canopy 9.3

Python Client for the Canopy Simulation API

Installation

Versioning

This library uses SimVer versioning, where a change in the major version number indicates a breaking change and a change in the minor version number indicates a non-breaking change (such as an additional feature or bug fix).

Changelog

The changelog is available here.

Requirements.

This library has been tested on Python 3.6 and higher.

pip install

pip install canopy

You may need to run pip with root permission: sudo pip install canopy.

From a Jupyter Notebook you can run !pip install canopy.

Install from source

Install from source using pip:

pip install .

You may need to run pip with root permission: sudo pip install .

Running Tests

Unit tests can be run with:

pytest canopy

Integration tests can be run with:

pytest integration_tests

To run the integration tests you'll need to ensure you have an environment variable called CANOPY_PYTHON_INTEGRATION_TEST_CREDENTIALS containing the string <client_id>|<client_secret>|<username>|<tenant_name>|<password>.

Getting Started

Example Usage

See the Canopy Python Examples repository for example usage.

Introduction

This package is designed for customers of Canopy Simulations who would like to access the Canopy API from Python, for example using Jupyter Notebooks.

Currently the library is split into two parts:

• The client generated using the OpenAPI toolset is located in the "canopy/openapi" folder. We don't have a great deal of control over how this code looks, but it should give a fairly complete interface to the main API.

• One folder up from that in the "canopy" folder we are adding helper functions which wrap common use cases in simple functions. You can also use these functions as a reference to using the OpenAPI generated code.

When using the library you generally start by creating a canopy.Session object. The session object manages authentication, and the caching of user settings. Calling session.authentication.authenticate() before calling OpenAPI generated client functions ensures that you are authenticated and that any expired access tokens are refreshed. Our helper functions will handle calling authenticate before making any calls, so if you are only using our helper functions you won't need to call it yourself.

The session should generally be created once per application. It will automatically dispose itself when the application shuts down. Alternatively you can enclose it in an async with or a with block if you need to create multiple sessions, as shown in the examples below.

If you are using the OpenAPI generated code then you can pass the session.async_client or session.sync_client into the OpenAPI generated API client instance as the api_client parameter as shown below. Passing in async_client will cause it to use asyncio, and you will need to await the calls. Passing in sync_client will cause the calls to complete synchronously.

Our helper functions all use asyncio for efficient parallelisation of downloads, and must therefore be awaited.

The following example shows how to create a session and request some output channels from a study using our helper function:

import canopy
import asyncio

async with canopy.Session(client_id='<your_client_id>', username='<your_username>') as session:
    study_data = await canopy.load_study(session, '<study_id>', 'DynamicLap', ['sRun', 'vCar'])

    # Using the OpenAPI generated client directly:
    study_api = canopy.openapi.StudyApi(session.async_client)
    job_result = await study_api.study_get_study_job_metadata(
        session.authentication.tenant_id,
        '<study_id>',
        0)

    # Using asyncio.ensure_future() to enable us to perform multiple calls in parallel
    job_result_task = asyncio.ensure_future(study_api.study_get_study_job_metadata(
        session.authentication.tenant_id,
        '<study_id>',
        0))

    job_result_2 = await job_result_task

When running this code you will be prompted for your client secret and your password if it is the first time session.authentication.authenticate() has been called for this session instance. Alternatively you can pass the client secret and password into the Session class (after fetching them from a secure location) to avoid being prompted.

