A RESTful API template (built with Go)

The goal of this project is to be an example of a relational database-backed REST HTTP Web Server that has characteristics needed to ensure success in a high volume environment. This project co-opts the DIY ethos of the Go community and does its best to "use the standard library" whenever possible, bringing in third-party libraries when not doing so would be unduly burdensome (structured logging, Oauth2, etc.).

I struggled a lot with parsing the myriad different patterns people have for package layouts and have tried to coalesce what I've learned from others into my own take on a package layout. Below, I hope to communicate how this structure works. If you have any questions, open an issue or send me a note - I'm happy to help! Also, if you disagree or have suggestions, please do the same, I really enjoy getting both positive and negative feedback.

The following is an in-depth walkthrough of this project. This is a demo API, so the "business" intent of it is to support basic CRUD (Create, Read, Update, Delete) operations for a movie database. All paths to files or directories are from the project root.

• Go
• PostgreSQL - Database
• Google OAuth 2.0 - authentication
• Task - task runner for build and script execution
• CUE - config file generation

Briefly, the data model for this project is set up to enable a B2B multi-tenant SAAS, which is overkill for a simple CRUD app, however, it's the model I wanted to create and teach myself. That said, it can serve only one tenant just fine.

• Person: from Wikipedia: "A person (plural people or persons) is a being that has certain capacities or attributes such as reason, morality, consciousness or self-consciousness, and being a part of a culturally established form of social relations such as kinship, ownership of property, or legal responsibility. The defining features of personhood and, consequently, what makes a person count as a person, differ widely among cultures and contexts."
• User: from Wikipedia: "A user is a person who utilizes a computer or network service." In the context of this project, given that we allow Persons to authenticate with multiple providers, a User is akin to a persona (Wikipedia - "The word persona derives from Latin, where it originally referred to a theatrical mask. On the social web, users develop virtual personas as online identities.") and as such, a Person can have one or many Users (for instance, I can have a GitHub user and a Google user, but I am just one Person). As a general, practical matter, most operations are considered at the User level. For instance, roles are assigned at the user level instead of the Person level, which allows for more fine-grained access control.
• App: is an application that interacts with the system. An App always belongs to just one Org.
• Org: represents an Organization (company, institution or any other organized body of people with a particular purpose). An Org can have multiple Persons/Users and Apps.

The following are basic instructions for getting started.

Clone the code:

$ git clone https://github.com/gilcrest/diygoapi.git
Cloning into 'diygoapi'...

or use the Github CLI (also written in Go!):

$ gh repo clone gilcrest/diygoapi
Cloning into 'diygoapi'...

All requests with this demo webserver require authentication. I have chosen to use Google's Oauth2 solution for these APIs. To use this, you need to setup a Client ID and Client Secret and obtain an access token. The instructions here are great.

After Oauth2 setup with Google, I recommend the Google Oauth2 Playground to obtain fresh access tokens for testing.

Once a user has authenticated through this flow, all calls to services require that the Google access token be sent as a Bearer token in the Authorization header.

• If there is no token present, an HTTP 401 (Unauthorized) response will be sent and the response body will be empty.
• If a token is properly sent, the Google Oauth2 v2 API is used to validate the token. If the token is invalid, an HTTP 401 (Unauthorized) response will be sent and the response body will be empty.

Note: For more details on the authentication model, see the Authentication Detail section below.

If the user's Bearer token is valid, the user must be authorized. Users must first register with the system and be given a role. Currently, the SelfRegister service accommodates this and automatically assigns a default role movieAdmin (functionality will be added eventually for one person registering another).

Roles are assigned permissions and permissions are assigned to resources (service endpoints). The system uses a role-based access control model. The user's role is used to determine if the user is authorized to access a particular endpoint/resource.

The movieAdmin role is set up to grant access to all resources. It's a demo... so why not?

Note: For more details on the authorization model, see the Authorization Detail section below.

All programs in this project (the web server, database tasks, etc.) use the ff library from Peter Bourgon for configuration. The priority order is: CLI flags > environment variables > config file > defaults. The config file defaults to ./config/config.json, so the simplest path, for local development, is to create that file.

If you are using Claude Code, you can run /db-init-config to be guided through configuration setup interactively — it collects your database connection values, generates a fresh encryption key, writes config/config.cue, and runs task gen-config for you.

Regardless of which configuration approach you choose, you need a 256-bit ciphertext string, which can be parsed to a 32 byte encryption key. Generate the ciphertext with task new-key:

$ task new-key
Key Ciphertext: [31f8cbffe80df0067fbfac4abf0bb76c51d44cb82d2556743e6bf1a5e25d4e06]

Copy the key ciphertext between the brackets to your clipboard to use in one of the options below

Security Disclaimer: Config files make local development easier, however, putting any credentials (encryption keys, username and password, etc.) in a config file is a bad idea from a security perspective. At a minimum, you should have the config/ directory added to your .gitignore file so these configs are not checked in. As this is a template repo, I have checked this all in for example purposes only. The data there is bogus. In an upcoming release, I will integrate with a secrets management platform like GCP Secret Manager or HashiCorp Vault See Issue 91.

The config uses a multi-target layout where each target (e.g. local, staging) has its own settings. Create or edit the JSON file at ./config/config.json. Update the encryption_key, database fields (host, port, name, user, password, search_path) and other settings as appropriate for your PostgreSQL installation.

{
    "default_target": "local",
    "targets": [
        {
            "target": "local",
            "server_listener_port": 8080,
            "logger": {
                "min_log_level": "trace",
                "log_level": "debug",
                "log_error_stack": false
            },
            "encryption_key": "31f8cbffe80df0067fbfac4abf0bb76c51d44cb82d2556743e6bf1a5e25d4e06",
            "database": {
                "host": "localhost",
                "port": 5432,
                "name": "dga_local",
                "user": "demo_user",
                "password": "REPLACE_ME",
                "search_path": "demo"
            }
        }
    ]
}

Setting the schema search path properly is critical as the objects in the migration scripts intentionally do not have qualified object names and will therefore use the search path when creating or dropping objects (in the case of the db down migration).

