Uploaded by
Published on

A powerful general purpose type for writing applications.

For an extensive overview checkout the tests.


spago install yoga-om


Check out the minimal example!

What can it do?

An Om consists of three parts. A context, potential errors, and the value of the computation.

Conventionally in code, we abbreviate this as Om ctx errs a.

ctx: Dependency injection

Om supports writing code that asks for dependencies that you only provide once at the start of the application.

myOmThatNeedsDbName = do
  { dbName } <- ask

This is an Om of the shape Om { dbName :: String } _ _. We are only concerned about the ctx part of Om in this section, hence the _s.

errors: Mix-and match errors

Dealing with failure in Om is just wonderful.

Throwing errors

To throw an error, simply use the throwError function with a record that has the name of the error as its label and the error data as its value:

myOmThatThrows = Om.throw { myError: "This failed" }

This is an Om _ ( myError :: String ) _

Handling errors

To catch one or multiple errors use handleErrors

myOmHandled = handleErrors { myError: \text -> Console.warn text } myOmThatThrows

As stated earlier myOmThatThrows has type: Om _ (myError :: String) _.

By handling all potential myError errors we've produced myOmHandled which has the type Om _ () _.

Combining errors

A powerful feature of Om is that you can easily combine different Om computations that can throw different errors.

om1 :: forall otherErrors. Om _ ( ioError :: Int | otherErrors ) _
om1 = throw { ioError: -8 }

om2 :: forall otherErrors. Om _ ( userError :: String | otherErrors ) _
om2 = throw { userError: "Your last name can't be shorter than 0 characters" }

om3 :: forall errs. Om _ ( ioError :: Int, userError :: String | errs ) _ 
om3 = do 

This means that you can and should only tag a function with the errors it can actually throw and not the complete set of errors that might happen anywhere in your program.

The compiler helps you out with this.

Working with Aff and Effect

In order to transform any Aff a into an Om _ _ a you may use liftAff, or fromAff. To do the same with Effect you may use liftEffect.

Running an Om

Let's bring it all back home. Eventually you want to actually run an Om. Most probably at the start of your application, in a main :: Effect Unit function for example.

That's the right time to supply the dependencies to your Om and to handle any remaining possible errors:

module Main where

import Prelude
import Node.Process (lookupEnv)
import Effect.Class.Console as Console
import Effect (Effect)
import Yoga.Om as Om
import Yoga.Om (Om)

main :: Effect Unit
main = do  
  name <- lookupEnv "NAME"
    # Om.launchOm_ 
      { name }
      { exception: 
          \e -> Console.error ("Unexpected exception: " <> show e)   
      , nameNotFound:
          \_ -> Console.error "Make sure the $NAME env variable is set"

greet :: Om { envName :: Maybe String } ( nameNotFound :: Unit ) Unit
greet = do
  { envName } <- ask
  name <- envName # Om.note { nameNotFound: unit }
  Console.log $ "Welcome " <> name 

Parallel computations

You can run different Oms in parallel.

Either the fastest one wins:

Om.race [ Om.delay (1.0 # Seconds) *> pure "slow" , pure "fast" ]

Or you look at all the results and get an Om _ _ (Array _):

    [ Om.delay (9.0 # Milliseconds) *> pure "1"
    , Om.delay (1.0 # Milliseconds) *> pure "2"


What if I need State and Writer?

Since Om combines the powers of Reader and supports Effectful computations it is less clutter (especially on the type signature) to bolt this functionality on ad-hoc via Refs in the ctx

It's so tedious to repeat the labels and keys in ctx and errs

You will probably prefer to define type aliases:

module Main where

import DB as DB
import Cache as Cache
-- ...
import Type.Row (type (+))

myWholeApp :: Om (DB.Ctx + Cache.Ctx ()) (DB.Errs + Cache.Errs ()) Unit
myWholeApp = do
  DB.writeToDB "'); DROP TABLE orders;--"
module DB where

writeToDB :: forall ctx errs. String -> Om (Ctx ctx) (Errs errs) Unit
writeToDB s = do
  -- ...
  pure unit 

type Ctx r = (dbCtx :: { port :: Int, hostname :: String } | r)

type Errs r = IOErr + TimeoutErr + r
type IOErr r = ( ioError :: { message :: String, code :: Int, details :: String } | r )
type TimeoutErr r = ( timeout :: { query :: String } | r )

Technical behind the scenes info

Om is a monad transformer stack built from the stack safe continuation passing RWSET without State or Writer with errors specialised to polymorphic Variants.