path: root/doc/haddock.sgml

<!DOCTYPE BOOK PUBLIC "-//OASIS//DTD DocBook V3.1//EN">

<book id="haddock">
  <bookinfo>
    <date>2002-4-10</date>
    <title>Haddock User Guide</title>
    <author>
      <firstname>Simon</firstname>
      <surname>Marlow</surname>
    </author>
    <address><email>simonmar@microsoft.com</email></address>
    <copyright>
      <year>2002</year>
      <holder>Simon Marlow</holder>
    </copyright>
    <abstract>
      <para>This document describes Haddock, a Haskell documentation
      tool.</para>
    </abstract>
  </bookinfo>

  <!-- Table of contents -->
  <toc></toc>
  
  <chapter id="introduction">
    <title>Introduction</title>

    <para>This is Haddock, a tool for automatically generating
    documentation from annotated Haskell source code.  Haddock was
    designed with several goals in mind:</para>

    <itemizedlist>
      <listitem>
	<para>When documenting APIs, it is desirable to keep the
	documentation close to the actual interface or implementation
	of the API, preferably in the same file, to reduce the risk
	that the two become out of sync.  Haddock therefore lets you
	write the documentation for an entity (function, type, or
	class) next to the definition of the entity in the source
	code.</para>
      </listitem>
      <listitem>
	<para>There is s tremendous amount of useful API documentation
	that can be extracted from just the bare source code,
	including types of exported functions, definitions of data
	types and classes, and so on.  Haddock can therefore generate
	documentation from a set of straight Haskell 98 modules, and
	the documentation will contain precisely the interface that is
	available to a programmer using those modules.</para>
      </listitem>
      <listitem>
	<para>Documentation annotations in the source code should be
	easy on the eye when editing the source code itself, so as not
	to obsure the code and to make reading and writing
	documentation annotations easy.  The easier it is to write
	documentation, the more likely the programmer is to do it.
	Haddock therefore uses lightweight markup in its annotations,
	taking several ideas from <ulink
	url="http://www.cse.unsw.edu.au/~chak/haskell/idoc/">IDoc</ulink>.
	In fact, Haddock can understand IDoc-annotated source
	code.</para>
      </listitem>
      <listitem>
	<para>The documentation should not expose any of the structure
	of the implementation, or to put it another way, the
	implementer of the API should be free to structure the
	implementation however he or she wishes, without exposing any
	of that structure to the consumer.  In practical terms, this
	means that while an API may internally consist of several
	Haskell modules, we often only want to expose a single module
	to the user of the interface, where this single module just
	re-exports the relevant parts of the implementation
	modules.</para>

	<para>Haddock therefore understands the Haskell module system
	and can generate documentation which hides not only
	non-exported entities from the interface, but also the
	internal module structure of the interface.  A documentation
	annotation can still be placed next to the implementation, and
	it will be propagated to the external module in the generated
	documentation.</para>
      </listitem>
      <listitem>
	<para>Being able to move around the documentation by following
	hyperlinks is essential.  Documentation generated by Haddock
	is therefore littered with hyperlinks: every type and class
	name is a link to the corresponding definition, and
	user-written documentation annotations can contain identifiers
	which are linked automatically when the documentation is
	generated.</para>
      </listitem>
      <listitem>
	<para>We might want documentation in multiple formats - online
	and printed, for example.  Haddock comes with HTML and DocBook
	backends, and it is structured in such a way that adding new
	back-ends is straightforward.</para>
      </listitem>
    </itemizedlist>

    <section id="obtaining">
      <title>Obtaining Haddock</title>
      
      <para>Distributions (source & binary) of Haddock can be obtained
      from its <ulink url="http://www.haskell.org/haddock/">web
      site</ulink>.</para>

      <para>Up-to-date sources can also be obtained from CVS.  The
      Haddock sources are under <literal>fptools/haddock</literal> in
      the <literal>fptools</literal> CVS repository, which also
      contains GHC, Happy, and several other projects.  See <ulink
      url="http://www.haskell.org/ghc/docs/latest/building/sec-cvs.html">Using
      The CVS Repository</ulink> for information on how to access the
      CVS repository.  Note that you need to check out the
      <literal>fpconfig</literal> module first to get the generic
      build system (the <literal>fptools</literal> directory), and
      then check out <literal>fptools/haddock</literal> to get the
      Haddock sources.</para>
    </section>