If you prefer, you can generate the JSON config file using CUE.

The CUE-based config uses a split layout:

• config/cue/schema.cue -- the shared validation schema (checked into git)
• config/config.cue -- local config values with credentials (gitignored)
• config/config.json -- generated output (gitignored)

Edit the ./config/config.cue file. Update the encryption_key, database fields (host, port, name, user, password, search_path) and other settings as appropriate for your PostgreSQL installation.

After modifying the CUE file, run the following from project root:

$ task gen-config

This should produce the JSON config file mentioned above (at ./config/config.json).

As an alternative, you can set environment variables directly through bash or whatever strategy you use. Environment variables override config file values. An example bash script:

#!/bin/bash

# encryption key
export ENCRYPT_KEY="31f8cbffe80df0067fbfac4abf0bb76c51d44cb82d2556743e6bf1a5e25d4e06"

# server listen port
export PORT="8080"

# logger environment variables
export LOG_LEVEL_MIN="trace"
export LOG_LEVEL="debug"
export LOG_ERROR_STACK="false"

# Database Environment variables
export DB_HOST="localhost"
export DB_PORT="5432"
export DB_NAME="dga_local"
export DB_USER="demo_user"
export DB_PASSWORD="REPLACE_ME"
export DB_SEARCH_PATH="demo"

For full control, you can pass command line flags directly when running a program. Flags take the highest priority, overriding both environment variables and config file values. The following table lists all available flags, their equivalent environment variables, and defaults:

For example:

$ go run ./cmd/diy/main.go -db-name=dga_local -db-user=demo_user -db-password=REPLACE_ME -db-search-path=demo -encrypt-key=31f8cbffe80df0067fbfac4abf0bb76c51d44cb82d2556743e6bf1a5e25d4e06

The following steps create the database objects and initialize data needed for running the web server. As a convenience, database migration programs which create these objects and load initial data can be executed using Task. To understand database migrations and how they are structured in this project, you can watch this talk I gave to the Boston Golang meetup group in February 2022. The below examples assume you have already setup PostgreSQL and know what user, database and schema you want to install the objects.

If you want to create an isolated database and schema, you can find examples of doing that at ./scripts/db/db_init.sql.

All database tasks read connection info from ./config/config.json by default, using the default_target defined in the config. To target a different environment, pass --target via CLI args, e.g. task db-up -- --target staging. You can also override the target with the TARGET environment variable.

Database initialization (creating the user, database, and schema) requires elevated privileges. Define a local-admin target in your config with a superuser (or a role that has CREATEROLE and CREATEDB), then run:

$ task db-init -- --db-admin-config-target local-admin

This idempotently creates the database user, database, and schema using values from the config's default target. To specify a different app target: task db-init -- --db-admin-config-target local-admin --app-config-target staging.

Fifteen database migration scripts are run as part of the up migration:

$ task db-up

Note: At any time, you can drop all the database objects created as part of the up migration using the down migration scripts in ./scripts/db/migrations/down/.

There are a number of tables that require initialization of data to facilitate things like: authentication through role based access controls, tracking which applications/users are interacting with the system, etc. I have bundled this initialization into a Genesis service, which can be run only once per database.

To call Genesis, send a POST request to /api/v1/genesis. The Genesis endpoint uses a special authentication middleware (genesisAuthHandler) that only validates the user's Bearer token — no app authentication is required, since apps don't exist yet. An example request body is provided at ./config/genesis/request.json:

$ curl --location --request POST 'http://127.0.0.1:8080/api/v1/genesis' \
--header 'Content-Type: application/json' \
--header 'x-auth-provider: google' \
--header 'Authorization: Bearer <REPLACE WITH ACCESS TOKEN>' \
--data @./config/genesis/request.json

The request body defines:

• user: The OAuth2 provider and token for the user calling Genesis (this user becomes the system's first admin).
• org: The user-initiated organization and app to create (your organization for interacting with the Movie APIs).
• permissions: The set of permissions (resource + operation pairs) to create, e.g. POST /api/v1/movies.
• roles: The roles to create, each with a list of permissions to assign to it.

Edit ./config/genesis/request.json before calling the service — replace the user.token with a valid Google OAuth2 access token and org.app.oauth2_provider_client_id with your Google OAuth2 Client ID.

This initial data setup as part of Genesis creates a Principal organization, a Test organization and apps/users within those as well as sets up permissions and roles for access for the user input into the service. The Principal org is created solely for the administrative purpose of creating other organizations, apps and users. The Test organization is where all tests are run for test data isolation, etc.

Most importantly, a user initiated organization and app is created based on your input. The response details of this organization (located within the userInitiated node of the response) are those which are needed to run the various Movie APIs (create movie, read movie, etc.)

Genesis creates the following seed data within a single database transaction:

1. Principal Org — The administrative organization, with a "Developer Dashboard" app created within it.
2. Test Org — The test organization, with a "Test App" and a test user created within it.
3. User-Initiated Org — Your organization (name and description from the request body), with the app you specified created within it.
4. Permissions — All resource/operation pairs from the request (e.g. GET /api/v1/movies, POST /api/v1/movies).
5. Roles — All roles from the request (e.g. sysAdmin, movieAdmin), each linked to their specified permissions.
6. Role Grants — The authenticated user (the person calling Genesis) is granted all requested roles in both the Principal and User-Initiated orgs. A test role is granted to the test user in the Test org.