If you can't use asyncio and async/await you can instead instantiate the session object synchronously and use the canopy.run method when calling our async helper methods. You can pass session.sync_client into the OpenAPI client classes instead of session.async_client to make them return results synchronously.

import canopy

with canopy.Session(client_id='<your_client_id>', username='<your_username>') as session:
    # Note we are using canopy.run(..) to force the async method to run synchronously.
    # This is a wrapper for asyncio.get_event_loop().run_until_complete(..).
    study_data = canopy.run(canopy.load_study(session, '<study_id>', 'DynamicLap', ['sRun', 'vCar']))

    # Using the OpenAPI generated client synchronously by passing in sync_client:
    study_api = canopy.openapi.StudyApi(session.sync_client)
    job_result = study_api.study_get_study_job_metadata(
        session.authentication.tenant_id,
        '<study_id>',
        0)

    # You can still run synchronous OpenAPI client methods asynchronously using threads if you need to:
    job_result_thread = study_api.study_get_study_job_metadata(
        session.authentication.tenant_id,
        '<study_id>',
        0,
        async_req=True)

    job_result_2 = job_result_thread.get()

Proxy Servers

You can configure your proxy server by passing in a proxy argument to the canopy.Session object:

async with canopy.Session(authentication_data, proxy=canopy.ProxyConfiguration('http://some.proxy.com', 'user', 'pass')) as session:

Updating the OpenAPI Client

This needs to be tidied up, improved, and automated.

Additional options can be found here: https://openapi-generator.tech/docs/generators/openapi/

• e.g. enumUnknownDefaultCase could be useful if the remaining exposed enums change in future.

You can use the Dockerfile in this repository to create a docker image to generate the new API stubs.

You can open this project from VSCode running in windows or in a container, but keep this source in a windows share
(copying from a container into a WSL share is problematic and didn't work for me)

This process defaults to using the production api as the source of the client. If you wish to run against a local build there are extra steps to follow, see below

step by step

1. Create the docker image to host the java runtime

docker image build -t canopy-python-gen:1 .

2. open a session in the new container and bind it to the source folder

docker container run -i -t --mount type=bind,src='<path>/<to>/canopy/canopy-python',dst=/canopy/repo canopy-python-gen:1 /bin/bash

if the source is in C:\users\username\source\Canopy-Python the command would be:

docker container run -i -t --mount type=bind,src='C:\users\username\source\Canopy-Python',dst=/canopy/repo canopy-python-gen:1 /bin/bash

3. run the script to generate the client and copy it into the source folder:

./generate_client.sh

4. check that the url is correct and if so type y:

Generated URL:
  https://api.canopysimulations.com/swagger/v1/swagger.json

Proceed? (y/n):

Note: The openapi/configuration.py file will need to be manually modified to add the default API host URL.
Note: The openapi_asyncio/rest.py file will need to be manually modified to support proxy servers after generation.
Note: The openapi_asyncio/api_client.py and openapi/api_client.py files will need to be manually modified to support numpy array serialization after generation.
Note: The availability_api.py, membership_api.py and study_api.py files will need reverting to specify 'Bearer' in AuthSettings

Using a local version of the API as source

I had difficulty connecting to the local version of the api when it was served under https. While users can choose to ignore the risk when running in a browser that is not the case when connecting from the docker container. If found it was simpler to modify the API project to serve via http by making the following changes:

1. edit .devcontainer/devcontainer.json and replace the forwarded ports:

	"forwardPorts": [23911]

2. edit Canopy.Api.App/Properties/launchSettings.json:

	"sslPort": 0  
    applicationUrl: "http://localhost:23911",  

3. edit Canopy.Api.App/appsettings.json:

  "WEBSITE_BASE_URL": "http://localhost:4200/",  
  "API_BASE_URL": "http://localhost:23911/",

4. edit Canopy.Identity.App/appsettings.json:

  "WEBSITE_BASE_URL": "http://localhost:4200/",  
  "API_BASE_URL": "http://localhost:23911/",

5. edit docker-compose.yml

      - ASPNETCORE_URLS=http://localhost:23911

If you then build and deploy the api it shoud be accessible on http://localhost:23911/swagger/index.html

To use that endpoint when generating the client:

./generate_client.sh http://host.docker.internal:23911

If you know an easier way to sidestep this issue, for example making the client generator ignore certificate errors, then please update this document!

Documentation for OpenAPI Generated Client

OpenAPI generated documentation can be found here.