    <section id="license">
      <title>License</title>

      <para>The following license covers this documentation, and the
      Haddock source code, except where otherwise indicated.</para>

      <blockquote>
	<para>Copyright 2002, Simon Marlow.  All rights reserved.</para>

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

	<itemizedlist>
	  <listitem>
	    <para>Redistributions of source code must retain the above
            copyright notice, this list of conditions and the
            following disclaimer.</para>
	  </listitem>
	  <listitem>
	    <para>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.</para>
	  </listitem>
	</itemizedlist>

	<para>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.</para>
      </blockquote>
    </section>

    <section>
      <title>Acknowledgements</title>
      
      <para>Several documentation systems provided the inspiration for
      Haddock, most notably:</para>

      <itemizedlist>
	<listitem>
	  <para><ulink
          url="http://www.cse.unsw.edu.au/~chak/haskell/idoc/">
          IDoc</ulink></para>
	</listitem>
	<listitem>
	  <para><ulink
	  url="http://www.fmi.uni-passau.de/~groessli/hdoc/">HDoc</ulink></para>
	</listitem>
	<listitem>
	  <para><ulink url="http://www.stack.nl/~dimitri/doxygen/">
          Doxygen</ulink></para>
	</listitem>
      </itemizedlist>

      <para>and probably several others I've forgotten.</para>

      <para>Thanks to the following people for useful feedback,
      discussion, patches, packaging, and moral support: Simon Peyton
      Jones, Mark Shields, Manuel Chakravarty, Ross Patterson, Brett
      Letner, Sven Panne, Hal Daume, George Russell, Oliver Braun,
      Ashley Yakeley, Malcolm Wallace, Krasimir Angelov, the members
      of <email>haskelldoc@haskell.org</email>, and everyone who
      contributed to the many libraries that Haddock makes use
      of.</para>
    </section>

  </chapter>

  <chapter id="invoking">
    <title>Invoking Haddock</title>
    <para>Haddock is invoked from the command line, like so:</para>

    <cmdsynopsis>
      <command>haddock</command>
      <arg rep=repeat><replaceable>option</replaceable></arg>
      <arg rep=repeat choice=plain><replaceable>file</replaceable></arg>
    </cmdsynopsis>

    <para>Where each <replaceable>file</replaceable> is a filename
    containing a Haskell source module.  All the modules specified on
    the command line will be processed together.  When one module
    refers to an entity in another module being processed, the
    documentation will link directly to that entity.</para>

    <para>Entities that cannot be found, for example because they are
    in a module that isn't being processed as part of the current
    batch, simply won't be hyperlinked in the generated
    documentation.  Haddock will emit warnings listing all the
    indentifiers it couldn't resolve.</para>

    <para>The modules should <emphasis>not</emphasis> be mutually
    recursive, as Haddock don't like swimming in circles.</para> 

    <para>The following options are available:</para>

    <variablelist>
      <varlistentry>
	<term><option>-d</option></term>
	<term><option>--docbook</option></term>
	<indexterm><primary><option>-d</option></primary></indexterm>
	<indexterm><primary><option>--docbook</option></primary></indexterm>
	<listitem>
	  <para>Output documentation in SGML DocBook format.  NOTE: at
	  time of writing this is only partially implemented and
	  doesn't work.</para>
	</listitem>
      </varlistentry>

      <varlistentry>
	<term><option>--dump-interface</option>=<replaceable>file</replaceable></term>
	<indexterm><primary><option>--dump-interface</option></primary>
	</indexterm>
	<listitem>
	  <para>Produce an <firstterm>interface
	  file</firstterm><footnote><para>Haddock interface files are
	  not the same as Haskell interface files, I just couldn't
	  think of a better name.</para> </footnote>
	  in the file <replaceable>file</replaceable>.  An interface
	  file contains information Haddock needs to produce more
	  documentation that refers to the modules currently being
	  processed - see the <option>--read-interface</option> option
	  for more details.  The interface file is in a binary format;
	  don't try to read it.</para>
	</listitem>
      </varlistentry>