The response contains three nodes — principal, test, and userInitiated — each with the org and app details:

{
  "principal": {
    "external_id": "qAG9Gn34ruud86a_",
    "name": "Principal",
    "kind_description": "principal",
    "description": "The Principal org represents the first organization created in the database...",
    "app": {
      "external_id": "iK6yM7pBTTSXJXCF",
      "name": "Developer Dashboard",
      "api_keys": [{ "key": "j-DZh4olLswgodsPDA2NsA==", "deactivation_date": "2099-12-31 ..." }]
    }
  },
  "test": {
    "external_id": "h-o4hvaClqlM0V3d",
    "name": "Test Org",
    "kind_description": "test",
    "app": {
      "external_id": "upbfVbnyqByyuOUs",
      "name": "Test App",
      "api_keys": [{ "key": "sD-i1kYNWtGaFNauXhKZ6A==", "deactivation_date": "2099-12-31 ..." }]
    }
  },
  "userInitiated": {
    "external_id": "a2gmurv3P9Ws1ybk",
    "name": "Movie Makers Unlimited",
    "kind_description": "standard",
    "app": {
      "external_id": "zakvyaRpCu4zmt1-",
      "name": "Movie Makers App",
      "api_keys": [{ "key": "2BJd-AYbXJdzHWmiphtxxA==", "deactivation_date": "2099-12-31 ..." }]
    }
  }
}

The full response is saved to ./config/genesis/response.json and can be retrieved later via GET /api/v1/genesis.

Most importantly, the userInitiated node contains the app external_id and api_keys[0].key values needed as the x-app-id and x-api-key headers for all subsequent API calls (see Step 7).

The project tests require that Genesis has been run successfully. If all went well in step 4, you can run the following command to validate:

$ task test

Note: Some tests require a running database with Genesis data. Packages without database dependencies can be tested independently.

With configuration handled in Step 3, start the web server with Task:

$ task run
{"level":"info","time":1675700939,"severity":"INFO","message":"minimum accepted logging level set to trace"}
{"level":"info","time":1675700939,"severity":"INFO","message":"logging level set to debug"}
{"level":"info","time":1675700939,"severity":"INFO","message":"log error stack via github.com/pkg/errors set to false"}
{"level":"info","time":1675700939,"severity":"INFO","message":"sql database opened for localhost on port 5432"}
{"level":"info","time":1675700939,"severity":"INFO","message":"sql database Ping returned successfully"}
{"level":"info","time":1675700939,"severity":"INFO","message":"database version: PostgreSQL 14.6 on aarch64-apple-darwin20.6.0, compiled by Apple clang version 12.0.5 (clang-1205.0.22.9), 64-bit"}
{"level":"info","time":1675700939,"severity":"INFO","message":"current database user: demo_user"}
{"level":"info","time":1675700939,"severity":"INFO","message":"current database: dga_local"}
{"level":"info","time":1675700939,"severity":"INFO","message":"current search_path: demo"}
{"level":"info","time":1675700939,"severity":"INFO","message":"server listening on :8080"}

You can also run directly with go run ./cmd/diy/main.go, passing flags or relying on environment variables / config file as described in Step 3.

With the server up and running, the easiest service to interact with is the ping service. This service is a simple health check that returns a series of flags denoting health of the system (queue depths, database up boolean, etc.). For right now, the only thing it checks is if the database is up and pingable. I have left this service unauthenticated so there's at least one service that you can get to without having to have an authentication token, but in actuality, I would typically have every service behind a security token.

Use cURL GET request to call ping:

$ curl --location --request GET 'http://127.0.0.1:8080/api/v1/ping' \
--header 'x-auth-provider: google' \
--header 'Authorization: Bearer <REPLACE WITH ACCESS TOKEN>'
{"db_up":true}

The values for the x-app-id and x-api-key headers needed for all below services are found in the /api/v1/genesis service response. The response can be found at ./config/genesis/response.json:

• APP ID (x-app-id): userInitiated.app.external_id
• API Key (x-api-key): userInitiated.app.api_keys[0].key

The Bearer token for the Authorization header needs to be generated through Google's OAuth2 mechanism. Assuming you've completed setup mentioned in Step 2, you can generate a new token at the Google OAuth2 Playground

Create Movie - use the POST HTTP verb at /api/v1/movies:

$ curl --location --request POST 'http://127.0.0.1:8080/api/v1/movies' \
--header 'Content-Type: application/json' \
--header 'x-app-id: <REPLACE WITH APP ID>' \
--header 'x-api-key: <REPLACE WITH API KEY>' \
--header 'x-auth-provider: google' \
--header 'Authorization: Bearer <REPLACE WITH ACCESS TOKEN>' \
--data-raw '{
    "title": "Repo Man",
    "rated": "R",
    "release_date": "1984-03-02T00:00:00Z",
    "run_time": 92,
    "director": "Alex Cox",
    "writer": "Alex Cox"
}'
{"external_id":"IUAtsOQuLTuQA5OM","title":"Repo Man","rated":"R","release_date":"1984-03-02T00:00:00Z","run_time":92,"director":"Alex Cox","writer":"Alex Cox","create_app_extl_id":"nBRyFTHq6PALwMdx","create_username":"[email protected]","create_user_first_name":"Otto","create_user_last_name":"Maddox","create_date_time":"2022-06-30T15:26:02-04:00","update_app_extl_id":"nBRyFTHq6PALwMdx","update_username":"[email protected]","update_user_first_name":"Otto","update_user_last_name":"Maddox","update_date_time":"2022-06-30T15:26:02-04:00"}

Read (Single Record) - use the GET HTTP verb at /api/v1/movies/:extl_id with the movie external_id from the create (POST) response as the unique identifier in the URL. I try to never expose primary keys, so I use something like an external id as an alternative key.

