/usr/share/doc/camlp5/html/opretty.html

<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.1//EN"
 "http://www.w3.org/TR/xhtml11/DTD/xhtml11.dtd">
<html xmlns="http://www.w3.org/1999/xhtml">
<head>
  <!-- opretty.html,v -->
  <!-- Copyright (c) INRIA 2007-2014 -->
  <title>Extensions of printing</title>
  <meta http-equiv="Content-Type" content="text/html; charset=utf-8" />
  <meta http-equiv="Content-Style-Type" content="text/css" />
  <link rel="stylesheet" type="text/css" href="styles/base.css"
        title="Normal" />
  <link rel="alternate" type="application/rss+xml" href="rss/camlp5.rss" 
        title="Camlp5"/>
</head>
<body>

<div id="menu">
  <h1>- <a href="http://pauillac.inria.fr/~ddr/camlp5">Camlp5</a> -</h1>
  <p class="subtitle">Version 6.16</p>
  <ul>
    <li><a href="index.html">Introduction</a></li>
    <li><a href="strict.html">Transitional and Strict</a></li>
    <li><a href="ptools.html">Parsing and printing tools</a></li>
  </ul>
  <ul>
    <li>Parsing tools
      <ul>
        <li><a href="parsers.html">Stream parsers</a></li>
        <li><a href="lexers.html">Stream lexers</a></li>
        <li><a href="fparsers.html">Functional parsers</a></li>
        <li><a href="bparsers.html">Backtracking parsers</a></li>
        <li><a href="grammars.html">Extensible grammars</a></li>
      </ul>
    </li>
    <li>Printing tools
      <ul>
        <li><a href="printers.html">Extensible printers</a></li>
        <li><a href="pprintf.html">Pprintf</a></li>
        <li><a href="pretty.html">Pretty print</a></li>
      </ul>
    </li>
    <li>Language extensions
      <ul>
        <li><a href="locations.html">Locations</a></li>
        <li><a href="ml_ast.html">Syntax tree</a></li>
        <li><a href="ast_transi.html">Syntax tree - transi</a></li>
        <li><a href="ast_strict.html">Syntax tree - strict</a></li>
        <li><a href="q_ast.html">Quotation kit q_ast.cmo</a></li>
        <li><a href="pcaml.html">The Pcaml module</a></li>
        <li><a href="directives.html">Directives</a></li>
        <li><a href="syntext.html">Extensions of syntax</a></li>
        <li><a href="opretty.html">Extensions of printing</a></li>
        <li><a href="redef.html">Redefining OCaml syntax</a></li>
        <li><a href="quot.html">Quotations</a></li>
        <li><a href="revsynt.html">Revised syntax</a></li>
        <li><a href="scheme.html">Scheme syntax</a></li>
        <li><a href="macros.html">Macros</a></li>
        <li><a href="pragma.html">Pragma directive</a></li>
        <li><a href="extfun.html">Extensible functions</a></li>
      </ul>
    </li>
    <li>Appendix
      <ul>
        <li><a href="commands.html">Commands and Files</a></li>
        <li><a href="library.html">Library</a></li>
        <li><a href="sources.html">Camlp5 sources</a></li>
        <li><a href="about.html">About Camlp5</a></li>
      </ul>
    </li>
  </ul>
</div>

<div id="content">

<h1 class="top">Extensions of printing</h1>

<p>Camlp5 provides extensions kits to pretty print programs in revised
  syntax and normal syntax. Some other extensions kits also allow to
  rebuild the parsers, or the EXTEND statements in their initial
  syntax. The pretty print system is itself extensible, by adding new
  rules. We present here how it works in the Camlp5 sources.</p>

<p>The pretty print system of Camlp5 uses the library modules
  <a href="pretty.html">Pretty</a>, an original system to format
  output) and <a href="extfun.html">Extfun</a>, another original
  system of extensible functions.</p>

<p>This chapter is designed for programmers that want to understand
  how the pretty printing of OCaml programs work in Camlp5, want to
  adapt, modify or debug it, or want to add their own pretty printing
  extensions.</p>