      <varlistentry>
	<term><option>-h</option></term>
	<term><option>--html</option></term>
	<indexterm><primary><option>-h</option></primary></indexterm>
	<indexterm><primary><option>--html</option></primary></indexterm>
	<listitem>
	  <para>Generate documentation in HTML format.  Several files
	  will be generated into the current directory (or the
	  specified directory if the <option>-o</option> option is
	  given), including the following:</para>
	  <variablelist>
	    <varlistentry>
	      <term><filename>index.html</filename></term>
	      <listitem>
		<para>The top level page of the documentation: lists
		the modules available, using indentation to represent
		the hierarchy if the modules are hierarchical.</para>
	      </listitem>
	    </varlistentry>
	    <varlistentry>
	      <term><filename>haddock.css</filename></term>
	      <listitem>
		<para>The stylesheet used by the generated HTML.  Feel
		free to modify this to change the colors or
		layout, or even specify your own stylesheet using the
		<option>--css</option> option.</para>
	      </listitem>
	    </varlistentry>
	    <varlistentry>
	      <term><filename><replaceable>module</replaceable>.html</filename></term>
	      <listitem>
		<para>An HTML page for each
		<replaceable>module</replaceable>.</para>
	      </listitem>
	    </varlistentry>
	    <varlistentry>
	      <term><filename>doc-index.html</filename></term>
	      <term><filename>doc-index-XX.html</filename></term>
	      <listitem>
		<para>The index, split into two
		(functions/constructors and types/classes, as per
		Haskell namespaces) and further split
		alphabetically.</para>
	      </listitem>
	    </varlistentry>
	  </variablelist>
	</listitem>
      </varlistentry>

      <varlistentry>
	<term><option>-i</option> <replaceable>path</replaceable>,<replaceable>file</replaceable></term>
	<term><option>--read-interface</option>=<replaceable>path</replaceable>,<replaceable>file</replaceable></term>
	<indexterm><primary><option>-i</option></primary></indexterm>
	<indexterm><primary><option>--read-interface</option></primary></indexterm>
	<listitem>
	  <para>Read the interface file in
	  <replaceable>file</replaceable>, which must have been
	  produced by running Haddock with the
	  <option>--dump-interface</option> option.  The interface
	  describes a set of modules whose HTML documentation is
	  located in <replaceable>path</replaceable> (which may be a
	  relative pathname).  The <replaceable>path</replaceable> is
	  optional, and defaults to <quote>.</quote>.</para>

	  <para>This option allows Haddock to produce separate sets of
	  documentation with hyperlinks between them.  The
	  <replaceable>path</replaceable> is used to direct hyperlinks
	  to point to the right files; so make sure you don't move the
	  HTML files later or these links will break.  Using a
	  relative <replaceable>path</replaceable> means that a
	  documentation subtree can still be moved around without
	  breaking links.</para>

	  <para>Multiple <option>--read-interface</option> options may
	  be given.</para>
	</listitem>
      </varlistentry>

      <varlistentry>
	<term><option>--ms-help</option></term>
	<indexterm><primary><option>--ms-help</option></primary>
	</indexterm>
	<listitem>
	  <para>(In HTML mode only) Produce extra contents and index
	  files for the Microsoft HTML Help system.  The files created
	  will be named <filename>index.hhc</filename> and
	  <filename>index.hhk</filename> respectively.</para>

	  <para>Using the Microsoft HTML Help system provides two
	  advantages over plain HTML: the help viewer gives you a nice
	  hierarchical folding contents pane on the left, and the
	  documentation files are compressed and therefore much
	  smaller (roughly a factor of 10).  The disadvantages are
	  that the viewer is only available on Windows, and the help
	  can't be viewed over the web.</para>

	  <para>In order to create a compiled Microsoft help file, you
	  also need the Microsoft HTML Help compiler, which is
	  available free from <ulink
	  url="http://www.microsoft.com/">http://www.microsoft.com/</ulink>
	  (search for <quote>HTML Help compiler</quote>).</para>
	</listitem>
      </varlistentry>