$ curl --location --request GET 'http://127.0.0.1:8080/api/v1/movies/IUAtsOQuLTuQA5OM' \
--header 'x-app-id: <REPLACE WITH APP ID>' \
--header 'x-api-key: <REPLACE WITH API KEY>' \
--header 'x-auth-provider: google' \
--header 'Authorization: Bearer <REPLACE WITH ACCESS TOKEN>' \
{"external_id":"IUAtsOQuLTuQA5OM","title":"Repo Man","rated":"R","release_date":"1984-03-02T00:00:00Z","run_time":92,"director":"Alex Cox","writer":"Alex Cox","create_app_extl_id":"QfLDvkZlAEieAA7u","create_username":"[email protected]","create_user_first_name":"Otto","create_user_last_name":"Maddox","create_date_time":"2022-06-30T15:26:02-04:00","update_app_extl_id":"QfLDvkZlAEieAA7u","update_username":"[email protected]","update_user_first_name":"Otto","update_user_last_name":"Maddox","update_date_time":"2022-06-30T15:26:02-04:00"}

Read (All Records) - use the GET HTTP verb at /api/v1/movies:

$ curl --location --request GET 'http://127.0.0.1:8080/api/v1/movies' \
--header 'x-app-id: <REPLACE WITH APP ID>' \
--header 'x-api-key: <REPLACE WITH API KEY>' \
--header 'x-auth-provider: google' \
--header 'Authorization: Bearer <REPLACE WITH ACCESS TOKEN>' \

Update - use the PUT HTTP verb at /api/v1/movies/:extl_id with the movie external_id from the create (POST) response as the unique identifier in the URL.

$ curl --location --request PUT 'http://127.0.0.1:8080/api/v1/movies/IUAtsOQuLTuQA5OM' \
--header 'Content-Type: application/json' \
--header 'x-app-id: <REPLACE WITH APP ID>' \
--header 'x-api-key: <REPLACE WITH API KEY>' \
--header 'x-auth-provider: google' \
--header 'Authorization: Bearer <REPLACE WITH ACCESS TOKEN>' \
--data-raw '{
    "title": "Repo Man",
    "rated": "R",
    "release_date": "1984-03-02T00:00:00Z",
    "run_time": 91,
    "director": "Alex Cox",
    "writer": "Alex Cox"
}'
{"external_id":"IUAtsOQuLTuQA5OM","title":"Repo Man","rated":"R","release_date":"1984-03-02T00:00:00Z","run_time":91,"director":"Alex Cox","writer":"Alex Cox","create_app_extl_id":"QfLDvkZlAEieAA7u","create_username":"[email protected]","create_user_first_name":"Otto","create_user_last_name":"Maddox","create_date_time":"2022-06-30T15:26:02-04:00","update_app_extl_id":"nBRyFTHq6PALwMdx","update_username":"[email protected]","update_user_first_name":"Otto","update_user_last_name":"Maddox","update_date_time":"2022-06-30T15:38:42-04:00"}

Delete - use the DELETE HTTP verb at /api/v1/movies/:extl_id with the movie external_id from the create (POST) response as the unique identifier in the URL.

$ curl --location --request DELETE 'http://127.0.0.1:8080/api/v1/movies/IUAtsOQuLTuQA5OM' \
--header 'x-app-id: <REPLACE WITH APP ID>' \
--header 'x-api-key: <REPLACE WITH API KEY>' \
--header 'x-auth-provider: google' \
--header 'Authorization: Bearer <REPLACE WITH ACCESS TOKEN>' \
{"extl_id":"IUAtsOQuLTuQA5OM","deleted":true}

The above image is a high-level view of an example request that is processed by the server (creating a movie). To summarize, after receiving an http request, the request path, method, etc. is matched to a registered route in the Server's standard library multiplexer (aka ServeMux, initialization of which, is part of server startup in the cmd package as part of the routes.go file in the server package). The request is then sent through a sequence of middleware handlers for setting up request logging, response headers, authentication and authorization. Finally, the request is routed through a bespoke app handler, in this case handleMovieCreate.

diygoapi package layout is based on several projects, but the primary source of inspiration is the WTF Dial app repo and accompanying blog from Ben Johnson. It's really a wonderful resource and I encourage everyone to read it.

Handling errors is really important in Go. Errors are first class citizens and there are many different approaches for handling them. I have based my error handling on a blog post from Rob Pike and have modified it to meet my needs. The post is many years old now, but I find the lessons there still hold true at least for my requirements.

My requirements for REST API error handling are the following:

• Requests for users who are not properly authenticated should return a 401 Unauthorized error with a WWW-Authenticate response header and an empty response body.
• Requests for users who are authenticated, but do not have permission to access the resource, should return a 403 Forbidden error with an empty response body.
• All requests which are due to a client error (invalid data, malformed JSON, etc.) should return a 400 Bad Request and a response body which looks similar to the following:

{
    "error": {
        "kind": "input_validation_error",
        "param": "director",
        "message": "director is required"
    }
}

• All requests which incur errors as a result of an internal server or database error should return a 500 Internal Server Error and not leak any information about the database or internal systems to the client. These errors should return a response body which looks like the following:

{
    "error": {
        "kind": "internal_error",
        "message": "internal server error - please contact support"
    }
}

All errors should return a Request-Id response header with a unique request id that can be used for debugging to find the corresponding error in logs.

All errors should be raised using custom errors from the errs package. The three custom errors correspond directly to the requirements above.

