servant-docs-simple

Introduction
This library uses Data.Typeable to generate documentation for Servant API types.
It relies on the typeRep of Servant's combinators and other datatypes used in the API to generate the documentation.
Why do we need this?
-
We need the API Format types inside the documentation. They can be used as keys to look up examples, fields, and other miscellaneous details.
-
We want to generate an overview of our endpoints from the information our API type provides, without having to write instances.
In-depth explanation here
Tutorials
Example usage
Using this script
-- stack runghc --package servant-docs-simple {-# LANGUAGE DataKinds #-} {-# LANGUAGE TypeApplications #-} {-# LANGUAGE TypeOperators #-} module Main where import Servant.API ((:>), Post, ReqBody) import Servant.Docs.Simple (writeDocsJson, rriteDocsPlainText) -- Our API type type API = "hello" :> "world" :> Request :> Response type Request = ReqBody '[()] () type Response = Post '[()] () main :: IO () main = do -- Writes to the file $PWD/docs.json writeDocsJson @API "docs.json" -- Writes to the file $PWD/docs.txt writeDocsPlainText @API "docs.txt"
Expected Output
Files should be generated relative to $PWD
$ ls | grep docs docs.json docs.txt
docs.json
{ "/hello/world": { "Response": { "Format": "[()]", "ContentType": "()" }, "RequestType": "'POST", "RequestBody": { "Format": "[()]", "ContentType": "()" } } }
docs.txt
/hello/world: RequestBody: Format: [()] ContentType: () RequestType: 'POST Response: Format: [()] ContentType: ()
Tutorials
Click on these links for tutorials
Generating plaintext/JSON documentation from api types
Generating the intermediate documentation structure
Writing our own rendering format
Parsing custom API combinators
To run tutorial scripts
# clone this repository git clone git@github.com:Holmusk/servant-docs-simple.git # run the examples stack examples/<source file>
FAQ
What's the usecase for this?
-
Lightweight documentation solution by providing an overview of endpoints.
-
Keeping documentation for types in one place
Suppose we have a library of format types. These are shared between different languages and formats. If we centralize all the information in this library, we have one source of truth.
-- --- Format Types --- -- | | -- v v type SingleAPI = "hello" :> ReqBody '[JSON] User :> POST '[JSON] Message data User = User Name Password deriving Typeable data Messsage = Message Id Content deriving Typeable
We use these format types in our API as illustrated above. Developers can use these format types to index the library, to find information (examples, instances, ...) they need about the type.
What is required to achieve this?
-
We need documentation to be generated from the API type, without writing extra instances for each format type we use.
We can do this with typeRep from Data.Typeable.
$ stack ghci Prelude> import Data.Typeable Prelude Data.Typeable> data Alignment = Good | Bad deriving Typeable Prelude Data.Typeable> typeRep (Proxy :: Proxy Alignment) Alignment
-
We need access to the API format types (Email, Users, ...) in documentation.
This is so we can index our library with these types to get the relevant fields, examples for these.
How does our usecase differ from Servant.Docs?
-
Servant.Docs generates documentation only if Format Types have implemented the necessary instances.
-- Instances for format types; they provide examples for format types instance ToSample User where toSamples _ = <some example> instance ToSample Message where toSamples _ = <some example>
Hence, we cannot generate documentation solely from the Format Type.
-
In documentation generated by Servant.Docs, the format types mentioned above (User, Message) are not included. This means we can't use them to index our library to look for the relevant information.
What are the trade-offs?
-
Currently Servant.Docs.Simple does not support as many formats as Servant.Docs (via Servant.Docs.Pandoc).
-
Examples are not mandatory as you do not have to write instances for your format types.
This means that if developers don't provide examples, the generated documentation would be without them.
We can still provide these examples through: