Simple things in FSharp
Hello there, this is the 4th entry in Simple Things F#.
Today we'll be talking about producing HTML strings (useful to create reports, pdfs and things like those). Producing HTML in F# is quite simple to be honest, there are a bunch of libraries that allow that we'll be looking at some I've used at some point
The first two are F# DSL's to build HTML, the last one is an HTML scripting language for .NET (there's also Giraffe.Razor but I'll skip that one for today)
Giraffe
When it comes to HTML DSL's this is the most "traditional" one since it's not a flavor liked my many these days but that doesn't take away it's usefulness. Giraffe View Engine uses XmlNode as it's building block so you can produce XML as well with Giraffe View Engine! (points for that)
the structure of a node function in Giraffe is as follows
tagName [(* attribute list *)] [(* node list *)]
for example creating <div class="my-class"></div>
would be something like this
div [ _class "my-class" ] []
attributes are prefixed with an underscore
_
to prevent clashing with reserved words in F#
Let's take a look at a simple page with a header
#r "nuget: Giraffe.ViewEngine"
open Giraffe.ViewEngine
let view =
html [] [
head [] [ title [] [ str "Giraffe" ] ]
body [] [ header [] [ str "Giraffe" ] ]
]
let document = RenderView.AsString.htmlDocument view
printfn "%s" document
To Run this, copy this content into a file named
script.fsx
(or whatever name you prefer) and type:
dotnet fsi run script.fsx
we assign the result of the html function to view, then we just render the document as a string, we can then follow up and create a file with the IO API's we saw earlier in this series.
If you're using Saturn/Giraffe as your web framework you don't need to do the manual render they provide helpers that take care of this (as you'll see on the next entry in this series)
you can also create functions to pre-define aspects of your views and override values if you deem it necessary
#r "nuget: Giraffe.ViewEngine"
open Giraffe.ViewEngine
let card attributes =
article [ yield! attributes; _class "card is-green"]
let cardFooter attributes =
footer [ yield! attributes; _class "card-footer is-rounded"]
let cardHeader attributes =
header [ yield! attributes; _class "card-header no-icons"]
let mySection =
div [] [
card [ _id "my-card" ] [
cardHeader [] [
h1 [] [ str "This is my custom card"]
img [ _src "https://some-image.com"; _class "card-header-image" ]
]
p [] [ str "this is the body of the card" ]
cardFooter [ _data "my-attr" "extra attributes" ] [
p [] [ str "This is my footer"]
]
]
]
// note that we used htmlNode this time instead of htmlDocument
let document = RenderView.AsString.htmlNode mySection
printfn "%s" document
To Run this, copy this content into a file named
script.fsx
(or whatever name you prefer) and type:
dotnet fsi run script.fsx
so creating new "tags" is basically just creating a new function that accepts attributes and contents.
Feliz
The original Feliz DSL was made by @zaidajaj (he produces AMAZING OSS work for F# you should check his github profile) to use in Fable applications which are built most of the time on top of React.js so if you take a look at the following content you'll see that view
is a type of ReactElement
but don't worry it's just a type, there's no javascript running here it's just a way to make use of a really good DSL (which reduces typing and improves readability) to produce HTML as well in the backend.
I'm not ultra well versed in composition techniques with Feliz so take the following examples with a grain of salt (you can also check the Elmish book https://zaid-ajaj.github.io/the-elmish-book/) for more information about Fable and Feliz
#r "nuget: Feliz.ViewEngine"
open Feliz.ViewEngine
let view =
Html.html [
Html.head [ Html.title "Feliz" ]
Html.body [
Html.header [ prop.text "Feliz" ]
]
]
let document = Render.htmlDocument view
printfn "%s" document
To Run this, copy this content into a file named
script.fsx
(or whatever name you prefer) and type:
dotnet fsi run script.fsx
when you open the Feliz.ViewEngine
namespace you have access to the Html
and prop
types these have all the tags and attributes you may need to construct HTML content. if you are feeling tired of writing prop
and Html
you can even open type
those classes and access all of their static methods
#r "nuget: Feliz.ViewEngine"
open Feliz.ViewEngine
open type Html
open type prop
let view =
html [
head [ title "Feliz" ]
body [
header [ text "Feliz" ]
]
]
let document = Render.htmlDocument view
printfn "%s" document
To Run this, copy this content into a file named
script.fsx
(or whatever name you prefer) and type:
dotnet fsi run script.fsx
that will save you a few keystrokes as well! you might run into some naming clashes but you can also just use qualified prop
and Html
if you need it.