      <varlistentry>
	<term><option>--no-implicit-prelude</option></term>
	<indexterm><primary><option>--no-implicit-prelude</option></primary>
	</indexterm>
	<listitem>
	  <para>Don't automatically import <literal>Prelude</literal>
	  in every module.  Used when producing documentation for the
	  <literal>Prelude</literal> itself.</para>
	</listitem>
      </varlistentry>

      <varlistentry>
	<term><option>-o</option> <replaceable>dir</replaceable></term>
	<term><option>--odir</option>=<replaceable>dir</replaceable></term>
	<indexterm><primary><option>-o</option></primary></indexterm>
	<indexterm><primary><option>--odir</option></primary></indexterm>
	<listitem>
	  <para>Generate files into <replaceable>dir</replaceable>
	  instead of the current directory.</para>
	</listitem>
      </varlistentry>

      <varlistentry>
	<term><option>-p</option> <replaceable>file</replaceable></term>
	<term><option>--prologue</option>=<replaceable>file</replaceable></term>
	<indexterm><primary><option>-p</option></primary></indexterm>
	<indexterm><primary><option>--prologue</option></primary></indexterm>
	<listitem>
	  <para>Specify a file containing documentation which is
	  placed on the main contents page under the heading
	  &ldquo;Description&rdquo;.  The file is parsed as a normal
	  Haddock doc comment (but the comment markers are not
	  required).</para>
	</listitem>
      </varlistentry>

      <varlistentry>
	<term><option>-s</option> <replaceable>URL</replaceable></term>
	<term><option>--source</option>=<replaceable>URL</replaceable></term>
	<indexterm><primary><option>-s</option></primary></indexterm>
	<indexterm><primary><option>--source</option></primary></indexterm>
	<listitem>
	  <para>Include links to the source files in the generated
	  documentation, where <replaceable>URL</replaceable> is the
	  base URL where the source files can be found.</para>
	</listitem>
      </varlistentry>

      <varlistentry>
	<term><option>-t</option> <replaceable>title</replaceable></term>
	<term><option>--title</option>=<replaceable>title</replaceable></term>
	<indexterm><primary><option>-t</option></primary></indexterm>
	<indexterm><primary><option>--title</option></primary></indexterm>
	<listitem>
	  <para>Use <replaceable>title</replaceable> as the page
	  heading for each page in the documentation.This will
	  normally be the name of the library being documented.</para>

	  <para>The title should be a plain string (no markup
	  please!).</para>
	</listitem>
      </varlistentry>

      <varlistentry>
	<term><option>-v</option></term>
	<term><option>--verbose</option></term>
	<indexterm><primary><option>-v</option></primary></indexterm>
	<indexterm><primary><option>--verbose</option></primary></indexterm>
	<listitem>
	  <para>Reserved for future expansion.</para>
	</listitem>
      </varlistentry>

      <varlistentry>
	<term><option>--css</option>=<replaceable>filename</replaceable></term>
	<indexterm><primary><option>--css</option></primary></indexterm>
	<listitem>
	  <para>Specify a stylesheet to use instead of the default one
	  that comes with Haddock.  It should specify certain classes:
	  see the default stylesheet for details.</para>
	</listitem>
      </varlistentry>
    </variablelist>

  </chapter>
  
  <chapter id="markup">
    <title>Documentation and Markup</title>

    <para>Haddock understands special documentation annotations in the
    Haskell source file and propagates these into the generated
    documentation.  The annotations are purely optional: if there are
    no annotations, Haddock will just generate documentation that
    contains the type signatures, data type declarations, and class
    declarations exported by each of the modules being
    processed.</para>

    <section>
      <title>Documenting a top-level declaration</title>

      <para>The simplest example of a documentation annotation is for
      documenting any top-level declaration (function type signature,
      type declaration, or class declaration).  For example, if the
      source file contains the following type signature:</para>

<programlisting>
square :: Int -> Int
square x = x * x
</programlisting>

    <para>Then we can document it like this:</para>

<programlisting>
-- |The 'square' function squares an integer.
square :: Int -> Int
square x = x * x
</programlisting>


      <para>The <quote><literal>-- |</literal></quote> syntax begins a
      documentation annotation, which applies to the
      <emphasis>following</emphasis> declaration in the source file.
      Note that the annotation is just a comment in Haskell &mdash; it
      will be ignored by the Haskell compiler.</para>