Typical errors raised throughout diygoapi are the custom errs.Error, which look like:

// Error is the type that implements the error interface.
// It contains a number of fields, each of different type.
// An Error value may leave some values unset.
type Error struct {
   // Op is the operation being performed, usually the name of the method
   // being invoked.
   Op Op
   // User is the name of the user attempting the operation.
   User UserName
   // Kind is the class of error, such as permission failure,
   // or "Other" if its class is unknown or irrelevant.
   Kind Kind
   // Param represents the parameter related to the error.
   Param Parameter
   // Code is a human-readable, short representation of the error
   Code Code
   // Realm is a description of a protected area, used in the WWW-Authenticate header.
   Realm Realm
   // The underlying error that triggered this one, if any.
   Err error
}

This custom error type is raised using the E function from the errs package. errs.E is taken from Rob Pike's upspin errors package (but has been changed based on my requirements). The errs.E function call is variadic and can take several different types to form the custom errs.Error struct.

Here is a simple example of creating an error using errs.E:

err := errs.E("seems we have an error here")

When a string is sent, a new error will be created and added to the Err element of the struct. In the above example, Op, User, Kind, Param, Realm and Code would all remain unset.

By convention, we create an op constant to denote the method or function where the error is occuring (or being returned through). This op constant should always be the first argument in each call, though it is not actually required to be.

package opdemo

import (
    "fmt"

    "github.com/gilcrest/diygoapi/errs"
)

// IsEven returns an error if the number given is not even
func IsEven(n int) error {
    const op errs.Op = "opdemo/IsEven"

    if n%2 != 0 {
        return errs.E(op, fmt.Sprintf("%d is not even", n))
    }
    return nil
}

You can set any of these custom errs.Error fields that you like, for example:

var released time.Time
released, err = time.Parse(time.RFC3339, r.Released)
if err != nil {
    return nil, errs.E(op, errs.Validation,
        errs.Code("invalid_date_format"),
        errs.Parameter("release_date"),
        err)
}

Above, we used errs.Validation to set the errs.Kind as Validation. Valid error Kind are:

const (
    Other          Kind = iota // Unclassified error. This value is not printed in the error message.
    Invalid                    // Invalid operation for this type of item.
    IO                         // External I/O error such as network failure.
    Exist                      // Item already exists.
    NotExist                   // Item does not exist.
    Private                    // Information withheld.
    Internal                   // Internal error or inconsistency.
    BrokenLink                 // Link target does not exist.
    Database                   // Error from database.
    Validation                 // Input validation error.
    Unanticipated              // Unanticipated error.
    InvalidRequest             // Invalid Request
    // Unauthenticated is used when a request lacks valid authentication credentials.
    //
    // For Unauthenticated errors, the response body will be empty.
    // The error is logged and http.StatusUnauthorized (401) is sent.
    Unauthenticated // Unauthenticated Request
    // Unauthorized is used when a user is authenticated, but is not authorized
    // to access the resource.
    //
    // For Unauthorized errors, the response body should be empty.
    // The error is logged and http.StatusForbidden (403) is sent.
    Unauthorized
)

errs.Code represents a short code to respond to the client with for error handling based on codes (if you choose to do this) and is any string you want to pass.

errs.Parameter represents the parameter that is being validated or has problems, etc.

Note in the above example, instead of passing a string and creating a new error inside the errs.E function, I am directly passing the error returned by the time.Parse function to errs.E. The error is then added to the Err field using errors.WithStack from the github.com/pkg/errors package, which enables stacktrace retrieval later.

There are a few helpers in the errs package as well, namely the errs.MissingField function which can be used when validating missing input on a field. This idea comes from this Mat Ryer post and is pretty handy.

Here is an example in practice:

// IsValid performs validation of the struct
func (m *Movie) IsValid() error {
    const op errs.Op = "diygoapi/Movie.IsValid"

    switch {
    case m.Title == "":
        return errs.E(op, errs.Validation, errs.Parameter("title"), errs.MissingField("title"))

The error message for the above would read title is required

There is also errs.InputUnwanted which is meant to be used when a field is populated with a value when it is not supposed to be.

As errors created with errs.E move up the call stack, the op can just be added to the error, like the following:

func outer() error {
    const op errs.Op = "opdemo/outer"

    err := middle()
    if err != nil {
        return errs.E(op, err)
    }
    return nil
}

func middle() error {
    err := inner()
    if err != nil {
        return errs.E(errs.Op("opdemo/middle"), err)
    }
    return nil
}

func inner() error {
    const op errs.Op = "opdemo/inner"

    return errs.E(op, "seems we have an error here")
}

Note that errs.Op can be created inline as part of the error instead of creating a constant as done in the middle function, I just prefer to create the constant in most cases.

In addition, you can add context fields (errs.Code, errs.Parameter, errs.Kind) as the error moves up the stack, however, I try to add as much context as possible at the point of error origin and only do this in rare cases.

At the top of the program flow for each route is the handler (for example, Server.handleMovieCreate). In this handler, any error returned from any function or method is sent through the errs.HTTPErrorResponse function along with the http.ResponseWriter and a zerolog.Logger.

For example:

response, err := s.CreateMovieService.Create(r.Context(), rb, u)
if err != nil {
    errs.HTTPErrorResponse(w, logger, err)
    return
}

errs.HTTPErrorResponse takes the custom errs.Error type and writes the response to the given http.ResponseWriter and logs the error using the given zerolog.Logger.

return must be called immediately after errs.HTTPErrorResponse to return the error to the client.

If an errs.Error type is sent to errs.HTTPErrorResponse, the function writes the HTTP response body as JSON using the errs.ErrResponse struct.

// ErrResponse is used as the Response Body
type ErrResponse struct {
    Error ServiceError `json:"error"`
}

// ServiceError has fields for Service errors. All fields with no data will be omitted
type ServiceError struct {
    Kind    string `json:"kind,omitempty"`
    Code    string `json:"code,omitempty"`
    Param   string `json:"param,omitempty"`
    Message string `json:"message,omitempty"`
}

When the error is returned to the client, the response body JSON looks like the following:

{
    "error": {
        "kind": "input validation error",
        "param": "title",
        "message": "title is required"
    }
}

In addition, the error is logged. By default, the error stack is built using the op context added to errors and added to the log as a string array in the stack field (see below). For the majority of cases, I believe this is sufficient.

