cr-connect-workflow/example_data.py

115 lines
4.7 KiB
Python

import datetime
import glob
import os
from crc import app, db, session
from crc.models.file import FileType, FileModel, FileDataModel
from crc.models.study import StudyModel
from crc.models.workflow import WorkflowSpecModel
class ExampleDataLoader:
def make_data(self):
studies = [
StudyModel(
id=1,
title='The impact of fried pickles on beer consumption in bipedal software developers.',
last_updated=datetime.datetime.now(),
protocol_builder_status='in_process',
primary_investigator_id='dhf8r',
sponsor='Sartography Pharmaceuticals',
ind_number='1234'
),
StudyModel(
id=2,
title='Requirement of hippocampal neurogenesis for the behavioral effects of soft pretzels',
last_updated=datetime.datetime.now(),
protocol_builder_status='in_process',
primary_investigator_id='dhf8r',
sponsor='Makerspace & Co.',
ind_number='5678'
),
]
workflow_specifications = \
self.create_spec(id="random_fact",
name="random_fact",
display_name="Random Fact Generator",
description='Displays a random fact about a topic of your choosing.')
workflow_specifications += \
self.create_spec(id="two_forms",
name="two_forms",
display_name="Two dump questions on two separate tasks",
description='the name says it all')
workflow_specifications += \
self.create_spec(id="decision_table",
name="decision_table",
display_name="Form with Decision Table",
description='the name says it all')
workflow_specifications += \
self.create_spec(id="parallel_tasks",
name="parallel_tasks",
display_name="Parallel tasks",
description='Four tasks that can happen simultaneously')
# workflow_specifications += \
# self.create_spec(id="docx",
# name="docx",
# display_name="Form with document generation",
# description='the name says it all')
all_data = studies + workflow_specifications
return all_data
def create_spec(self, id, name, display_name, description):
"""Assumes that a directory exists in static/bpmn with the same name as the given id.
further assumes that the [id].bpmn is the primary file for the workflow.
returns an array of data models to be added to the database."""
models = []
spec = WorkflowSpecModel(id=id,
name=name,
display_name=display_name,
description=description)
models.append(spec)
filepath = os.path.join(app.root_path, 'static', 'bpmn', id, "*")
files = glob.glob(filepath)
for file_path in files:
noise, file_extension = os.path.splitext(file_path)
filename = os.path.basename(file_path)
if file_extension.lower() == '.bpmn':
type = FileType.bpmn
elif file_extension.lower() == '.dmn':
type = FileType.dmn
elif file_extension.lower() == '.svg':
type = FileType.svg
# elif file_extension.lower() == '.docx':
# type = FileType.docx
else:
raise Exception("Unsupported file type:" + file_path)
continue
is_primary = filename.lower() == id + ".bpmn"
file_model = FileModel(name=filename, type=type, content_type='text/xml', version="1",
last_updated=datetime.datetime.now(), primary=is_primary,
workflow_spec_id=id)
models.append(file_model)
try:
file = open(file_path, "rb")
models.append(FileDataModel(data=file.read(), file_model=file_model))
finally:
file.close()
return models
@staticmethod
def clean_db():
session.flush() # Clear out any transactions before deleting it all to avoid spurious errors.
for table in reversed(db.metadata.sorted_tables):
session.execute(table.delete())
session.flush()
def load_all(self):
session.add_all(self.make_data())
session.commit()
session.flush()