Ormolu.Printer.Combinators

Printing combinators. The definitions here are presented in such an order so you can just go through the Haddocks and by the end of the file you should have a pretty good idea how to program rendering logic.

The R monad

data R a

The R monad hosts combinators that allow us to describe how to render AST.

Instances

Monad R

Defined in Ormolu.Printer.Internal

Methods

(>>=) :: R a -> (a -> R b) -> R b

(>>) :: R a -> R b -> R b

return :: a -> R a

fail :: String -> R a

Functor R

Defined in Ormolu.Printer.Internal

Methods

fmap :: (a -> b) -> R a -> R b

(<$) :: a -> R b -> R a

Applicative R

Defined in Ormolu.Printer.Internal

Methods

pure :: a -> R a

(<*>) :: R (a -> b) -> R a -> R b

liftA2 :: (a -> b -> c) -> R a -> R b -> R c

(*>) :: R a -> R b -> R b

(<*) :: R a -> R b -> R a

runR

Arguments

:: R ()

Monad to run

-> SpanStream

Span stream

-> CommentStream

Comment stream

-> Anns

Annotations

-> Bool

Use Record Dot Syntax

-> Text

Resulting rendition

Run an R monad.

getAnns :: SrcSpan -> R [AnnKeywordId]

For a given span return AnnKeywordIds associated with it.

getEnclosingSpan

Arguments

:: (RealSrcSpan -> Bool)

Predicate to use

-> R (Maybe RealSrcSpan)

Get the first enclosing SrcSpan that satisfies given predicate.

Combinators

Basic

txt

Arguments

:: Text

Text to output

-> R ()

Output a fixed Text fragment. The argument may not contain any line breaks. txt is used to output all sorts of “fixed” bits of syntax like keywords and pipes | in functional dependencies.

To separate various bits of syntax with white space use space instead of txt " ". To output Outputable Haskell entities like numbers use atom.

atom :: Outputable a => a -> R ()

Output Outputable fragment of AST. This can be used to output numeric literals and similar. Everything that doesn't have inner structure but does have an Outputable instance.

space :: R ()

This primitive does not necessarily output a space. It just ensures that the next thing that will be printed on the same line will be separated by a single space from the previous output. Using this combinator twice results in at most one space.

In practice this design prevents trailing white space and makes it hard to output more than one delimiting space in a row, which is what we usually want.

newline :: R ()

Output a newline. First time newline is used after some non-newline output it gets inserted immediately. Second use of newline does not output anything but makes sure that the next non-white space output will be prefixed by a newline. Using newline more than twice in a row has no effect. Also, using newline at the very beginning has no effect, this is to avoid leading whitespace.

Similarly to space, this design prevents trailing newlines and makes it hard to output more than one blank newline in a row.

inci :: R () -> R ()

Increase indentation level by one indentation step for the inner computation. inci should be used when a part of code must be more indented relative to the parts outside of inci in order for the output to be valid Haskell. When layout is single-line there is no obvious effect, but with multi-line layout correct indentation levels matter.

located

Arguments

:: Data a

=> Located a

Thing to enter

-> (a -> R ())

How to render inner value

-> R ()

Enter a Located entity. This combinator handles outputting comments and sets layout (single-line vs multi-line) for the inner computation. Roughly, the rule for using located is that every time there is a Located wrapper, it should be “discharged” with a corresponding located invocation.

located'

Arguments

:: Data a

=> (a -> R ())

How to render inner value

-> Located a

Thing to enter

-> R ()

A version of located with arguments flipped.

locatedPat :: (Data (Pat pass), XXPat pass ~ Located (Pat pass)) => Pat pass -> (Pat pass -> R ()) -> R ()

A version of located that works on Pat.

Starting from GHC 8.8, LPat == Pat. Located Pats are always constructed with the XPat constructor, containing a Located Pat.

Most of the time, we can just use p_pat directly, because it handles located Pats. However, sometimes we want to use the location to render something other than the given Pat.

If given Pat does not contain a location, we error out.

