DrawUML Backend

Endpoints

version 1

GET /v1/files/{filename} to get a specific filename

GET /v1/files to get all the filenames in the output folder

GET /v1/servers to get the list of the server files in the server output

GET /v1/servers/{filename} to get a specific file in the server output folder

POST /v1/servers upload an SQL file made from draw SQL to create SQLALchemy classes

POST /v1/files/existing upload existing code and improve commenting or formatting

POST /v1/files/new upload a response in the Domain Specific Language of drawUML to create test, classes and documentation

Design Overview

The software can be divided in this 3 main layers The dataprocessing layer processes the classes which arrive from the front end one by one the file generation layer generates the files such as the diagram file and the _types file this can be organised by routines such that makes the design more modular and easily testable

The api endpoints can foolow the following convention /method/version/file_type/file_type_metadata for instance: /create/new/diagram get/existing/python/code create/new/python/type

in the source code the response_code_path refers to the json generated from the existing source code new_response code path is the json generated from the UI and this should be converted into the response_code_path into the common format

Note

for generating class diagrams you can also run

pyreverse -o png -p draw_uml_backend src/draw_uml_backend

make sure that you have Graphviz installed if you don't run

sudo apt-get install graphviz

Using the drawSQL plugin

for automating the generation of SQLAlchemy classes from drawSQL we assume that relationships are built into the tables i.e

therefore we want to keep tables in singular i.e. User and in camel notation such as UnitOperation, have foreign keys map to ids of other tables and to be named <table_reference>_id and to set columns which map to other tables i.e. if I want to pass an owner which is a user object to an Item class

item = Item(owner=User(name="Paul"))

then we can name the column with a leading slash i.e. owner_. This will map to generating a relationship on the class which are used to pass objects rather than values

Example of using the database models generated through the drawSQL function

Session = sessionmaker(bind=create_engine(
    f"sqlite:///__main__.sqlite3", echo=True))

session = Session()
names = [
    {"name": "Mark"},
    {"name": "Luci"},
    {"name": "Julia"}
]

users = [User(**name) for name in names]
items = [Items(user) for user in users]

session.add_all(users)
session.add_all(items)
session.commit()

users: List[User] = session.query(User).all()
print([user.name for user in users])
for user in users:
    print(user.items)

For modifying the schema.py generated file add a property that is unique within the table on the TableCreate object of the schema. Add a property on the constructor for all the rows which are relationships in tables which have a foreignkey, this will allow to relate the table to at least one of its foreign table as the argument will be required tio be passed.

Tests

the only test is the integration test test_routines.py however a better way to test the code is to run the gunicorn instance locally, and then either copy and paste the response of the diagram from the frontend or copy the code that you want to paste as existing code to check the files being downloaded you can do it directly on the browser by opening the file on a seperate window to run automatic tests

to run manual tests

python -m pytest tests