mirror of
https://github.com/sartography/cr-connect-workflow.git
synced 2025-02-23 21:28:32 +00:00
Improving the study_info script documentation to provide detailed examples of values returned based on arguments. Making the tests a little more targetted and less subject to breaking through better mocks. Allow all tests to pass even when ther protocol builder mock isn't running locally. Removing the duplication of reference files in tests and static, as this seems silly to me at the moment.
142 lines
5.2 KiB
Python
142 lines
5.2 KiB
Python
import enum
|
|
from typing import cast
|
|
|
|
from marshmallow_enum import EnumField
|
|
from marshmallow_sqlalchemy import SQLAlchemyAutoSchema
|
|
from sqlalchemy import func, Index
|
|
from sqlalchemy.dialects.postgresql import UUID
|
|
|
|
from crc import db
|
|
|
|
|
|
class FileType(enum.Enum):
|
|
bpmn = "bpmm"
|
|
csv = 'csv'
|
|
dmn = "dmn"
|
|
doc = "doc"
|
|
docx = "docx"
|
|
gif = 'gif'
|
|
jpg = 'jpg'
|
|
md = 'md'
|
|
pdf = 'pdf'
|
|
png = 'png'
|
|
ppt = 'ppt'
|
|
pptx = 'pptx'
|
|
rtf = 'rtf'
|
|
svg = "svg"
|
|
svg_xml = "svg+xml"
|
|
txt = 'txt'
|
|
xls = 'xls'
|
|
xlsx = 'xlsx'
|
|
xml = 'xml'
|
|
zip = 'zip'
|
|
|
|
|
|
CONTENT_TYPES = {
|
|
"bpmn": "text/xml",
|
|
"csv": "text/csv",
|
|
"dmn": "text/xml",
|
|
"doc": "application/msword",
|
|
"docx": "application/vnd.openxmlformats-officedocument.wordprocessingml.document",
|
|
"gif": "image/gif",
|
|
"jpg": "image/jpeg",
|
|
"md" : "text/plain",
|
|
"pdf": "application/pdf",
|
|
"png": "image/png",
|
|
"ppt": "application/vnd.ms-powerpoint",
|
|
"pptx": "application/vnd.openxmlformats-officedocument.presentationml.presentation",
|
|
"rtf": "application/rtf",
|
|
"svg": "image/svg+xml",
|
|
"svg_xml": "image/svg+xml",
|
|
"txt": "text/plain",
|
|
"xls": "application/vnd.ms-excel",
|
|
"xlsx": "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet",
|
|
"xml": "application/xml",
|
|
"zip": "application/zip"
|
|
}
|
|
|
|
class FileDataModel(db.Model):
|
|
__tablename__ = 'file_data'
|
|
id = db.Column(db.Integer, primary_key=True)
|
|
md5_hash = db.Column(UUID(as_uuid=True), unique=False, nullable=False)
|
|
data = db.Column(db.LargeBinary)
|
|
version = db.Column(db.Integer, default=0)
|
|
last_updated = db.Column(db.DateTime(timezone=True), default=func.now())
|
|
file_model_id = db.Column(db.Integer, db.ForeignKey('file.id'))
|
|
file_model = db.relationship("FileModel")
|
|
|
|
|
|
class FileModel(db.Model):
|
|
__tablename__ = 'file'
|
|
id = db.Column(db.Integer, primary_key=True)
|
|
name = db.Column(db.String)
|
|
type = db.Column(db.Enum(FileType))
|
|
is_status = db.Column(db.Boolean)
|
|
content_type = db.Column(db.String)
|
|
is_reference = db.Column(db.Boolean, nullable=False, default=False) # A global reference file.
|
|
primary = db.Column(db.Boolean, nullable=False, default=False) # Is this the primary BPMN in a workflow?
|
|
primary_process_id = db.Column(db.String, nullable=True) # An id in the xml of BPMN documents, critical for primary BPMN.
|
|
workflow_spec_id = db.Column(db.String, db.ForeignKey('workflow_spec.id'), nullable=True)
|
|
workflow_id = db.Column(db.Integer, db.ForeignKey('workflow.id'), nullable=True)
|
|
study_id = db.Column(db.Integer, db.ForeignKey('study.id'), nullable=True)
|
|
task_id = db.Column(db.String, nullable=True)
|
|
irb_doc_code = db.Column(db.String, nullable=True) # Code reference to the irb_documents.xlsx reference file.
|
|
form_field_key = db.Column(db.String, nullable=True)
|
|
latest_version = db.Column(db.Integer, default=0)
|
|
|
|
|
|
class FileModelSchema(SQLAlchemyAutoSchema):
|
|
class Meta:
|
|
model = FileModel
|
|
load_instance = True
|
|
include_relationships = True
|
|
include_fk = True # Includes foreign keys
|
|
type = EnumField(FileType)
|
|
|
|
|
|
class LookupFileModel(db.Model):
|
|
"""Takes the content of a file (like a xlsx, or csv file) and creates a key/value
|
|
store that can be used for lookups and searches. This table contains the metadata,
|
|
so we know the version of the file that was used, and what key column, and value column
|
|
were used to generate this lookup table. ie, the same xls file might have multiple
|
|
lookup file models, if different keys and labels are used - or someone decides to
|
|
make a change. We need to handle full text search over the label and value columns,
|
|
and not every column, because we don't know how much information will be in there. """
|
|
__tablename__ = 'lookup_file'
|
|
id = db.Column(db.Integer, primary_key=True)
|
|
label_column = db.Column(db.String)
|
|
value_column = db.Column(db.String)
|
|
file_data_model_id = db.Column(db.Integer, db.ForeignKey('file_data.id'))
|
|
|
|
|
|
class LookupDataModel(db.Model):
|
|
__tablename__ = 'lookup_data'
|
|
id = db.Column(db.Integer, primary_key=True)
|
|
lookup_file_model_id = db.Column(db.Integer, db.ForeignKey('lookup_file.id'))
|
|
lookup_file_model = db.relationship(LookupFileModel)
|
|
value = db.Column(db.String)
|
|
label = db.Column(db.String)
|
|
# In the future, we might allow adding an additional "search" column if we want to search things not in label.
|
|
data = db.Column(db.JSON) # all data for the row is stored in a json structure here, but not searched presently.
|
|
|
|
# Assure there is a searchable index on the label column, so we can get fast results back.
|
|
# query with:
|
|
# search_results = LookupDataModel.query.filter(LookupDataModel.label.match("INTERNAL")).all()
|
|
|
|
__table_args__ = (
|
|
Index(
|
|
'ix_lookupdata_tsv',
|
|
func.to_tsvector('simple', label), # Use simple, not english to keep stop words in place.
|
|
postgresql_using='gin'
|
|
),
|
|
)
|
|
|
|
|
|
class LookupDataSchema(SQLAlchemyAutoSchema):
|
|
class Meta:
|
|
model = LookupDataModel
|
|
load_instance = True
|
|
include_relationships = False
|
|
include_fk = False # Includes foreign keys
|
|
|