- The server part wraps Flask to leverage the type system of python 3
- The client part uses the swagger file generated by the server to create a proxy object
model.py
@dataclass
class TestDataObject:
name: str = field()
id: int = field()server.py
flsk = flask.Flask('test')
app = ServerService(flsk, base='/my_service')
app.register(TestDataObject)
@app.get('/test/<p>')
def test_add_one(p: int=1) -> int:
"""
Example 1 for a unit test
:param p: example of description
:return: return data
"""
return p + 1
@app.get('/test2')
def test_obj(p: TestDataObject) -> TestDataObject:
return p
start_server(flsk, 5000)client.py
builder = ClientService('http://localhost:5000/my_service')
builder.register(TestDataObject)
serv = builder.build()
res = serv.test_add_one(2)
assert(res == 3)
res = serv.test_obj(TestDataObject(name='example', id=1))
assert(res == TestDataObject(name='example', id=1))The client and the server will:
-
cast and check all the parameters to know if
- they have the right type
- they are mandatory
-
handle the content type, complex types use marshmallow:
-
encoding is automatically adapted
-
authenticate if necessary
-
swagger 2 definition is generated and based on the method signature and docstring
-
type annotation is optional
-
pandas dataframe can be used