{
   "level": "error",
   "remote_ip": "127.0.0.1:60382",
   "user_agent": "PostmanRuntime/7.30.1",
   "request_id": "cfgihljuns2hhjb77tq0",
   "stack": [
      "diygoapi/Movie.IsValid",
      "service/MovieService.Create"
   ],
   "error": "title is required",
   "http_statuscode": 400,
   "Kind": "input validation error",
   "Parameter": "title",
   "Code": "",
   "time": 1675700438,
   "severity": "ERROR",
   "message": "error response sent to client"
}

If you feel you need the full error stack trace, you can set the flag, environment variable on startup or call the PUT method for the {{base_url}}/api/v1/logger service to update zerolog.ErrorStackMarshaler and set it to log error stacks (more about this below). The logger will log the full error stack, which can be super helpful when trying to identify issues.

The error log will look like the following (I cut off parts of the stack for brevity):

{
    "level": "error",
    "ip": "127.0.0.1",
    "user_agent": "PostmanRuntime/7.26.8",
    "request_id": "bvol0mtnf4q269hl3ra0",
    "stack": [{
        "func": "E",
        "line": "172",
        "source": "errs.go"
    }, {
        "func": "(*Movie).SetReleased",
        "line": "76",
        "source": "movie.go"
    }, {
        "func": "(*MovieController).CreateMovie",
        "line": "139",
        "source": "create.go"
    }, {
    ...
    }],
    "error": "parsing time \"1984a-03-02T00:00:00Z\" as \"2006-01-02T15:04:05Z07:00\": cannot parse \"a-03-02T00:00:00Z\" as \"-\"",
    "HTTPStatusCode": 400,
    "Kind": "input_validation_error",
    "Parameter": "release_date",
    "Code": "invalid_date_format",
    "time": 1609650267,
    "severity": "ERROR",
    "message": "Response Error Sent"
}

Note: E will usually be at the top of the stack as it is where the errors.New or errors.WithStack functions are being called.

There is logic within errs.HTTPErrorResponse to return a different response body if the errs.Kind is Internal or Database. As per the requirements, we should not leak the error message or any internal stack, etc. when an internal or database error occurs. If an error comes through and is an errs.Error with either of these error Kind or is an unknown error type in any way, the response will look like the following:

{
    "error": {
        "kind": "internal_error",
        "message": "internal server error - please contact support"
    }
}

The spec for 401 Unauthorized calls for a WWW-Authenticate response header along with a realm. The realm should be set when creating an Unauthenticated error.

Unauthenticated errors should only be raised at points of authentication as part of a middleware handler. I will get into application flow in detail later, but authentication for diygoapi happens in middleware handlers prior to calling the final app handler for the given route.

The example below demonstrates returning an Unauthenticated error if the Authorization header is not present. This is done using the errs.E function (common to all errors in this repo), but the errs.Kind is sent as errs.Unauthenticated. An errs.Realm type should be added as well. For now, the constant defaultRealm is set to diygoapi in the server package and is used for all unauthenticated errors. You can set this constant to whatever value you like for your application.

// parseAuthorizationHeader parses/validates the Authorization header and returns an Oauth2 token
func parseAuthorizationHeader(realm string, header http.Header) (*oauth2.Token, error) {
    const op errs.Op = "server/parseAuthorizationHeader"

    // Pull the token from the Authorization header by retrieving the
    // value from the Header map with "Authorization" as the key
    //
    // format: Authorization: Bearer
    headerValue, ok := header["Authorization"]
    if !ok {
        return nil, errs.E(op, errs.Unauthenticated, errs.Realm(realm), "unauthenticated: no Authorization header sent")
    }
...

Per requirements, diygoapi does not return a response body when returning an Unauthenticated error. The error response from cURL looks like the following:

HTTP/1.1 401 Unauthorized
Request-Id: c30hkvua0brkj8qhk3e0
Www-Authenticate: Bearer realm="diygoapi"
Date: Wed, 09 Jun 2021 19:46:07 GMT
Content-Length: 0

If the user is not authorized to use the API, an HTTP 403 (Forbidden) response will be sent and the response body will be empty.

Unauthorized errors are raised when there is a permission issue for a user attempting to access a resource. diygoapi currently has a custom database-driven RBAC (Role Based Access Control) authorization implementation (more about this later). The below example demonstrates raising an Unauthorized error and is found in the DBAuthorizer.Authorize method.

return errs.E(errs.Unauthorized, fmt.Sprintf("user %s does not have %s permission for %s", adt.User.Username, r.Method, pathTemplate))

Per requirements, diygoapi does not return a response body when returning an Unauthorized error. The error response from cURL looks like the following:

HTTP/1.1 403 Forbidden
Request-Id: c30hp2ma0brkj8qhk3f0
Date: Wed, 09 Jun 2021 19:54:50 GMT
Content-Length: 0

diygoapi uses the zerolog library from Olivier Poitrey. The mechanics for using zerolog are straightforward and are well documented in the library's README. zerolog takes an io.Writer as input to create a new logger; for simplicity in diygoapi, I use os.Stdout.

When starting diygoapi, there are several flags which setup the logger:

As mentioned in Step 3, diygoapi uses the ff library from Peter Bourgon, which allows for using flags, environment variables, or a config file. Going forward, we'll assume you've chosen flags.