This should become unnecessary if https://gitlab.haskell.org/ghc/ghc/issues/17330 is ever fixed.

switchLayout

Arguments

:: [SrcSpan]

Span that controls layout

-> R ()

Computation to run with changed layout

-> R ()

Set layout according to combination of given SrcSpans for a given. Use this only when you need to set layout based on e.g. combined span of several elements when there is no corresponding Located wrapper provided by GHC AST. It is relatively rare that this one is needed.

Given empty list this function will set layout to single line.

data Layout

Layout options.

Constructors

SingleLine

Put everything on single line

MultiLine

Use multiple lines

Instances

Eq Layout

Defined in Ormolu.Printer.Internal

Methods

(==) :: Layout -> Layout -> Bool

(/=) :: Layout -> Layout -> Bool

Show Layout

Defined in Ormolu.Printer.Internal

Methods

showsPrec :: Int -> Layout -> ShowS

show :: Layout -> String

showList :: [Layout] -> ShowS

vlayout

Arguments

:: R a

Single line

-> R a

Multi line

-> R a

Do one or another thing depending on current Layout.

getLayout :: R Layout

Get current Layout.

breakpoint :: R ()

Insert a space if enclosing layout is single-line, or newline if it's multiline.

breakpoint = vlayout space newline

breakpoint' :: R ()

Similar to breakpoint but outputs nothing in case of single-line layout.

breakpoint' = vlayout (return ()) newline

Formatting lists

sep

Arguments

:: R ()

Separator

-> (a -> R ())

How to render an element

-> [a]

Elements to render

-> R ()

Render a collection of elements inserting a separator between them.

sepSemi

Arguments

:: (a -> R ())

How to render an element

-> [a]

Elements to render

-> R ()

Render a collection of elements layout-sensitively using given printer, inserting semicolons if necessary and respecting useBraces and dontUseBraces combinators.

useBraces $ sepSemi txt ["foo", "bar"]
  == vlayout (txt "{ foo; bar }") (txt "foo\nbar")

dontUseBraces $ sepSemi txt ["foo", "bar"]
  == vlayout (txt "foo; bar") (txt "foo\nbar")

canUseBraces :: R Bool

Return True if we can use braces in this context.

useBraces :: R () -> R ()

Make the inner computation use braces around single-line layouts.

dontUseBraces :: R () -> R ()

Make the inner computation omit braces around single-line layouts.

Wrapping

data BracketStyle

BracketStyle controlling how closing bracket is rendered.

Constructors

N

Normal

S

Shifted one level

sitcc :: R () -> R ()

Set indentation level for the inner computation equal to current column. This makes sure that the entire inner block is uniformly "shifted" to the right. Only works (and makes sense) when enclosing layout is multi-line.

backticks :: R () -> R ()

Surround given entity by backticks.

banana :: R () -> R ()

Surround given entity by banana brackets (i.e., from arrow notation.)

braces :: BracketStyle -> R () -> R ()

Surround given entity by curly braces { and }.

brackets :: BracketStyle -> R () -> R ()

Surround given entity by square brackets [ and ].

parens :: BracketStyle -> R () -> R ()

Surround given entity by parentheses ( and ).

parensHash :: BracketStyle -> R () -> R ()

Surround given entity by (# and #).

pragmaBraces :: R () -> R ()

Braces as used for pragmas: {-# and #-}.

pragma

Arguments

:: Text

Pragma text

-> R ()

Pragma body

-> R ()

Surround the body with a pragma name and pragmaBraces.

Literals

comma :: R ()

Print ,.

Comments

data HaddockStyle

Haddock string style.

Constructors

Pipe

-- |

Caret

-- ^

Asterisk Int

-- *

Named String

-- $

setLastCommentSpan

Arguments

:: Maybe HaddockStyle

HaddockStyle or Nothing if it's a non-Haddock comment

-> RealSrcSpan

Location of last printed comment

-> R ()

Set span of last output comment.

getLastCommentSpan :: R (Maybe (Maybe HaddockStyle, RealSrcSpan))

Get span of last output comment.