<div id="tableofcontents">
  <ol>
    <li><a href="#a:Introduction">Introduction</a></li>
    <li><a href="#a:Principles">Principles</a>
      <ul>
        <li><a href="#b:Using-module-Pretty">Using module Pretty</a></li>
        <li><a href="#b:Using-EXTEND_PRINTER-statement">Using EXTEND_PRINTER statement</a></li>
        <li><a href="#b:Dangling-else--bar--semicolon">Dangling else, bar, semicolon</a></li>
        <li><a href="#b:By-level">By level</a></li>
      </ul>
    </li>
    <li><a href="#a:The-Prtools-module">The Prtools module</a>
      <ul>
        <li><a href="#b:Comments">Comments</a></li>
        <li><a href="#b:Meta-functions-for-lists">Meta functions for lists</a></li>
        <li><a href="#b:Miscellaneous">Miscellaneous</a></li>
      </ul>
    </li>
    <li><a href="#a:Example-:-repeat--until">Example : repeat..until</a>
      <ul>
        <li><a href="#b:The-code">The code</a></li>
        <li><a href="#b:Compilation">Compilation</a></li>
        <li><a href="#b:Testing">Testing</a></li>
      </ul>
    </li>
  </ol>
</div>

<h2 id="a:Introduction">Introduction</h2>

<p>The files doing the pretty printings are located in Camlp5 sources
in the directory "etc". Peruse them if you are interested in creating
new ones. The main ones are:</p>

<ul>
  <li>"etc/pr_r.ml": pretty print in revised syntax.</li>
  <li>"etc/pr_o.ml": pretty print in normal syntax.</li>
  <li>"etc/pr_rp.ml": rebuilding parsers in their original revised
    syntax.</li>
  <li>"etc/pr_op.ml": rebuilding parsers in their original normal
    syntax.</li>
  <li>"etc/pr_extend.ml": rebuilding EXTEND in its original
    syntax.</li>
</ul>

<p>We present here how this system works inside these files. First,
  the general principles. Second, more details of the
  implementation.</p>

<h2 id="a:Principles">Principles</h2>

<h3 id="b:Using-module-Pretty">Using module Pretty</h3>

<p>All functions in OCaml pretty printing take a parameter named "the
printing context" (variable <code>pc</code>). It is a record holding :</p>

<ul>
  <li>The current indendation : <code>pc.ind</code></li>
  <li>What should be printed before, in the same line
  : <code>pc.bef</code></li>
  <li>What should be printed after, in the same line
    : <code>pc.aft</code></li>
  <li>The dangling token, useful in normal syntax to know whether
    parentheses are necessary : <code>pc.dang</code></li>
</ul>

<p>A typical pretty printing function calls the
  function <code>horiz_vertic</code> of the library
  module <a href="pretty.html">Pretty</a>. This function takes two
  functions as parameter:</p>

<ul>
  <li>The way to print the data in one only line (<i>horizontal</i>
    printing)</li>
  <li>The way to print the data in two or more lines (<i>vertical</i>
    printing)</li>
</ul>

<p>Both functions catenate the strings by using the
  function <code>sprintf</code> of the library
  module <code>Pretty</code> which controls whether the printed data
  holds in the line or not. They generally call, recursively, other
  pretty printing functions with the same behaviour.</p>

<p>Let us see an example (fictitious) of printing an OCaml
  application. Let us suppose we have an application expression
  "<code>e1 e2</code>" to pretty print where <code>e1</code>
  and <code>e2</code> are sub-expressions. If both expressions and
  their application holds on one only line, we want to see:</p>

<pre>
  e1 e2
</pre>

<p>On the other hand, if they do not hold on one only line, we want to
  see <code>e2</code> in another line with, say, an indendation of 2
  spaces:</p>

<pre>
  e1
    e2
</pre>

<p>Here is a possible implementation. The function has been
named <code>expr_app</code> and can call the function <code>expr</code> to
print the sub-expressions <code>e1</code> and <code>e2</code>:</p>