The log-level flag sets the Global logging level for your zerolog.Logger.

zerolog allows for logging at the following levels (from highest to lowest):

• panic (zerolog.PanicLevel, 5)
• fatal (zerolog.FatalLevel, 4)
• error (zerolog.ErrorLevel, 3)
• warn (zerolog.WarnLevel, 2)
• info (zerolog.InfoLevel, 1)
• debug (zerolog.DebugLevel, 0)
• trace (zerolog.TraceLevel, -1)

The log-level-min flag sets the minimum accepted logging level, which means, for example, if you set the minimum level to error, the only logs that will be sent to your chosen output will be those that are greater than or equal to error (error, fatal and panic).

The log-error-stack boolean flag tells whether to log full stack traces for each error. If true, the zerolog.ErrorStackMarshaler will be set to pkgerrors.MarshalStack which means, for errors raised using the github.com/pkg/errors package, the error stack trace will be captured and printed along with the log. All errors raised in diygoapi are raised using github.com/pkg/errors if this flag is set to true.

After parsing the command line flags, zerolog.Logger is initialized in cmd/cmd.go

// setup logger with appropriate defaults
lgr := logger.NewWithGCPHook(os.Stdout, minlvl, true)

and subsequently used to initialize the server.Server struct.

// initialize Server enfolding a http.Server with default timeouts,
// a mux router and a zerolog.Logger
s := server.New(http.NewServeMux(), server.NewDriver(), lgr)

The Server.registerRoutes method is responsible for registering routes and corresponding middleware/handlers to the Server's multiplexer (aka router). For each route registered to the handler, upon execution, the initialized zerolog.Logger struct is added to the request context through the Server.loggerChain method.

// register routes/middleware/handlers to the Server ServeMux
func (s *Server) registerRoutes() {

	// Match only POST requests at /api/v1/movies
	// with Content-Type header = application/json
	s.mux.Handle("POST /api/v1/movies",
		s.loggerChain().
			Append(s.addRequestHandlerPatternContextHandler).
			Append(s.enforceJSONContentTypeHandler).
			Append(s.appHandler).
			Append(s.authHandler).
			Append(s.authorizeUserHandler).
			Append(s.jsonContentTypeResponseHandler).
			ThenFunc(s.handleMovieCreate))

...

The Server.loggerChain method sets up the logger with pre-populated fields, including the request method, url, status, size, duration, remote IP, user agent, referer. A unique Request ID is also added to the logger, context and response headers.

func (s *Server) loggerChain() alice.Chain {
	ac := alice.New(hlog.NewHandler(s.Logger),
		hlog.AccessHandler(func(r *http.Request, status, size int, duration time.Duration) {
			hlog.FromRequest(r).Info().
				Str("method", r.Method).
				Stringer("url", r.URL).
				Int("status", status).
				Int("size", size).
				Dur("duration", duration).
				Msg("request logged")
		}),
		hlog.RemoteAddrHandler("remote_ip"),
		hlog.UserAgentHandler("user_agent"),
		hlog.RefererHandler("referer"),
		hlog.RequestIDHandler("request_id", "Request-Id"),
	)

	return ac
}

For every request, you'll get a request log that looks something like the following:

{
   "level": "info",
   "remote_ip": "127.0.0.1:60382",
   "user_agent": "PostmanRuntime/7.30.1",
   "request_id": "cfgihljuns2hhjb77tq0",
   "method": "POST",
   "url": "/api/v1/movies",
   "status": 400,
   "size": 90,
   "duration": 85.747943,
   "time": 1675700438,
   "severity": "INFO",
   "message": "request logged"
}

All error logs will have the same request metadata, including request_id. The Request-Id is also sent back as part of the error response as a response header, allowing you to link the two. An error log will look something like the following:

{
    "level": "error",
    "remote_ip": "127.0.0.1",
    "user_agent": "PostmanRuntime/7.28.0",
    "request_id": "c3nppj6a0brt1dho9e2g",
    "error": "googleapi: Error 401: Request is missing required authentication credential. Expected OAuth 2 access token, login cookie or other valid authentication credential. See https://developers.google.com/identity/sign-in/web/devconsole-project., unauthorized",
    "http_statuscode": 401,
    "realm": "diygoapi",
    "time": 1626315981,
    "severity": "ERROR",
    "message": "Unauthenticated Request"
}

The above error log demonstrates a log for an error with stack trace turned off.

If the Logger is to be used beyond the scope of the handler, it should be pulled from the request context in the handler and sent as a parameter to any inner calls. The Logger is added only to the request context to capture request related fields with the Logger and be able to pass the initialized logger and middleware handlers easier to the app/route handler. Additional use of the logger should be directly called out in function/method signatures so there are no surprises. All logs from the logger passed down get the benefit of the request metadata though, which is great!

You can retrieve and update the state of these flags using the {{base_url}}/api/v1/logger endpoint.

To retrieve the current logger state use a GET request:

curl --location --request GET 'http://127.0.0.1:8080/api/v1/logger' \
--header 'Authorization: Bearer <REPLACE WITH ACCESS TOKEN>'

and the response will look something like:

{
    "logger_minimum_level": "debug",
    "global_log_level": "error",
    "log_error_stack": false
}

In order to update the logger state use a PUT request:

curl --location --request PUT 'http://127.0.0.1:8080/api/v1/logger' \
--header 'Content-Type: application/json' \
--header 'Authorization: Bearer <REPLACE WITH ACCESS TOKEN>' \
--data-raw '{
    "global_log_level": "debug",
    "log_error_stack": "true"
}'

and the response will look something like:

{
    "logger_minimum_level": "debug",
    "global_log_level": "debug",
    "log_error_stack": true
}

The PUT response is the same as the GET response, but with updated values. In the examples above, I used a scenario where the logger state started with the global logging level (global_log_level) at error and error stack tracing (log_error_stack) set to false. The PUT request then updates the logger state, setting the global logging level to debug and the error stack tracing. You might do something like this if you are debugging an issue and need to see debug logs or error stacks to help with that.

