PPrint, a prettier printer

The library is based on a single way to concatenate documents, which is associative and has both a left and right unit. This simple design leads to an efficient and short implementation. The simplicity is reflected in the predictable behaviour of the combinators which make them easy to use in practice.

Copyright 2000, Daan Leijen. All rights reserved.

Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:

• Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
• Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.

This software is provided by the copyright holders “as is” and any express or implied warranties, including, but not limited to, the implied warranties of merchantability and fitness for a particular purpose are disclaimed. In no event shall the copyright holders be liable for any direct, indirect, incidental, special, exemplary, or consequential damages (including, but not limited to, procurement of substitute goods or services; loss of use, data, or profits; or business interruption) however caused and on any theory of liability, whether in contract, strict liability, or tort (including negligence or otherwise) arising in any way out of the use of this software, even if advised of the possibility of such damage.

• The nil document is called empty.
• The above combinator is called <$>. The operator </> is used for soft line breaks.
• There are three new primitives: align, fill and fillBreak. These are very useful in practice.
• Lots of other useful combinators, like fillSep and list.
• There are two renderers, renderPretty for pretty printing and renderCompact for compact output. The pretty printing algorithm also uses a ribbon-width now for even prettier output.
• There are two displayers, displayS for strings and displayIO for file based output.
• There is a Pretty class.
• The implementation uses optimised representations and strictness annotations.

The concatenation operator <> is associative and has empty as a left and right unit.

x <> (y <> z)       = (x <> y) <> z
x <> empty          = x
empty <> x          = x

The text combinator is a homomorphism from string concatenation to document concatenation. The combinator char behaves like one-element text.

text (s ++ t)       = text s <> text t
text ""             = empty
char c              = text [c]

The nest combinator is a homomorphism from addition to document composition. nest also distributes through document concatenation and is absorbed by text and align.

nest (i+j) x        = nest i (nest j x)
nest 0 x            = x
nest i (x <> y)     = nest i x <> nest j y
nest i empty        = empty
nest i (text s)     = text s
nest i (align x)    = align x

The group combinator is absorbed by empty. It is commutative with nest and align.

group (empty)       = empty
group (text s <> x) = text s <> group x
group (nest i x)    = nest i (group x)
group (align x)     = align (group x)

The align combinator is absorbed by empty and text.

align (empty)       = empty
align (text s)      = text s 
align (align x)     = align x

From the laws of the primitive combinators, we can derive many other laws for the derived combinators. For example, the above operator <$> is defined as:

x <$> y             = x <> line <> y

It follows that <$> is associative and that <$> and <> associate with each other.

x <$> (y <$> z)     = (x <$> y) <$> z
x <> (y <$> z)      = (x <> y) <$> z
x <$> (y <> z)      = (x <$> y) <> z

The same laws also hold for the other line break operators </>, <$$> and <//>.

Reference guide

The abstract data type Doc represents pretty documents.

Doc is an instance of the Show class. (show doc) pretty prints document doc with a page width of 100 characters and a ribbon width of 40 characters.

show (text "hello" <$> text "world")


Which would return the string "hello\nworld", i.e.

hello
world


The action (putDoc doc) pretty prints document doc to the standard output. with a page width of 100 characters and a ribbon width of 40 characters.

main :: IO ()
main = do{ putDoc (text "hello" <+> text "world") }


Which would output

hello world


(hPutDoc handle doc) pretty prints document doc to the file handle handle with a page width of 100 characters and a ribbon width of 40 characters.

main    = do{ handle <- openFile "MyFile" WriteMode
            ; hPutDoc handle (vcat (map text 
                              ["vertical","text"]))
            ; hClose handle
            }


The empty document is, indeed, empty. Allthough empty has no content, it does have a 'height' of 1 and behaves exactly like (text "") (and is therefore not a unit of <$>).

The document (char c) contains the literal character c. The character shouldn't be a newline ('\n'), the function line should be used for line breaks.

The document (text s) contains the literal string s. The string shouldn't contain any newline ('\n') characters. If the string contains newline characters, the function string should be used.

The document (x <> y) concatenates document x and document y. It is an associative operation having empty as a left and right unit.

The document (nest i x) renders document x with the current indentation level increased by i (See also hang, align and indent).

nest 2 (text "hello" <$> text "world") <$> text "!"


outputs as:

hello
  world
!


The line document advances to the next line and indents to the current nesting level. Document line behaves like (text " ") if the line break is undone by group.

The linebreak document advances to the next line and indents to the current nesting level. Document linebreak behaves like empty if the line break is undone by group.

The group combinator is used to specify alternative layouts. The document (group x) undoes all line breaks in document x. The resulting line is added to the current line if that fits the page. Otherwise, the document x is rendered without any changes.

The document softline behaves like space if the resulting output fits the page, otherwise it behaves like line.

softline  = group line


The document softbreak behaves like empty if the resulting output fits the page, otherwise it behaves like line.

softbreak  = group linebreak


The combinators in this section can not be descibed by Wadler's original combinators. They align their output relative to the current output position – in contrast to nest which always aligns to the current nesting level. This deprives these combinators from being `optimal'. In practice however they prove to be very useful. The combinators in this section should be used with care, since they are more expensive than the other combinators. For example, align shouldn't be used to pretty print all top-level declarations of a language, but using hang for let expressions is fine.