<pre>
  value expr_app pc e1 e2 =
    horiz_vertic
      (fun () ->
         let s1 = expr {(pc) with aft = ""} e1 in
         let s2 = expr {(pc) with bef = ""} e2 in
         sprintf "%s %s" s1 s2)
      (fun () ->
         let s1 = expr {(pc) with aft = ""} e1 in
         let s2 =
           expr
             {(pc) with
                ind = pc.ind + 2;
                bef = tab (pc.ind + 2)}
             e2
         in
         sprintf "%s\n%s" s1 s2)
  ;
</pre>

<p>The first function is the <i>horizontal</i> printing. It ends with a
  <code>sprintf</code> separating the printing of <code>e1</code>
  and <code>e2</code> by a space. The possible "before part"
  (<code>pc.bef</code>) and "after part" (<code>pc.aft</code>) are
  transmitted in the calls of the sub-functions.</p>

<p>The second function is the <i>vertical</i> printing. It ends with a
  <code>sprintf</code> separating the printing of <code>e1</code>
  and <code>e2</code> by a newline. The second line starts with an
  indendation, using the "before part" (<code>pc.bef</code>) of the
  second call to <code>expr</code>.</p>

<p>The pretty printing library
  function <code>Pretty.horiz_vertic</code> calls the first
  (<i>horizontal</i>) function, and if it fails (either
  because <code>s1</code> or <code>s2</code> are too long or hold
  newlines, or because the final string produced
  by <code>sprintf</code> is too long), calls the second
  (<i>vertical</i>) function.</p>

<p>Notice that the parameter <code>pc</code> contains a
  field <code>pc.bef</code> (what should be printed before in the same
  line), which in both cases is transmitted to the printing
  of <code>e1</code> (since the syntax <code>{(pc) with aft =
  ""}</code> is a record with <code>pc.bef</code> kept). Same for the
  field <code>pc.aft</code> transmitted to the printing
  of <code>e2</code>.</p>

<h3 id="b:Using-EXTEND_PRINTER-statement">Using EXTEND_PRINTER statement</h3>

<p>This system is combined to the <a href="printers.html">extensible
  printers</a> to allow the extensibility of the pretty printing.</p>

<p>The code above actually looks like:</p>

<pre>
  EXTEND_PRINTER
    pr_expr:
      [ [ &lt;:expr&lt; $e1$ $e2$ &gt;&gt; ->
            horiz_vertic
              (fun () ->
                 let s1 = curr {(pc) with aft = ""} e1 in
                 let s2 = next {(pc) with bef = ""} e2 in
                 sprintf "%s %s" s1 s2)
              (fun () ->
                 let s1 = curr {(pc) with aft = ""} e1 in
                 let s2 =
                   next
                     {(pc) with
                        ind = pc.ind + 2;
                        bef = tab (pc.ind + 2)}
                     e2
                 in
                 sprintf "%s\n%s" s1 s2) ] ]
    ;
  END;
</pre>

<p>The variable "<tt>pc</tt>" is implicit in the semantic actions of
  the syntax "<tt>EXTEND_PRINTER</tt>", as well as two other
  variables: "<tt>curr</tt>" and "<tt>next</tt>".</p>

<p>These parameters, "<tt>curr</tt>" and "<tt>next</tt>", correspond
  to the pretty printing of, respectively, the current level and the
  next level. Since the application in OCaml is left associative, the
  first sub-expression is printed at the same (current) level and the
  second one is printed at the next level. We also see a call
  to <code>next</code> in the last (2nd) case of the function to treat
  the other cases in the next level.</p>

<h3 id="b:Dangling-else--bar--semicolon">Dangling else, bar, semicolon</h3>

<p>In normal syntax, there are cases where it is necessary to enclose
  expressions between parentheses (or
  between <span class="parenthesis">begin</span>
  and <span class="parenthesis">end</span>, which is equivalent in
  that syntax). Three tokens may cause problems: the
  "<code>else</code>", the vertical bar "<code>|</code>" and the
  semicolon "<code>;</code>". Here are examples where the presence of
  these tokens constrains the previous expression to be
  parenthesized. In these three examples, removing
  the <span class="parenthesis">begin</span>..<span class="parenthesis">end</span>
  enclosers would change the meaning of the expression because the
  dangling token would be included in that expression:</p>

