Let Container.write_dataframe accept a dict

[danielballan]

It is convenient that Container.write_array accepts a list, like:

$ tiled serve catalog --temp --api-key=secret

In [1]: from tiled.client import from_uri

In [2]: c = from_uri('http://localhost:8000', api_key='secret')

In [3]: c
Out[3]: <Container {}>

In [4]: c.write_array([1,2,3], key='x')
Out[4]: <ArrayClient shape=(3,) chunks=((3,),) dtype=int64>

It would be convenient if

c.write_dataframe({'a': [1,2,3], 'b':, [4,5,6]}, key='y')

were likewise acceptable. Currently, a pandas.DataFrame or dask.dataframe object must be passed.

[danielballan]

This needs two additions:

1. Add a branch to this conditional that checks for isinstance(dataframe, dict) and, if so, attempts dataframe = pandas.DataFrame(dataframe).

tiled/tiled/client/container.py

Lines 943 to 946 in 07c8925

	if isinstance(dataframe, dask.dataframe.DataFrame):
	structure = TableStructure.from_dask_dataframe(dataframe)
	else:
	structure = TableStructure.from_pandas(dataframe)

2. Add a variation on this test that passes the dict data to write_dataframe instead of the pandas.DataFrame df.

tiled/tiled/_tests/test_writing.py

Lines 125 to 147 in 07c8925

	def test_write_dataframe_full(tree):
	with Context.from_app(
	build_app(tree, validation_registry=validation_registry)
	) as context:
	client = from_context(context)
	data = {f"Column{i}": (1 + i) * numpy.ones(5) for i in range(5)}
	df = pandas.DataFrame(data)
	metadata = {"scan_id": 1, "method": "A"}
	specs = [Spec("SomeSpec")]
	with record_history() as history:
	client.write_dataframe(df, metadata=metadata, specs=specs)
	# one request for metadata, one for data
	assert len(history.requests) == 1 + 1
	results = client.search(Key("scan_id") == 1)
	result = results.values().first()
	result_dataframe = result.read()
	pandas.testing.assert_frame_equal(result_dataframe, df)
	assert result.metadata == metadata
	assert result.specs == specs