A modern template for a Servant application.

The project aims to provide a template for a Servant project featuring:

• nix support via flakes.
• database interaction with rel8;
• JWT authentication with servant-auth;
• logging with co-log-core;
• TOML configuration using tomland;
• first class records in Servant API using NamedRecords.

The application allows users to categorify contents by tags. Any content can have many tags and any tag could be used for several contents.

It allows also to retrieve contents by a set of tags.

A more in depth description of the architecture of the application can be found in ARCHITECTURE.md.

Configuration of the application is managed using TOML. The application requires a configuration file with the following format:

[database]
  host     = "localhost"
  port     = 5432
  dbname   = "tagger-db"
  user     = "tagger-user"
  password = "tagger-pwd"

[api]
  port = 8080

By default, the file is located in config.toml, but the path is actually configurable with the config option.

The main endpoints of the application are protected by JWT authentication. To access them you first need to get an authorization token for a user.

To get it you first need to register a user by calling the register endpoint.

Next, you can obtain a token by calling the login endpoint with the same data provided to the register endpoint.

Eventually, you should pass your token in the Authorization header for the relevant endpoints to access them.

The project is setup to be built with Cabal, though with dependencies provided via Nix. Tasks are provided using a GNU Make-like tool called Task and are supposed to be run inside a Nix shell.

Available tasks can be seen with task --list:

❯ task --list
task: Available tasks for this project:
* api:build:        Build the API
* api:dev:          Typecheck the API in a loop
* api:docs:         Build Haddock docs
* api:repl:         Start a cabal REPL
* api:serve:        Serve the API
* api:test:         Run API tests
* db:destroy:       Destroy the database
* db:setup:         Setup a postgres database using the config file
* fe:build:         Build the frontend app
* fe:serve:         Serve the frontend app

The usage of tasks is completely optional, and direct invocations of cabal (e.g. cabal build, cabal test, etc.) and elm commands are also valid.

ℹ️ If direnv is installed, it is also possible to use it with nix. The provided .envrc file is already configured to use nix and only needs to be enabled by issuing direnv allow in the project root once.

Setting up the runtime dependencies for development, such as the database, is taken care of by task setup. This calls individual components' setup scripts such as task db:setup under the hood, if you prefer to call it directly.

This task (and others in general) assumes the existence of a few packages (toml2json, jq, postgres, etc.) provided by the nix shell.

In the root of the project you can find a docker-compose.yml file which provides a Postgresql database and a web interface to it, exposed on port 8081. You can initialise the schema of the database by running the schema.sql which is also provided.

To build the API, run

# With task
task api:build

# With cabal
cabal build

⚠️ Note for non-nix users: the build requires the presence of the pg_config executable which is made available by installing Postgresql. Nix takes care of this automatically.

You can launch the web server using

# With task
task api:serve

### Running the API tests

To run the tests, run

```sh
# With task
task api:test

# With Cabal
cabal test

which will expose the service on port defined in configuration.

⚠️ Note for non-nix users: serving the API with hot-reloading requires the presence of the watchexec utility which is made available by the nix shell. Install it manually if you wish to use this script.

The executable accepts two options:

• --config, which allows to customize the path of the configuration file
• --jwk, which allows to customize the path of the file where the JWK is stored

The Haskell files are formatted using ormolu. The Elm source code is formatted using elm-format. The executables are provided in the nix shell.

Git commit hooks are installed by the nix shell and run checks before a commit. These include linting, formatting, etc.

These checks can also be run manually with pre-commit run.

You can generate the documentation of the project using

# With task
task api:docs

# With Cabal
cabal haddock

You can access the OpenAPI documentation just by visiting the docs endpoint (by default https://localhost:8080/docs)

This repository contains also a client Elm application to interact in a human-friendly way with the Tagger api.

You can find more details in elm/README.md, but there are convenience commands:

# With task
task fe:build

# With nix
nix-shell --run 'cd elm; elm make src/Main.elm'

# With npm
cd elm; npx elm make src/Main.elm

# With task
task fe:serve

# With nix
nix-shell --run 'cd elm; elm-live src/Main.elm'

# With npm
cd elm; npx elm-live src/Main.elm