def setUp(self):
self.ds = [dat.DatasetStructure(p) for p in Path(project_path).children]
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
self.datasetdata = dat.DatasetStructure(self.path)
def pipeline(self):
if hasattr(self, '_pipeline'):
return self._pipeline
datasetdata = self.datasetdata
dsc = datasetdata.cache.dataset
if dsc is None:
raise exc.NotInDatasetError
dataset = dat.DatasetStructure(dsc) # FIXME should be able to go back
class ProtocolHelper(ProtocolData): # FIXME so ... bad ...
@property
def protocol_uris(self,
outer_self=self
): # FIXME this needs to be pipelined
try:
yield from adops.get(outer_self.data,
['meta', 'protocol_url_or_doi'])
except exc.NoSourcePathError:
pass
ph = ProtocolHelper()
# FIXME except for the adds, everything here should be a pipeline
class Lifters: # do this to prevent accidental data leaks
# context
id = dsc.id # in case we are somewhere else
folder_name = dsc.name
uri_api = dsc.uri_api
uri_human = dsc.uri_human
# dataset metadata
#submission = property(lambda s: (_ for _ in dataset.submission))
#dataset_description = property(lambda s: (_ for _ in dataset.dataset_description))
#subjects = property(lambda s: (_ for _ in dataset.subjects))
# protocols
protocol = ph.protocol
#protcur = self.protcur
# aux
organ = self.organ
member = self.member
modality = self.organs_sheet.modality(id)
award_manual = self.organs_sheet.award_manual(id)
#sheets
organ_term = self.organs_sheet.organ_term(id)
class RuntimeContext:
""" utility for logging etc. """
id = dsc.id # TODO can we eliminate the need for lifters this way?
uri_api = dsc.uri_api # FIXME black blackfynn ids their own thing
uri_human = dsc.uri_human # so uri_api and uri_human can be computed from them ...
path = self.path
helpers = {}
helper_key = object()
lifters = Lifters()
#sources = dataset.data
# helper key was bad because it means that pipelines have to
# know the structure of their helper pipelines and maintain them
# across changes, better to use the lifters as other pipelines that
# can start from a single piece of data and return a formatted result
# and track their own provenance, the concept of a helper that holds
# all the prior existing data is not a bad one, it just just that it
# is better to create objects that can take the information already
# in the data for the current pipeline and return an expanded verion
self._pipeline = pipes.PipelineEnd(dataset, lifters, RuntimeContext())
return self._pipeline