<p>Dangling else:</p>

<pre>
  if a then <span class="parenthesis">begin</span> if b then c <span class="parenthesis">end</span> <span class="dangling">else</span> d
</pre>

<p>Dangling bar:</p>

<pre>
  function
    A ->
      <span class="parenthesis">begin</span> match a with
        B -> c
      | D -> e
      <span class="parenthesis">end</span>
  <span class="dangling">|</span> F -> g
</pre>

<p>Dangling semicolon:</p>

<pre>
  if a then b
  else <span class="parenthesis">begin</span>
    let c = d in
    e
  <span class="parenthesis">end</span><span class="dangling">;</span>
  f
</pre>

<p>The information is transmitted by the
  value <code>pc.dang</code>. In the first example above, while
  displaying the "<code>then</code>" part of the outer
  "<code>if</code>", the sub-expression is called with the
  value <code>pc.dang</code> set to
  <code>"else"</code> to inform the last sub-sub-expression that it is
  going to be followed by that token. When a "<code>if</code>"
  expression should be displayed without "<code>else</code>" part, and
  that its "<code>pc.dang</code>" is "else", it should be enclosed
  with spaces.</p>

<p>This problem does not exist in revised syntax. While pretty
  printing in revised syntax, the parameter <code>pc.dang</code> is
  not necessary and remains the empty string.</p>

<h3 id="b:By-level">By level</h3>

<p>As explained in the chapter about
  the <a href="printers.html">extensible printers</a> (with the
  EXTEND_PRINTER statement), printers contain levels. The global
  printer variable of expressions is named "<tt>pr_expr</tt>" and
  contain all definitions for pretty printing expressions, organized
  by levels, just like the parsing of expressions. The definition of
  "<tt>pr_expr</tt>" actually looks like this:</p>

<pre>
  EXTEND_PRINTER
    pr_expr:
      [ "top"
        [ (* code for level "top" *) ]
      | "add"
        [ (* code for level "add" *) ]
      | "mul"
        [ (* code for level "mul" *) ]
      | "apply"
        [ (* code for level "apply" *) ]
      | "simple"
        [ (* code for level "add" *) ] ]
    ;
  END;
</pre>

<h2 id="a:The-Prtools-module">The Prtools module</h2>

<p>The Prtools module is defined inside Camlp5 for pretty printing
  kits. It provides variables and functions to process comments, and
  meta-functions to process lists (horizontally, vertically,
  paragraphly).</p>

<h3 id="b:Comments">Comments</h3>

<dl>
  <dt><tt>value comm_bef : int -> MLast.loc -> string;</tt></dt>
  <dd>"<tt>comm_bef ind loc</tt>" get the comment from the source just
    before the given location "<tt>loc</tt>". This comment may be
    reindented using "<tt>ind</tt>". Returns the empty string if no
    comment found.</dd>
</dl>

<dl>
  <dt><tt>value source : ref string;</tt></dt>
  <dd>The initial source string, which must be set by the pretty
    printing kit. Used by [comm_bef] above.</dd>
  <dt><tt>value set_comm_min_pos : int -> unit;</tt></dt>
  <dd>Set the minimum position of the source where comments can be
    found, (to prevent possible duplication of comments).</dd>
</dl>

<h3 id="b:Meta-functions-for-lists">Meta functions for lists</h3>

<dl>
  <dt><tt>type pr_fun 'a = pr_context -> 'a -> string;</tt></dt>
  <dd>Type of printer functions.</dd>
</dl>