      <para>The declaration following a documentation annotation
      should be one of the following:</para>
      <itemizedlist>
	<listitem>
	  <para>A type signature for a top-level function,</para>
	</listitem>
	<listitem>
	  <para>A <literal>data</literal> declaration,</para>
	</listitem>
	<listitem>
	  <para>A <literal>newtype</literal> declaration,</para>
	</listitem>
	<listitem>
	  <para>A <literal>type</literal> declaration, or</para>
	</listitem>
	<listitem>
	  <para>A <literal>class</literal> declaration.</para>
	</listitem>
      </itemizedlist>

      <para>If the annotation is followed by a different kind of
      declaration, it will probably be ignored by Haddock.</para>

      <para>Some people like to write their documentation
      <emphasis>after</emphasis> the declaration; this is possible in
      Haddock too:</para>
     
<programlisting>
square :: Int -> Int
-- ^The 'square' function squares an integer.
square x = x * x
</programlisting>

      <para>Note that Haddock doesn't contain a Haskell type system
      &mdash; if you don't write the type signature for a function,
      then Haddock can't tell what its type is and it won't be
      included in the documentation.</para>
      
      <para>Documentation annotations may span several lines; the
      annotation continues until the first non-comment line in the
      source file.  For example:</para>

<programlisting>
-- |The 'square' function squares an integer.
-- It takes one argument, of type 'Int'.
square :: Int -> Int
square x = x * x
</programlisting>

      <para>You can also use Haskell's nested-comment style for
      documentation annotations, which is sometimes more convenient
      when using multi-line comments:</para>

<programlisting>
{-|
  The 'square' function squares an integer.
  It takes one argument, of type 'Int'.
-}
square :: Int -> Int
square x = x * x
</programlisting>
      
    </section>
    <section>
      <title>Documenting parts of a declaration</title>
      
      <para>In addition to documenting the whole declaration, in some
      cases we can also document individual parts of the
      declaration.</para>

      <section>
	<title>Class methods</title>
	  
	<para>Class methods are documented in the same way as top
	level type signatures, by using either the
	<quote><literal>--&nbsp;|</literal></quote> or
	<quote><literal>--&nbsp;^</literal></quote>
	annotations:</para>
	  
<programlisting>
class C a where
   -- | This is the documentation for the 'f' method
   f :: a -> Int
   -- | This is the documentation for the 'g' method
   g :: Int -> a
</programlisting>
      </section>

      <section>
	<title>Constructors and record fields</title>

	<para>Constructors are documented like so:</para>

<programlisting>
data T a b
  -- | This is the documentation for the 'C1' constructor
  = C1 a b 
  -- | This is the documentation for the 'C2' constructor
  | C2 a b 
</programlisting>

	<para>or like this:</para>

<programlisting>
data T a b
  = C1 a b  -- ^ This is the documentation for the 'C1' constructor
  | C2 a b  -- ^ This is the documentation for the 'C2' constructor
</programlisting>

	<para>Record fields are documented using one of these
	styles:</para>

<programlisting>
data R a b = 
  C { -- | This is the documentation for the 'a' field
      a :: a,
      -- | This is the documentation for the 'b' field
      b :: b
    }

data R a b = 
  C { a :: a  -- ^ This is the documentation for the 'a' field
    , b :: b  -- ^ This is the documentation for the 'b' field
    }
</programlisting>

        <para>Alternative layout styles are generally accepted by
        Haddock - for example doc comments can appear before or after
        the comma in separated lists such as the list of record fields
        above.</para>
      </section>

      <section>
	<title>Function arguments</title>
	
	<para>Individual arguments to a function may be documented
	like this:</para>

<programlisting>
f  :: Int      -- ^ The 'Int' argument
   -> Float    -- ^ The 'Float' argument
   -> IO ()    -- ^ The return value
</programlisting>
      </section>
    </section>

    <section>
      <title>The module description</title>

      <para>A module may contain a documentation comment before the
      module header, in which case this comment is interpreted by
      Haddock as an overall description of the module itself, and
      placed in a section entitled <quote>Description</quote> in the
      documentation for the module.  For example:</para>

