shift requirements to dev

[prjemian]

Both pyarrow and pandas were added as package requirements in commit 672f579 as part of PR#301. Neither of these are imported by this package. Shift the both of them to requirements-dev.txt.

Motivation and Context

We ran into an installation dependency in prjemian/model_instrument#5 (comment). On inspection, this seems not to be a real requirement of this package.

[tacaswell]

#308 has a fix for the doc failure.

[tacaswell]

I suspect a few more can go (like mpl, skimage, and databroker).

[prjemian]

databroker is used by worker:

(test) prjemian@arf:~/.../Bluesky/bluesky-queueserver$ git grep databroker | grep import
bluesky_queueserver/manager/tests/test_gen_lists.py:from databroker.v0 import Broker
bluesky_queueserver/manager/tests/test_profile_ops.py:from databroker.v0 import Broker
bluesky_queueserver/manager/tests/test_profile_ops.py:from databroker.v0 import Broker
bluesky_queueserver/manager/worker.py:                            from databroker import Broker

[prjemian]

Let's get these through. Probably enough for the PR that triggered this. Any others can be noted in a new issue/branch/PR.

[dmgav]

For the future, the following code was added at the very start of the project, serves no practical purpose and should be removed. The respective parameters should be removed. The instance of RE is always created in the startup code as well as all subscriptions. Built-in features only confuse beginners who look at the examples.

bluesky-queueserver/bluesky_queueserver/manager/worker.py

Lines 1478 to 1536 in d529fcb

	else:
	# Instantiate a new Run Engine and Data Broker (if needed)
	md = {}
	if self._config_dict["use_persistent_metadata"]:
	# This code is temporarily copied from 'nslsii' before better solution for keeping
	# continuous sequence Run ID is found. TODO: continuous sequence of Run IDs.
	directory = os.path.expanduser("~/.config/bluesky/md")
	os.makedirs(directory, exist_ok=True)
	md = PersistentDict(directory)
	self._RE = RunEngine(md)
	self._re_namespace["RE"] = self._RE
	def factory(name, doc):
	# Documents from each run are routed to an independent
	# instance of BestEffortCallback
	bec = BestEffortCallback()
	if not self._use_ipython_kernel or not ipython_matplotlib:
	bec.disable_plots()
	return [bec], []
	# Subscribe to Best Effort Callback in the way that works with multi-run plans.
	rr = RunRouter([factory])
	self._RE.subscribe(rr)
	# Subscribe RE to databroker if config file name is provided
	self._db = None
	if "databroker" in self._config_dict:
	config_name = self._config_dict["databroker"].get("config", None)
	if config_name:
	logger.info("Subscribing RE to Data Broker using configuration '%s'.", config_name)
	from databroker import Broker
	self._db = Broker.named(config_name)
	self._re_namespace["db"] = self._db
	self._RE.subscribe(self._db.insert)
	if "kafka" in self._config_dict:
	logger.info(
	"Subscribing to Kafka: topic '%s', servers '%s'",
	self._config_dict["kafka"]["topic"],
	self._config_dict["kafka"]["bootstrap"],
	)
	kafka_publisher = kafkaPublisher(
	topic=self._config_dict["kafka"]["topic"],
	bootstrap_servers=self._config_dict["kafka"]["bootstrap"],
	key="kafka-unit-test-key",
	# work with a single broker
	producer_config={"acks": 1, "enable.idempotence": False, "request.timeout.ms": 5000},
	serializer=partial(msgpack.dumps, default=mpn.encode),
	)
	self._RE.subscribe(kafka_publisher)
	if "zmq_data_proxy_addr" in self._config_dict:
	from bluesky.callbacks.zmq import Publisher
	publisher = Publisher(self._config_dict["zmq_data_proxy_addr"])
	self._RE.subscribe(publisher)

[prjemian]

@dmgav , @tacaswell -- Thanks!