<dl>
  <dt><tt>value hlist : pr_fun 'a -> pr_fun (list 'a);</tt></dt>
  <dd>[hlist elem pc el] returns the horizontally pretty printed
    string of a list of elements; elements are separated with
    spaces.<br/>The list is displayed in one only line. If this
    function is called in the context of the [horiz] function of the
    function [horiz_vertic] of the module Printing, and if the line
    overflows or contains newlines, the function internally fails (the
    exception is catched by [horiz_vertic] for a vertical pretty
    print).</dd>
  <dt><tt>value hlist2 : pr_fun 'a -> pr_fun 'a -> pr_fun (list 'a);</tt></dt>
  <dd>horizontal list with a different function from 2nd element on.</dd>
  <dt><tt>value hlistl : pr_fun 'a -> pr_fun 'a -> pr_fun (list 'a);</tt></dt>
  <dd>horizontal list with a different function for the last element.</dd>
</dl>

<dl>
  <dt><tt>value vlist : pr_fun 'a -> pr_fun (list 'a);</tt></dt>
  <dd>[vlist elem pc el] returns the vertically pretty printed string
    of a list of elements; elements are separated with newlines and
    indentations.</dd>
  <dt><tt>value vlist2 : pr_fun 'a -> pr_fun 'a -> pr_fun (list
   'a);</tt></dt>
  <dd>vertical list with different function from 2nd element on.</dd>
  <dt><tt>value vlist3 : pr_fun ('a * bool) -> pr_fun ('a * bool) -> pr_fun
      (list 'a);</tt></dt>
  <dd>vertical list with different function from 2nd element on, the
    boolean value being True for the last element of the list.</dd>
  <dt><tt>value vlistl : pr_fun 'a -> pr_fun 'a -> pr_fun (list
   'a);</tt></dt>
  <dd>vertical list with different function for the last element.</dd>
</dl>

<dl>
  <dt><tt>value plist : pr_fun 'a -> int -> pr_fun (list ('a *
   string));</tt></dt>
  <dd>[plist elem sh pc el] returns the pretty printed string of a
   list of elements with separators. The elements are printed
   horizontally as far as possible. When an element does not fit on
   the line, a newline is added and the element is displayed in the
   next line with an indentation of [sh]. [elem] is the function to
   print elements, [el] a list of pairs (element * separator) (the
   last separator being ignored).</dd>
  <dt><tt>value plistb : pr_fun 'a -> int -> pr_fun (list ('a *
      string));</tt></dt>
  <dd>[plist elem sh pc el] returns the pretty printed string of the
   list of elements, like with [plist] but the value of [pc.bef]
   corresponds to an element already printed, as it were on the
   list. Therefore, if the first element of [el] does not fit in the
   line, a newline and a tabulation is added after [pc.bef].</dd>
  <dt><tt>value plistl : pr_fun 'a -> pr_fun 'a -> int -> pr_fun (list
      ('a * string));</tt></dt>
  <dd>paragraph list with a different function for the last element.</dd>
</dl>

<dl>
  <dt><tt>value hvlistl : pr_fun 'a -> pr_fun 'a -> pr_fun (list
   'a);</tt></dt>
  <dd>applies "<tt>hlistl</tt>" if the context is horizontal; else
    applies "<tt>vlistl</tt>".</dd>
</dl>

<h3 id="b:Miscellaneous">Miscellaneous</h3>

<dl>
  <dt><tt>value tab : int -> string;</tt></dt>
  <dd>[tab ind] is equivalent to [String.make ind ' ']</dd>
</dl>

<dl>
  <dt><tt>value flatten_sequence : MLast.expr -> option (list
   MLast.expr);</tt></dt>
  <dd>[flatten_sequence e]. If [e] is an expression representing a
   sequence, return the list of expressions of the sequence. If some
   of these expressions are already sequences, they are flattened in
   the list. If that list contains expressions of the form let..in
   sequence, this sub-sequence is also flattened with the let..in
   applying only to the first expression of the sequence. If [e] is a
   let..in sequence, it works the same way. If [e] is not a sequence
   nor a let..in sequence, return None.</dd>
</dl>

<h2 id="a:Example-:-repeat--until">Example : repeat..until</h2>

<p>This pretty prints the example
  <a href="syntext.html#a:An-example-:-repeat..until">repeat..until</a>
  statement programmed in the chapter <a href="syntext.html">Syntax
  extensions</a> (first version generating a "<tt>while</tt>"
  statement).</p>