<programlisting>
-- | This is the description for module "Foo"
module Foo where
...
</programlisting>
    </section>
    
    <section>
      <title>Controlling the documentation structure</title>
	
      <para>Haddock produces interface documentation that lists only
      the entities actually exported by the module.  The documentation
      for a module will include <emphasis>all</emphasis> entities
      exported by that module, even if they were re-exported by
      another module.  The only exception is when Haddock can't see
      the declaration for the re-exported entity, perhaps because it
      isn't part of the batch of modules currently being
      processed.</para>

      <para>However, to Haddock the export list has even more
      significance than just specifying the entities to be included in
      the documentation.  It also specifies the
      <emphasis>order</emphasis> that entities will be listed in the
      generated documentation.  This leaves the programmer free to
      implement functions in any order he/she pleases, and indeed in
      any <emphasis>module</emphasis> he/she pleases, but still
      specify the order that the functions should be documented in the
      export list.  Indeed, many programmers already do this: the
      export list is often used as a kind of ad-hoc interface
      documentation, with headings, groups of functions, type
      signatures and declarations in comments.</para>

      <para>You can insert headings and sub-headings in the
      documentation by including annotations at the appropriate point
      in the export list.  For example:</para>

<programlisting>
module Foo (
  -- * Classes
  C(..),
  -- * Types
  -- ** A data type
  T,
  -- ** A record
  R,
  -- * Some functions
  f, g
  ) where
</programlisting>

      <para>Headings are introduced with the syntax
      <quote><literal>--&nbsp;*</literal></quote>,
      <quote><literal>--&nbsp;**</literal></quote> and so on, where
      the number of <literal>*</literal>s indicates the level of the
      heading (section, sub-section, sub-sub-section, etc.).</para>

      <para>If you use section headings, then Haddock will generate a
      table of contents at the top of the module documentation for
      you.</para>

      <para>The alternative style of placing the commas at the
      beginning of each line is also supported. eg.:</para>

<programlisting>
module Foo (
  -- * Classes
  , C(..)
  -- * Types
  -- ** A data type
  , T
  -- ** A record
  , R
  -- * Some functions
  , f
  , g
  ) where
</programlisting>

      <section>
	<title>Re-exporting an entire module</title>
	
	<para>Haskell allows you to re-export the entire contents of a
	module (or at least, everything currently in scope that was
	imported from a given module) by listing it in the export
	list:</para>

<programlisting>
module A (
  module B,
  module C
 ) where	  
</programlisting>

	<para>What will the Haddock-generated documentation for this
	module look like?  Well, it depends on how the modules
	<literal>B</literal> and <literal>C</literal> are imported.
	If they are imported wholly and without any
	<literal>hiding</literal> qualifiers, then the documentation
	will just contain a cross-reference to the documentation for
	<literal>B</literal> and <literal>C</literal>.  However, if
	the modules are not <emphasis>completely</emphasis>
	re-exported, for example:</para>

<programlisting>
module A (
  module B,
  module C
 ) where	  

import B hiding (f)
import C (a, b)
</programlisting>

	<para>then Haddock behaves as if the set of entities
	re-exported from <literal>B</literal> and <literal>C</literal>
	had been listed explicitly in the export
	list<footnote><para>NOTE: this is not fully implemented at the
	time of writing (version 0.2).  At the moment, Haddock always
	inserts a cross-reference.</para>
	  </footnote>.</para>

	<para>The exception to this rule is when the re-exported
	module is declared with the <literal>hide</literal> attribute
	(<xref linkend="module-attributes">), in which case the module
	is never cross-referenced; the contents are always expanded in
	place in the re-exporting module.</para>
      </section>
      
      <section>
	<title>Omitting the export list</title>
	
	<para>If there is no export list in the module, how does
	Haddock generate documentation?  Well, when the export list is
	omitted, e.g.:</para>

<programlisting>module Foo where</programlisting>

	<para>this is equivalent to an export list which mentions
	every entity defined at the top level in this module, and
	Haddock treats it in the same way.  Furthermore, the generated
	documentation will retain the order in which entities are
	defined in the module.  In this special case the module body
	may also include section headings (normally they would be
	ignored by Haddock).</para>
      </section>
    </section>