let's continue with the card component to see how you can compose different view elements. There are a few differences when you use the Feliz DSL, primarily these are because how the React API is defined, you can't mix properties with children, so you have to pass prop.children
to be able to define child elements for your components
#r "nuget: Feliz.ViewEngine"
open Feliz.ViewEngine
open type Html
open type prop
// custom card, but you can't customize it's classes only the children elements
let card (content: ReactElement seq) =
article [
className "card is-green"
children content
]
// this card footer will allou you to define any property of the footer element
let cardFooter content =
footer [
className "card-footer is-rounded"
yield! content
]
let slotedHeader (content: ReactElement seq) =
header [
className "card-header"
// pass the contents directly to the children property
children content
]
let customizableHeader content =
header [
className "card-header"
// allow any property to be set
yield! content
]
let card1 =
div [
card [
// our slottedHeader only allows to pass children not props
slotedHeader [
h1 [ text "This is my custom card"]
// className "" <- can't do this
]
p [ text "this is the body of the card" ]
cardFooter [
custom("data-my-attr", "extra attributes")
children (p [text "This is my footer"])
]
]
]
let card2 =
div [
card [
/// our customizable header allows us
/// to pass properties as well as children elements
customizableHeader [
children (h1 [ text "This is my custom card"])
className "custom class"
]
p [ text "this is the body of the card" ]
cardFooter [
custom("data-my-attr", "extra attributes")
children (p [text "This is my footer"])
]
]
]
let r1 = Render.htmlView card1
let r2 = Render.htmlView card2
printfn "%s\n\n%s" r1 r2
To Run this, copy this content into a file named
script.fsx
(or whatever name you prefer) and type:
dotnet fsi run script.fsx
But there's a thing that will happen here, unlike Giraffe.ViewEngine Feliz doesn't strip existing props so our header will end up like this
<header class="card-header" class="custom class">
<h1>This is my custom card</h1>
</header>
In HTML the last defined property always wins, so keep in mind that depending on your intent you might need properties or children ant that might be the deciding factor between using one way or the other.
Scriban
If you like me can't simply just leave HTML because of reasons there are also Text based alternatives like Scriban which allow you to just write an html file and just fill it with data
#r "nuget: Scriban"
open Scriban
type Product = { name: string; price: float; description: string }
let renderProducts products =
let html =
"""
<ul id='products'>
{{ for product in products }}
<li>
<h2>{{ product.name }}</h2>
Price: {{ product.price }}
{{ product.description | string.truncate 15 }}
</li>
{{ end }}
</ul>
"""
let result = Template.Parse(html)
result.Render({| products = products |})
let result =
renderProducts [
{ name = "Shoes"; price = 20.50; description = "The most shoes you'll ever see"}
{ name = "Potatoes"; price = 1.50; description = "The most potato you'll ever see" }
{ name = "Cars"; price = 10.3; description = "The most car you'll ever see" }
]
printfn "%s" result
To Run this, copy this content into a file named
script.fsx
(or whatever name you prefer) and type:
dotnet fsi run script.fsx
If you have ever used jinja, moustache, handlebars, liquid and similar template engines it will look familiar to you basically here we define an HTML string (which can be read from an HTML file in disk) we parse it, then render it with a data source (in case we need one)
scriban has a TON of helpers in it's scripting language (like those pipes that are truncating a string to 15 characters)
If you want to compose components in the scriban templates the approach is way way different
#r "nuget: Scriban"
open System
open Scriban
type Product =
{ name: string;
price: float;
details : {| description: string |} }
// create a fragment/component html string
let detailDiv =
"""
<details>
<summary> {{ product.details.description | string.truncate 15 }} <summary>
{{ product.details.description }}
</details>
"""
let renderProducts products =
let html =
// here with the help of sprintf
// and {{ "%s" | object.eval_template }}
// we use F# to pre-process the template
sprintf
"""
<ul id='products'>
{{ for product in products }}
<li>
<h2>{{ product.name }}</h2>
Price: {{ product.price }}
{{ "%s" | object.eval_template }}
</li>
{{ end }}
</ul>
"""
detailDiv
let result = Template.Parse(html)
result.Render({| products = products |})
let result =
renderProducts [
{ name = "Shoes"
price = 20.50
details =
{| description = "The most shoes you'll ever see" |} }
{ name = "Potatoes"
price = 1.50
details =
{| description = "The most potato you'll ever see" |} }
{ name = "Cars"
price = 10.3
details =
{| description = "The most car you'll ever see" |} }
]
printfn "%s" result
To Run this, copy this content into a file named
script.fsx
(or whatever name you prefer) and type:
dotnet fsi run script.fsx
now, keep in mind that you're using F# and this you can modify the strings before passing them to the final template before parsing it but this leads to handling strings here and there and possibly getting into regex territory and to be honest I don't like that. I think this approach is best suited to templates you already set and know what the model for them is and that they are not super dynamic, you can still perform a lot of dynamic operations with the scriban scripting capabilities inside the template.
also keep in mind that if you really need it, you can parse and compile multiple HTML templates and then pass them together to a layout and just do a final render but I'm not aware of how performant/useful that is in practice
Closing Thoughts
So that's it generating HTML isn't complex and it might be useful for you, perhaps you want to build a resume from JSON to HTML, perhaps you need to build a report from the last year sales or another similar use case.
As always feel free to ping me on Twitter or the comments below 😁