The document (align x) renders document x with the nesting level set to the current column. It is used for example to implement hang.

As an example, we will put a document right above another one, regardless of the current nesting level:

x $$ y  = align (x <$> y)

test    = text "hi" <+> (text "nice" $$ text "world")


which will be layed out as:

hi nice
   world


The hang combinator implements hanging indentation. The document (hang i x) renders document x with a nesting level set to the current column plus i. The following example uses hanging indentation for some text:

test  = hang 4 (fillSep (map text 
        (words "the hang combinator indents these words !")))


Which lays out on a page with a width of 20 characters as:

the hang combinator
    indents these
    words !


The hang combinator is implemented as:

hang i x  = align (nest i x)


The document (indent i x) indents document x with i spaces.

test  = indent 4 (fillSep (map text 
        (words "the indent combinator indents these words !")))


Which lays out with a page width of 20 as:

    the indent 
    combinator 
    indents these
    words !


The document (encloseSep l r sep xs) concatenates the documents xs seperated by sep and encloses the resulting document by l and r. The documents are rendered horizontally if that fits the page. Otherwise they are aligned vertically. All seperators are put in front of the elements. For example, the combinator list can be defined with encloseSep:

list xs  = encloseSep lbracket rbracket comma xs
test     = text "list" <+> (list (map int [10,200,3000]))


Which is layed out with a page width of 20 as:

list [10,200,3000]


But when the page width is 15, it is layed out as:

list [10
     ,200
     ,3000]


The document (list xs) comma seperates the documents xs and encloses them in square brackets. The documents are rendered horizontally if that fits the page. Otherwise they are aligned vertically. All comma seperators are put in front of the elements.

The document (tupled xs) comma seperates the documents xs and encloses them in parenthesis. The documents are rendered horizontally if that fits the page. Otherwise they are aligned vertically. All comma seperators are put in front of the elements.

The document (semiBraces xs) seperates the documents xs with semi colons and encloses them in braces. The documents are rendered horizontally if that fits the page. Otherwise they are aligned vertically. All semi colons are put in front of the elements.

The document (x <+> y) concatenates document x and y with a space in between.

The document (x <$> y) concatenates document x and y with a line in between.

The document (x </> y) concatenates document x and y with a softline in between. This effectively puts x and y either next to each other (with a space in between) or underneath each other.

The document (x <$$> y) concatenates document x and y with a linebreak in between.

The document (x <//> y) concatenates document x and y with a softbreak in between. This effectively puts x and y either right next to each other or underneath each other.

The document (hsep xs) concatenates all documents xs horizontally with (<+>).

The document (vsep xs) concatenates all documents xs vertically with (<$>). If a group undoes the line breaks inserted by vsep, all documents are seperated with a space.

someText = map text (words ("text to lay out"))

test     = text "some" <+> vsep someText


This is layed out as:

some text
to
lay
out


The align combinator can be used to align the documents under their first element

test     = text "some" <+> align (vsep someText)


Which is printed as:

some text
     to
     lay
     out


The document (fillSep xs) concatenates documents xs horizontally with (<+>) as long as its fits the page, than inserts a line and continues doing that for all documents in xs.

fillSep xs  = foldr (</>) empty xs


The document (sep xs) concatenates all documents xs either horizontally with (<+>), if it fits the page, or vertically with (<$>).

sep xs  = group (vsep xs)


The document (hcat xs) concatenates all documents xs horizontally with (<>).

The document (vcat xs) concatenates all documents xs vertically with (<$$>). If a group undoes the line breaks inserted by vcat, all documents are directly concatenated.

The document (fillCat xs) concatenates documents xs horizontally with (<>) as long as its fits the page, than inserts a linebreak and continues doing that for all documents in xs.

fillCat xs  = foldr (<//>) empty xs


The document (cat xs) concatenates all documents xs either horizontally with (<>), if it fits the page, or vertically with (<$$>).

cat xs  = group (vcat xs)


(punctuate p xs) concatenates all in documents xs with document p except for the last document.

someText = map text ["words","in","a","tuple"]
test     = parens (align (cat (punctuate comma someText)))


This is layed out on a page width of 20 as:

(words,in,a,tuple)


But when the page width is 15, it is layed out as:

(words,
 in,
 a,
 tuple)


(If you want put the commas in front of their elements instead of at the end, you should use tupled or, in general, encloseSep.)

