In an elm-pages app, each Route Module can define a value data which is a DataSource that will be resolved before init is called. That means it is also available
when the page's HTML is pre-rendered during the build step. You can also access the resolved data in head to use it for the page's SEO meta tags.
A DataSource lets you pull in data from:
DataSource.File)DataSource.Http)content/*.txt (DataSource.Glob)DataSource.Port)DataSource.succeed "Hello!")DataSource.map2, DataSource.andThen, or other combining/continuing helpers from this moduleDataSource's are always resolved before the page is rendered and sent to the browser. A DataSource is never executed in the Browser. Instead, the resolved data from the DataSource is passed down to the Browser - it has been resolved before any client-side JavaScript ever executes. In the case of a pre-rendered route, this is during the CLI build phase, and for server-rendered routes its DataSource is resolved on the server.
Effect's/Cmd's are never executed on the CLI or server, they are only executed in the Browser. The data from a Route Module's
init function is used to render the initial HTML on the server or build step, but the Effect isn't executed and update is never called
before the page is hydrated in the Browser. This gives a deterministic mental model of what the first render will look like,
and a nicely typed way to define the initial Data you have to render your initial view.
Because elm-pages hydrates into a full Elm single-page app, it does need the data in order to initialize the Elm app.
So why not just get the data the old-fashioned way, with elm/http, for example?
A few reasons:
elm-pages has a build step, you know that your DataSource.Http requests succeeded, your decoders succeeded, your custom DataSource validations succeeded, and everything went smoothly. If something went wrong, you get a build failure and can deal with the issues before the site goes live. That means your users won't see those errors, and as a developer you don't need to handle those error cases in your code! Think of it as "parse, don't validate", but for your entire build. In the case of server-rendered routes, a DataSource failure will render a 500 page, so more care needs to be taken to make sure all common errors are handled properly, but the tradeoff is that you can use DataSource's to pull in highly dynamic data and even render user-specific pages.DataSource data. Also, it will be served up extremely quickly without needing to wait for any database queries to be performed, andThen requests to be resolved, etc., because all of that work and waiting was done at build-time!You can think of a DataSource as a declarative (not imperative) definition of data. It represents where to get the data from, and how to transform it (map, combine with other DataSources, etc.).
This is very similar to Cmd's in Elm. You don't perform a Cmd just by running that code, as you might in a language like JavaScript. Instead, a Cmd will not do anything unless you pass it to The Elm Architecture to have it perform it for you.
You pass a Cmd to The Elm Architecture by returning it in init or update. So actually a Cmd is just data describing a side-effect that the Elm runtime can perform, and how to build a Msg once it's done.
DataSource's are very similar. A DataSource doesn't do anything just by "running" it. Just like a Cmd, it's only data that describes a side-effect to perform. Specifically, it describes a side-effect that the elm-pages runtime can perform.
There are a few places where we can pass a DataSource to the elm-pages runtime so it can perform it. Most commonly, you give a field called data in your Route Module's definition. Instead of giving a Msg when the side-effects are complete,
the page will render once all of the side-effects have run and all the data is resolved. elm-pages makes the resolved data available your Route Module's init, view, update, and head functions, similar to how a regular Elm app passes Msg's in
to update.
Any place in your elm-pages app where the framework lets you pass in a value of type DataSource is a place where you can give elm-pages a DataSource to perform (for example, Site.head where you define global head tags for your site).
Transform a request into an arbitrary value. The same underlying HTTP requests will be performed during the build step, but mapping allows you to change the resulting values by applying functions to the results.
A common use for this is to map your data into your elm-pages view:
import DataSource
import Json.Decode as Decode exposing (Decoder)
view =
DataSource.Http.get
(Secrets.succeed "https://api.github.com/repos/dillonkearns/elm-pages")
(Decode.field "stargazers_count" Decode.int)
|> DataSource.map
(\stars ->
{ view =
\model viewForPage ->
{ title = "Current stars: " ++ String.fromInt stars
, body = Html.text <| "⭐️ " ++ String.fromInt stars
, head = []
}
}
)This is useful for prototyping with some hardcoded data, or for having a view that doesn't have any StaticHttp data.
import DataSource
view :
List ( PagePath, Metadata )
->
{ path : PagePath
, frontmatter : Metadata
}
->
StaticHttp.Request
{ view : Model -> View -> { title : String, body : Html Msg }
, head : List (Head.Tag Pages.PathKey)
}
view siteMetadata page =
StaticHttp.succeed
{ view =
\model viewForPage ->
mainView model viewForPage
, head = head page.frontmatter
}Stop the StaticHttp chain with the given error message. If you reach a fail in your request,
you will get a build error. Or in the dev server, you will see the error message in an overlay in your browser (and in
the terminal).
Turn an Err into a DataSource failure.
Build off of the response from a previous DataSource request to build a follow-up request. You can use the data
from the previous response to build up the URL, headers, etc. that you send to the subsequent request.
import DataSource
import Json.Decode as Decode exposing (Decoder)
licenseData : DataSource String
licenseData =
DataSource.Http.get
(Secrets.succeed "https://api.github.com/repos/dillonkearns/elm-pages")
(Decode.at [ "license", "url" ] Decode.string)
|> DataSource.andThen
(\licenseUrl ->
DataSource.Http.get (Secrets.succeed licenseUrl) (Decode.field "description" Decode.string)
)Helper to remove an inner layer of Request wrapping.
Turn a list of StaticHttp.Requests into a single one.
import DataSource
import Json.Decode as Decode exposing (Decoder)
type alias Pokemon =
{ name : String
, sprite : String
}
pokemonDetailRequest : StaticHttp.Request (List Pokemon)
pokemonDetailRequest =
StaticHttp.get
(Secrets.succeed "https://pokeapi.co/api/v2/pokemon/?limit=3")
(Decode.field "results"
(Decode.list
(Decode.map2 Tuple.pair
(Decode.field "name" Decode.string)
(Decode.field "url" Decode.string)
|> Decode.map
(\( name, url ) ->
StaticHttp.get (Secrets.succeed url)
(Decode.at
[ "sprites", "front_default" ]
Decode.string
|> Decode.map (Pokemon name)
)
)
)
)
)
|> StaticHttp.andThen StaticHttp.combineA helper for combining DataSources in pipelines.
Like map, but it takes in two Requests.
view siteMetadata page =
StaticHttp.map2
(\elmPagesStars elmMarkdownStars ->
{ view =
\model viewForPage ->
{ title = "Repo Stargazers"
, body = starsView elmPagesStars elmMarkdownStars
}
, head = head elmPagesStars elmMarkdownStars
}
)
(get
(Secrets.succeed "https://api.github.com/repos/dillonkearns/elm-pages")
(Decode.field "stargazers_count" Decode.int)
)
(get
(Secrets.succeed "https://api.github.com/repos/dillonkearns/elm-markdown")
(Decode.field "stargazers_count" Decode.int)
)
A DataSource represents data that will be gathered at build time. Multiple
DataSources can be combined together using themapNfunctions, very similar to how you can manipulate values with Json Decoders in Elm.