    <section>
      <title>Named chunks of documentation</title>

      <para>Occasionally it is desirable to include a chunk of
      documentation which is not attached to any particular Haskell
      declaration.  There are two ways to do this:</para>

      <itemizedlist>
	<listitem>
	  <para>The documentation can be included in the export list
	  directly, e.g.:</para>

<programlisting>
module Foo (
   -- * A section heading

   -- | Some documentation not attached to a particular Haskell entity
   ...
 ) where
</programlisting>
	</listitem>
	
	<listitem>
	  <para>If the documentation is large and placing it inline in
	  the export list might bloat the export list and obscure the
	  structure, then it can be given a name and placed out of
	  line in the body of the module.  This is achieved with a
	  special form of documentation annotation
	  <quote><literal>--&nbsp;$</literal></quote>:</para>

<programlisting>
module Foo (
   -- * A section heading

   -- $doc
   ...
 ) where

-- $doc 
-- Here is a large chunk of documentation which may be referred to by
-- the name $doc.
</programlisting>

	  <para>The documentation chunk is given a name, which is the
	  sequence of alphanumeric characters directly after the
	  <quote><literal>--&nbsp;$</literal></quote>, and it may be
	  referred to by the same name in the export list.</para>
	</listitem>
      </itemizedlist>
    </section>

    <section>
      <title>Hyperlinking and re-exported entities</title> 
      
      <para>When Haddock renders a type in the generated
      documentation, it hyperlinks all the type constructors and class
      names in that type to their respective definitions.  But for a
      given type constructor or class there may be several modules
      re-exporting it, and therefore several modules whose
      documentation contains the definition of that type or class
      (possibly including the current module!) so which one do we link
      to?</para>

      <para>Let's look at an example.  Suppose we have three modules
      <literal>A</literal>, <literal>B</literal> and
      <literal>C</literal> defined as follows:</para>

<programlisting>
module A (T) where
data T a = C a

module B (f) where
import A
f :: T Int -> Int
f (C i) = i

module C (T, f) where
import A
import B
</programlisting>

      <para>Module <literal>A</literal> exports a datatype
      <literal>T</literal>.  Module <literal>B</literal> imports
      <literal>A</literal> and exports a function <literal>f</literal>
      whose type refers to <literal>T</literal>: the hyperlink in
      <literal>f</literal>'s signature will point to the definition of
      <literal>T</literal> in the documentation for module
      <literal>A</literal>.</para>

      <para>Now, module <literal>C</literal> exports both
      <literal>T</literal> and <literal>f</literal>.  We have a choice
      about where to point the hyperlink to <literal>T</literal> in
      <literal>f</literal>'s type: either the definition exported by
      module <literal>C</literal> or the definition exported by module
      <literal>A</literal>.  Haddock takes the view that in this case
      pointing to the definition in <literal>C</literal> is better,
      because the programmer might not wish to expose
      <literal>A</literal> to the programmer at all:
      <literal>A</literal> might be a module internal to the
      implementation of the library in which <literal>C</literal> is
      the external interface, so linking to definitions in the current
      module is preferrable over an imported module.</para>

      <para>The general rule is this: when attempting to link an
      instance of a type constructor or class to its definition, the
      link is made to</para>
      <itemizedlist>
	<listitem>
	  <para>the current module, if the current module exports the
	  relevant definition, or</para>
	</listitem>
	<listitem>
	  <para>the module that the entity was imported from,
	  otherwise.  If the entity was imported via multiple routes,
	  then Haddock picks the module listed earliest in the imports
	  of the current module.</para>
	</listitem>
      </itemizedlist>
    </section>

    <section id="module-attributes">
      <title>Module Attributes</title>
      
      <para>Certain attributes may be specified for each module which
      affects the way that Haddock generates documentation for that
      module.  Attributes are specified in a comma-separated list in a
      <literal>-- #</literal> (or
      <literal>{-&nbsp;#&nbsp;...&nbsp;-}</literal>) comment at the
      top of the module, either before or after the module
      description.  For example:</para>

<programlisting>
-- #hide, prune, ignore-exports
-- |Module description
module A where
...
</programlisting>

