Code formatting

This page demonstrates how to use FSharp.Formatting.CodeFormat to tokenize F# source code, obtain information about the source code (mainly tooltips from the type-checker) and how to turn the code into a nicely formatted HTML.

First, we need to load the assembly and open the necessary namespaces:

open FSharp.Formatting.CodeFormat
open System.Reflection

If you want to process multiple snippets, it is a good idea to keep the formatting agent around if possible. The agent needs to load the F# compiler (which needs to load various files itself) and so this takes a long time.

Processing F# source

The formatting agent provides a CodeFormatAgent.ParseAndCheckSource method (together with an asynchronous version for use from F# and also a version that returns a .NET Task for C#). To call the method, we define a simple F# code as a string:

let source =
    """
    let hello () =
      printfn "Hello world"
  """

let snippets, diagnostics =
    CodeFormatter.ParseAndCheckSource("C:\\snippet.fsx", source, None, None, ignore)

When calling the method, you need to specify a file name and the actual content of the script file. The file does not have to physically exist. It is used by the F# compiler to resolve relative references (e.g. #r) and to automatically name the module including all code in the file.

You can also specify additional parameters, such as *.dll references, by passing a third argument with compiler options (e.g. "-r:Foo.dll -r:Bar.dll").

This operation might take some time, so it is a good idea to use an asynchronous variant of the method. It returns two arrays - the first contains F# snippets in the source code and the second contains any errors reported by the compiler. A single source file can include multiple snippets using the same formatting tags as those used on fssnip.net as documented in the about page.

Working with returned tokens

Each returned snippet is essentially just a collection of lines, and each line consists of a sequence of tokens. The following snippet prints basic information about the tokens of our sample snippet:

// Get the first snippet and obtain list of lines
let (Snippet (title, lines)) = snippets |> Seq.head

// Iterate over all lines and all tokens on each line
for (Line (_, tokens)) in lines do
    for token in tokens do
        match token with
        | TokenSpan.Token (kind, code, tip) ->
            printf "%s" code

            tip
            |> Option.iter (fun spans -> printfn "%A" spans)
        | TokenSpan.Omitted _
        | TokenSpan.Output _
        | TokenSpan.Error _ -> ()

    printfn ""

The TokenSpan.Token is the most important kind of token. It consists of a kind (identifier, keyword, etc.), the original F# code and tooltip information. The tooltip is further formatted using a simple document format, but we simply print the value using the F# pretty printing, so the result looks as follows:

let hello[Literal "val hello : unit -> unit"; ...] () =
  printfn[Literal "val printfn : TextWriterFormat<'T> -> 'T"; ...] "Hello world"

The Omitted token is generated if you use the special (*[omit:...]*) command. The Output token is generated if you use the // [fsi:...] command to format output returned by F# interactive. The Error command wraps code that should be underlined with a red squiggle if the code contains an error.

Generating HTML output

Finally, the CodeFormat type also includes a method CodeFormat.FormatHtml that can be used to generate nice HTML output from an F# snippet. This is used, for example, on F# Snippets. The following example shows how to call it:

let prefix = "fst"
let html = CodeFormat.FormatHtml(snippets, prefix)

// Print all snippets, in case there is more of them
for snip in html.Snippets do
    printfn "%s" snip.Content

// Print HTML code that is generated for ToolTips
printfn "%s" html.ToolTip

If the input contains multiple snippets separated using the //[snippet:...] comment, e.g.:

1: 
2: 
3: 
4: 
5: 
6: 
7: 

// [snippet: First sample]
printf "The answer is: %A" 42
// [/snippet]
// [snippet: Second sample]
printf "Hello world!"
// [/snippet]

then the formatter returns multiple HTML blocks. However, the generated tooltips are shared by all snippets (to save space) and so they are returned separately.