<h3 id="b:The-code">The code</h3>

<p>The pattern generated by the "repeat" statement is recognized
  (sequence ending with a "<tt>while</tt>" whose contents is the same
  than the beginning of the sequence) by the function "is_repeat" and
  the repeat statement is pretty printed in its initial form using the
  function "horiz_vertic" of the Pretty module. File
  "<tt>pr_repeat.ml</tt>":</p>

<pre>
  #load "pa_extprint.cmo";
  #load "q_MLast.cmo";

  open Pcaml;
  open Pretty;
  open Prtools;

  value eq_expr_list el1 el2 =
    if List.length el1 <> List.length el2 then False
    else List.for_all2 eq_expr el1 el2
  ;

  value is_repeat el =
    match List.rev el with
    [ [<:expr< while not $e$ do { $list:el2$ } >> :: rel1] ->
        eq_expr_list (List.rev rel1) el2
    | _ -> False ]
  ;

  value semi_after pr pc = pr {(pc) with aft = sprintf "%s;" pc.aft};

  EXTEND_PRINTER
    pr_expr:
      [ [ <:expr< do { $list:el$ } >> when is_repeat el ->
            match List.rev el with
            [ [<:expr< while not $e$ do { $list:el$ } >> :: _] ->
                horiz_vertic
                  (fun () ->
                     sprintf "%srepeat %s until %s%s" pc.bef
                       (hlistl (semi_after curr) curr
                          {(pc) with bef = ""; aft = ""} el)
                       (curr {(pc) with bef = ""; aft = ""} e)
                       pc.aft)
                  (fun () ->
                     let s1 = sprintf "%srepeat" (tab pc.ind) in
                     let s2 =
                       vlistl (semi_after curr) curr
                         {(pc) with
                          ind = pc.ind + 2;
                          bef = tab (pc.ind + 2);
                          aft = ""}
                         el
                     in
                     let s3 =
                       sprintf "%suntil %s" (tab pc.ind)
                         (curr {(pc) with bef = ""} e)
                     in
                     sprintf "%s\n%s\n%s" s1 s2 s3)
            | _ -> assert False ] ] ]
    ;
  END;
</pre>

<h3 id="b:Compilation">Compilation</h3>

<pre>
  ocamlc -pp camlp5r -I +camlp5 -c pr_repeat.ml
</pre>

<h3 id="b:Testing">Testing</h3>

<p>Getting the same files "foo.ml" and "bar.ml" of the repeat syntax
  example:</p>

<pre>
  $ cat bar.ml
  #load "./foo.cmo";
  value x = ref 42;
  repeat
    print_int x.val;
    print_endline "";
    x.val := x.val + 3
  until x.val > 70;

  $ camlp
</pre>

<p>Without the pretty printing kit:</p>

<pre>
  $ camlp5r pr_r.cmo bar.ml
  #load "./foo.cmo";
  value x = ref 42;
  do {
    print_int x.val;
    print_endline "";
    x.val := x.val + 3;
    while not (x.val > 70) do {
      print_int x.val;
      print_endline "";
      x.val := x.val + 3
    }
  };
</pre>

<p>With the pretty printing kit:</p>

<pre>
  $ camlp5r pr_r.cmo ./pr_repeat.cmo bar.ml -l 75
  #load "./foo.cmo";
  value x = ref 42;
  repeat
    print_int x.val;
    print_endline "";
    x.val := x.val + 3
  until x.val > 70;
</pre>

<a class="toplink" href="opretty.html">↑</a>
<div class="trailer">

  <hr style="margin:0" /><div style="font-size: 80%"><em>Copyright 2007-2014
      Daniel de Rauglaudre (INRIA)</em></div>

  <p class="bottom">
    <a href="http://validator.w3.org/check?uri=referer"><img
       alt="Valid XHTML 1.1" height="31" width="88" /></a>
  </p>

</div>

</div>

</body>
</html>
camlp5 6.16-1 / usr / share / doc / camlp5 / html / opretty.html