The document (fill i x) renders document x. It than appends spaces until the width is equal to i. If the width of x is already larger, nothing is appended. This combinator is quite useful in practice to output a list of bindings. The following example demonstrates this.

types  = [("empty","Doc")
         ,("nest","Int -> Doc -> Doc")
         ,("linebreak","Doc")]

ptype (name,tp)
       = fill 6 (text name) <+> text "::" <+> text tp

test   = text "let" <+> align (vcat (map ptype types))


Which is layed out as:

let empty  :: Doc
    nest   :: Int -> Doc -> Doc
    linebreak :: Doc


The document (fillBreak i x) first renders document x. It than appends spaces untill the width is equal to i. If the width of x is already larger than i, the nesting level is increased by i and a line is appended. When we redefine ptype in the previous example to use fillBreak, we get a useful variation of the previous output:

ptype (name,tp)
       = fillBreak 6 (text name) <+> text "::" <+> text tp


The output will now be:

let empty  :: Doc
    nest   :: Int -> Doc -> Doc
    linebreak
           :: Doc


The document (enclose l r x) encloses document x between documents l and r using (<>).

enclose l r x   = l <> x <> r


Document (squotes x) encloses document x with single quotes "'".

Document (dquotes x) encloses document x with double quotes '"'.

Document (parens x) encloses document x in parenthesis, "(" and ")".

Document (angles x) encloses document x in angles, "<" and ">".

Document (braces x) encloses document x in braces, "{" and "}".

Document (brackets x) encloses document x in square brackets, "[" and "]".

The document lparen contains a left parenthesis, "(".

The document rparen contains a right parenthesis, ")".

The document langle contains a left angle, "<".

The document rangle contains a right angle, ">".

The document lbrace contains a left brace, "{".

The document rbrace contains a right brace, "}".

The document lbracket contains a left square bracket, "[".

The document rbracket contains a right square bracket, "]".

The document squote contains a single quote, "'".

The document dquote contains a double quote, '"'.

The document semi contains a semi colon, ";".

The document colon contains a colon, ":".

The document comma contains a comma, ",".

The document space contains a single space, "" "".

x <+> y   = x <> space <> y


The document dot contains a single dot, ".".

The document backslash contains a back slash, "\".

The document equals contains an equal sign, "=".

The document (string s) concatenates all characters in s using line for newline characters and char for all other characters. It is used instead of text whenever the text contains newline characters.

The document (int i) shows the literal integer i using text.

The document (integer i) shows the literal integer i using text.

The document (float f) shows the literal float f using text.

The document (double d) shows the literal double d using text.

The document (rational r) shows the literal rational r using text.

The class Pretty has two members:

class Pretty a where
  pretty     :: a -> Doc
  prettyList :: [a] -> Doc


The member prettyList is only used to define the instance Pretty a => Pretty [a]. In normal circumstances only the pretty function is used. Library defined instances of Pretty are (), Char, Int, Integer, Float, Double, Rational, Pretty a => Maybe a, Pretty a => [a], Pretty a,Pretty b => (a,b) and Pretty a,Pretty b,Pretty c => (a,b,c).

The data type SimpleDoc represents rendered documents and is used by the display functions.

data SimpleDoc  = SEmpty
                | SChar Char SimpleDoc
                | SText !Int String SimpleDoc
                | SLine !Int SimpleDoc


The Int in SText contains the length of the string. The Int in SLine contains the indentation for that line. The library provides two default display functions displayS and displayIO. You can provide your own display function by writing a function from a SimpleDoc to your own output format.

This is the default pretty printer which is used by show, putDoc and hPutDoc. (renderPretty ribbonfrac width x) renders document x with a page width of width and a ribbon width of (ribbonfrac * width) characters. The ribbon width is the maximal amount of non-indentation characters on a line. The parameter ribbonfrac should be between 0.0 and 1.0. If it is lower or higher, the ribbon width will be 0 or width respectively.

(renderCompact x) renders document x without adding any indentation. Since no 'pretty' printing is involved, this renderer is very fast. The resulting output contains fewer characters as a pretty printed version and can be used for output that is read by other programs.

(displayS simpleDoc) takes the output simpleDoc from a rendering function and transforms it to a ShowS type (for use in the Show class).

showWidth :: Int -> Doc -> String
showWidth w x   = displayS (renderPretty 0.4 w x) ""


(displayIO handle simpleDoc) writes simpleDoc to the file handle handle. This function is used for example by hPutDoc:

hPutDoc handle doc  = displayIO handle (renderPretty 0.4 100 doc)


References

1. John Hughes. (1995)
   The design of a pretty-printer library.
   In J. Jeuring and E. Meijer, editors, Advanced Functional Programming, Springer Verlag LNCS 925.
   http://www.cs.chalmers.se/~rjmh/Papers/pretty.ps.
2. Simon Peyton Jones. (1997)
   A Haskell pretty-printer library.
   
   http://www.research.microsoft.com/~simonpj/Haskell/pretty.html.
3. Doaitse Swierstra, Pablo Azero Alcocer and Joao Saraiva. (1998)
   Designing and Implementing Combinator Languages.
   Advanced Functional Programming, Springer-Verlag, LNCS 1608:150-206.
   http://www.cs.uu.nl/groups/ST/Software.
4. Philip Wadler. (April 1997, revised March 1998)
   A prettier printer.
   Draft paper.
   http://cm.bell-labs.com/cm/cs/who/wadler/topics/language-design.html.