github.com/France-ioi/AlgoreaBackend/v2

Backend for the new Algorea platform


License
MIT
Install
go get github.com/France-ioi/AlgoreaBackend/v2

Documentation

Algorea Backend

Running the app (for development)

Compile the app:

make

You can then run the app: (call ./bin/AlgoreaBackend to print the list of available commands)

./bin/AlgoreaBackend <cmd> <opt>

For instance, you can launch the web server using ./bin/AlgoreaBackend serve.

Running the setup (as API consumer)

The easiest way to run the backend for consumer it is to run it in a container with its database. To do that:

  • clone this repository (or download the docker-compose.yml file and replace build: . by image: franceioi/algoreabackend:latest to use the public image)

  • Seed the database:

    docker-compose run backend /bin/sh -c "sleep 1; ALGOREA_DATABASE__USER="root"; ALGOREA_DATABASE__PASSWD="a_root_db_password"; AlgoreaBackend db-restore && AlgoreaBackend db-migrate && AlgoreaBackend install"

  • Launch the docker compose setup (db+backend): docker-compose up

  • Visit http://127.0.0.1:8080/status with your browser, you should get a success status message.

If needed, you can connect on the MySQL CLI using:

docker exec -it algoreabackend_db_1 mysql -h localhost -u algorea -pa_db_password  --protocol=TCP algorea_db

Running the setup (as dev)

The application needs a database (MySQL) to run and requires it for a major part of its tests.

It is recommended to set

  • innodb_lock_wait_timeout=1,
  • max-allowed-packet=10485760

and provide at least 2Gb of memory to the MySQL server.

To make testing and development easier, a docker-compose file declares a database using the default configuration. Launch docker-compose up to run tests without any configuration efforts.

Configuration

The app configuration stands in the conf/config.yml file. The file conf/config.sample.yml is a sample configuration to start from, it is configured to work with the docker-compose configuration for local development. All configuration parameter can be also defined using environment variables (with an higher priority), see .circleci/config.yml for examples.

Environment-specific configurations can be defined using conf/config.ENV.yml files when ENV can be "prod", "dev" or "test.

Configuration of test environment

The test environment is used for running the tests. For the test environment, we don't fall back to the default configuration file, so you need to provide a conf/config.test.yml file. This is to avoid running tests on a production database by mistake and erasing data.

Creating the keys

openssl genrsa --out private_key.pem 4096
openssl rsa -in private_key.pem -pubout -out public_key.pem

Seeding the database

An empty dump (schema without data) can be loaded using the

./bin/AlgoreaBackend db-restore

followed by

./bin/AlgoreaBackend db-migrate

Probably you may also want to run

./bin/AlgoreaBackend install

to insert the data required by the config.

Also, after changing the DB data manually, you probably want to run

./bin/AlgoreaBackend db-recompute

in order to recompute DB caches.

Testing

make test

To execute all tests (unit and bdd) with race detection and collect the test code coverage you can run:

make test

This mode is the slowest one, it doesn't use the cache, and it always runs all tests. It is useful to run before pushing code to the repository.

make test-dev

To get test results faster during development, you may want to run all tests without race detection and without collecting the test code coverage to get advantage of Golang per-package caching:

make test-dev

make test-unit

You may want to run only unit tests that are marked with the "unit" tag:

make test-unit

make test-unit doesn't cache test results, all the matching tests will be run. For unit tests not marked as "unit", use make test or make test-dev instead.

make test-bdd

It is possible to run only Gherkin tests defined in *.feature files:

make test-bdd

or if you want only to run bdd tests with a specific tag, in a specific directory. Specifying the directory is mandatory when using tags.

DIRECTORY=./app/api/answers/ TAGS=wip make test-bdd

To add a tag to a test, just precede it by @wip on the line above it in the *.feature file. This is useful to only execute appropriate tests.

make test-bdd doesn't cache test results, all the matching tests will be run.

Tests filtering

For all make test* it is possible to filter with a certain directory and the name of the test function you want to run:

make DIRECTORY=./app/database FILTER=TestItemStore_TriggerBeforeInsert_SetsPlatformID test

Note that FILTER is not currently supported for make test-bdd.

Install the git hooks

Copy githooks/pre-commit to .git/hooks/pre-commit. You may want to adapt the content in case you have personal hooks.

Style

A .editorconfig file defines the basic editor style configuration to use. Check the "editorconfig" support for your favorite editor if it is not installed by default.

For the Go coding styles, we use the standard linters (many). You can install and run them with:

make lint

Code formatting

We use gofumpt to format the code. It is stricter than gofmt.

Generating the API documentation

The API documentation is an OpenAPI spec file generated automatically by Go-Swagger and swagger2openapi. Each time the code is pushed in the "master" branch, the CI generates the spec file and deploys it to the documentation server.

To perform the spec generation locally, install our patched version of Go-Swagger from source (requires Go 1.21+):

go install github.com/France-ioi/go-swagger/cmd/swagger@00200fa

and swagger2openapi:

npm install -g swagger2openapi

Also, you need to install redocly-cli to serve the documentation locally:

npm install -g @redocly/cli@1.25.15

After everything is installed, you can generate the specification file from code and validate it:

make swagger-generate

To view the documentation in a browser, start serving it:

make swagger-serve

and open http://127.0.0.1:8080 in your favorite browser.

Create a release

In order to create a release:

  • decide of a new version number (using semver)
  • update the changelog (add a new section, with the date of today and listing the fix and new features)
  • commit this change as a commit "Release vx.y.z"
  • tag the current commit "vx.y.z" (git tag -a -m "Release vx.y.z" vx.y.z)
  • push everything (git push origin master; git push origin vx.y.z)
  • the rest (github release, doc generation and deployment) is done by the CI

Software Walkthrough

Routing a request

  • The web app is defined in app.go which loads all the middlewares and routes. The routing part consists in mounting the API on / and giving a context to it (i.e., the app database connection)
  • The API routing (app/api/api.go) does the same for mounting all group of services.
  • A service group (e.g., app/api/groups/groups.go.) mounts all its services and pass the context again.
  • Each service has its dedicated file (e.g., app/api/groups/get-all.go). We try to separate the actual HTTP request parsing and response generation from the actual business logic and the call to the database.

How to profile a service

  1. Start the server in 'dev' environment
./bin/AlgoreaBackend serve dev
2019/07/10 00:15:39 Loading environment: test
INFO 2019/07/10 00:15:39 Starting application: environment = dev
INFO 2019/07/10 00:15:39 Loading environment: dev
INFO 2019/07/10 00:15:39 Configuring server...
INFO 2019/07/10 00:15:39 Starting server...
INFO 2019/07/10 00:15:39 Listening on :8080
  1. Start making many requests to the service you want to profile and wait for 10 seconds:
ab -k -c 1 -n 10000 -H "Authorization: Bearer 1" "http://127.0.0.1:8080/groups/5/team-descendants"

('-c 1' means 'concurrency = 1', you can try other values as well)

  1. Get the profile:
go tool pprof http://127.0.0.1:8080/debug/pprof/profile?seconds=10

Type 'web' as a pprof command to see the call graph with durations.