Can you confirm this is alright @danielballan

[prjemian]

Isn't the default to convert "" to "primary"?

I'm a user of yield from bps.create() in custom plans. Such change would break those custom plans?

[evvaaaa]

Since all descriptors are made with the event model compose methods, and this is a positional arg, I don't think it's possible that we'd ever be able to create a descriptor through the run engine which doesn't pass a name directly.

To clarify, I'm not suggesting we change any of the default stream name behavior of bluesky itself, just that the descriptor document has to have a name to be valid, and those defaults have to be given to it before it's validated. This came up because I know someone who was writing a consumer who saw the typing and assumed it wasn't a given that a descriptor would have a name.

In this commit Daniel mentions that it's optional, but I assume this has since changed?

Isn't the default to convert "" to "primary"?

Doesn't seem to be

def plan():
    device = Device()
    yield from bps.open_run()
    yield from bps.create(name="")
    yield from bps.read(device)
    yield from bps.save()
    yield from bps.close_run()


RE(plan())

{'configuration': {'foo': {'data': {}, 'data_keys': {}, 'timestamps': {}}},
  'data_keys': {'foo-bar': {'dtype': 'number',
                            'object_name': 'foo',
                            'shape': [],
                            'source': 'constant'}},
  'hints': {},
  'name': '',
  'object_keys': {'foo': ['foo-bar']},
  'run_start': 'd02679f9-ae0d-4f1f-81c8-1cc6c6b97f73',
  'time': 1769191307.6885548,
  'uid': '748b4b7a-4c84-46d3-bdc9-da0d473fae18'})

I'm a user of yield from bps.create() in custom plans. Such change would break those custom plans?

I don't think so unless I'm missing something?

1

bps.create() with no kwarg defaults to "primary". This wouldn't change that.

We don't have the option to pass either of the following into the run engine
a) Msg("create", name=None) # fails in the runengine and would fail in the current schema b) Msg("create") # fails in the runengine

It allows
c) Msg("create", name="") here.
d) Msg("create", "") here.

2

declare_stream disallows anything other than name="some_string" on the Msg, as defined in the plan stub.

[danielballan]

@evvaaaa's understanding matches mine. The name has been required in the Python API since at least 2016. It was left as optional in the schema because there are some very old documents at NSLS-II (maybe just CSX) that predate it. We didn't want to render those documents invalid in the schema.

I think the impact here is so narrow (and just on NSLS-II) that it takes sense to tighten the schema to match what the Python API required. So, from my point of view @evvaaaa, go ahead.

attn @tacaswell

[evvaaaa]

^ Ready to be merged whenever

[prjemian]

Thanks for clarifying.