Authentication is determined by validating the App making the request as well as the User.

An App (aka API Client) can be registered using the POST /api/v1/apps service. The App can be registered with an association to an Oauth2 Provider Client ID or be standalone.

An App has two possible methods of authentication.

1. The first method, which overrides the second, is using the X-APP-ID and X-API-KEY HTTP headers. The X-APP-ID is the app unique identifier and the X-API-KEY is the password. This method confirms the veracity of the App against values stored in the database (the password is encrypted in the db). If the App ID is not found or the API key does not match the stored API key, an HTTP 401 (Unauthorized) response will be sent and the response body will be empty. If the authentication is successful, the App details will be set to the request context for downstream use.
2. If there is no X-APP-ID header present, the second method is using the authorization Provider's Client ID associated with the Authorization header Bearer token. When a request is made only using the Authorization header, a callback to the provider's Oauth2 TokenInfo API is done to retrieve the associated Provider Client ID. The Provider Client ID is then used to find the associated App in the database. If the App is not found, an HTTP 401 (Unauthorized) response will be sent and the response body will be empty. If the App is found, the App details will be set to the request context for downstream use.

User authentication happens outside the API using an Oauth2 provider (e.g. Google, Github, etc.) sign-in page. After successful authentication, the user is given a Bearer token which is then used for service-level authentication.

In order to perform any actions with the diygoapi services, a User must Self-Register. The SelfRegister service creates a Person/User and stores them in the database. In addition, an Auth object, which represents the user's credentials is stored in the db. A search is done prior to Auth creation to determine if the user is already registered, and if so, the existing user is returned.

After a user is registered, they can perform actions (use resources/endpoints) using the services. For every call, the Authorization HTTP header with the user's Bearer token along with the X-AUTH-PROVIDER header used to denote the Provider (e.g. Google, Github, etc.) must be set.

The Bearer token is used to find the Auth object for the user in the database. Searching for the Auth object is done as follows:

• Search the database directly using the Bearer token.
  • If an Auth object already exists in the datastore which matches the Bearer token and the Bearer token is not past its expiration date, the existing Auth will be used to determine the User.
  • If no Auth object exists in the datastore for the request Bearer token, an attempt will be made to find the user's Auth with the Provider ID (e.g. Google) and unique person ID given by the provider (found by calling the provider API with the request Bearer token). If an Auth object exists given these attributes, it will be updated with the new Bearer token details and this Auth will be used to obtain the User details.

If the Auth object is not found, an HTTP 401 (Unauthorized) response will be sent and the response body will be empty. If the Auth object is found, the User details will be set to the request context for downstream use.

After a user is authenticated, their ability to access a given resource is determined by authorization. diygoapi implements a custom, database-driven Role Based Access Control (RBAC) model. Authorization is enforced via the authorizeUserHandler middleware, which runs after authentication in the middleware chain for every protected route:

s.mux.Handle("POST /api/v1/movies",
    s.loggerChain().
        Append(s.addRequestHandlerPatternContextHandler).
        Append(s.enforceJSONContentTypeHandler).
        Append(s.appHandler).              // authenticate app
        Append(s.authHandler).             // authenticate user
        Append(s.authorizeUserHandler).    // authorize user (RBAC)
        Append(s.jsonContentTypeResponseHandler).
        ThenFunc(s.handleMovieCreate))

The authorizeUserHandler middleware retrieves the authenticated User and App from the request context (set by earlier authentication middleware), then delegates to DBAuthorizationService.Authorize. This method extracts the handler pattern (e.g. POST /api/v1/movies) from the request context and splits it into the HTTP method (POST) and the resource path (/api/v1/movies). It then executes a database query to determine if the user is authorized.

If the user is not authorized, an HTTP 403 (Forbidden) response is returned with an empty response body.

The RBAC model is built on three core domain types, defined in the root package (auth.go):

• Permission: An approval of a mode of access to a resource. Each permission pairs a Resource (an HTTP route, e.g. /api/v1/movies) with an Operation (an HTTP method, e.g. POST). Permissions have an Active flag — only active permissions grant access.
• Role: A job function or title which defines an authority level (e.g. movieAdmin). Each role has a Code, a Description, and a list of Permissions.
• User: Roles are assigned at the User level (not the Person level), allowing for fine-grained access control across personas.

These are linked together through four database tables:

The users_role table's composite primary key of (user_id, role_id, org_id) means a user's roles are scoped per organization, enabling multi-tenant access control.

The authorization check is a single SQL query (IsAuthorized) that joins through the RBAC chain:

SELECT ur.user_id
FROM users_role ur
         INNER JOIN role_permission rp on rp.role_id = ur.role_id
         INNER JOIN permission p on p.permission_id = rp.permission_id
WHERE p.active = true
  AND p.resource = $1      -- e.g. '/api/v1/movies'
  AND p.operation = $2     -- e.g. 'POST'
  AND ur.user_id = $3      -- authenticated user
  AND ur.org_id = $4;      -- org from authenticated app

The organization context (org_id) is derived from the authenticated App — each app belongs to exactly one Org. This means authorization is determined by: does this user have a role (within this app's organization) that includes a permission matching the requested resource and HTTP method?

For example, when a user calls POST /api/v1/movies, the query checks: does users_role contain a row for this user and org? Does that role have a role_permission entry linking to a permission where resource = '/api/v1/movies' and operation = 'POST'? If so, the user is authorized.

If the query returns a matching user_id, the request proceeds to the handler. If not, the middleware returns an HTTP 403 (Forbidden) response with an empty body and logs the unauthorized attempt.

Note: For details on the 403 response format, see the Unauthorized Errors section above.