      <para>The following attributes are currently understood by
      Haddock:</para>

      <variablelist>
	<varlistentry>
	  <term><literal>hide</literal></term>
	  <indexterm><primary><literal>hide</literal></primary></indexterm>
	  <listitem>
	    <para>Omit this module from the generated documentation,
	    but nevertheless propagate definitions and documentation
	    from within this module to modules that re-export those
	    definitions.</para>
	  </listitem>
	</varlistentry>
	
	<varlistentry>
	  <term><literal>prune</literal></term>
	  <indexterm><primary><literal>hide</literal></primary></indexterm>
	  <listitem>
	    <para>Omit definitions that have no documentation
	    annotations from the generated documentation.</para>
	  </listitem>
	</varlistentry>
	
	<varlistentry>
	  <term><literal>ignore-exports</literal></term>
	  <indexterm><primary><literal>hide</literal></primary></indexterm>
	  <listitem>
	    <para>Ignore the export list.  Generate documentation as
	    if the module had no export list - i.e. all the top-level
	    declarations are exported, and section headings may be
	    given in the body of the module.</para>
	  </listitem>
	</varlistentry>
      </variablelist>
    </section>

    <section>
      <title>Markup</title>

      <para>Haddock understands certain textual cues inside
      documentation annotations that tell it how to render the
      documentation.  The cues (or <quote>markup</quote>) have been
      designed to be simple and mnemonic in ASCII so that the
      programmer doesn't have to deal with heavyweight annotations
      when editing documentation comments.</para>

      <section>
	<title>Paragraphs</title>
	
	<para>One or more blank lines separates two paragraphs in a
	documentation comment.</para>
      </section>

      <section>
	<title>Special characters</title>

	<para>The following characters have special meanings in
	documentation comments: <literal>/</literal>,
	<literal>'</literal>, <literal>`</literal>,
	<literal>"</literal>, <literal>@</literal>,
	<literal>&lt;</literal>.  To insert a literal occurrence of
	one of these special characters, precede it with a backslash
	(<literal>\</literal>).</para>

	<para>Additionally, the character <literal>&gt;</literal> has
	a special meaning at the beginning of a line, and the
	following characters have special meanings at the beginning of
	a paragraph:
	<literal>*</literal>, <literal>-</literal>.  These characters
	can also be escaped using <literal>\</literal>.</para>
      </section>

      <section>
	<title>Code Blocks</title>

	<para>Displayed blocks of code are indicated by surrounding a
	paragraph with <literal>@...@</literal> or by preceding each
	line of a paragraph with <literal>&gt;</literal> (we often
	call these &ldquo;bird tracks&rdquo;).  For
	example:</para>

<programlisting>
-- | This documentation includes two blocks of code:
--
-- @
--     f x = x + x
-- @
--
-- &gt;  g x = x * 42
</programlisting>

	<para>There is an important difference between the two forms
        of code block: in the bird-track form, the text to the right
        of the &lsquo;<literal>></literal>&rsquo; is interpreted
        literally, whereas the <literal>@...@</literal> form
        interprets markup as normal inside the code block.</para>
      </section>

      <section>
	<title>Hyperlinked Identifiers</title>

	<para>Referring to a Haskell identifier, whether it be a type,
	class, constructor, or function, is done by surrounding it
	with single quotes:</para>

<programlisting>
-- | This module defines the type 'T'.
</programlisting>

	<para>If there is an entity <literal>T</literal> in scope in
	the current module, then the documentation will hyperlink the
	reference in the text to the definition of
	<literal>T</literal> (if the output format supports
	hyperlinking, of course; in a printed format it might instead
	insert a page reference to the definition).</para>

	<para>It is also possible to refer to entities that are not in
	scope in the current module, by giving the full qualified name
	of the entity:</para>

<programlisting>
-- | The identifier 'M.T' is not in scope
</programlisting>

	<para>If <literal>M.T</literal> is not otherwise in scope,
	then Haddock will simply emit a link pointing to the entity
	<literal>T</literal> exported from module <literal>M</literal>
	(without checking to see whether either <literal>M</literal>
	or <literal>M.T</literal> exist).</para>

	<para>To make life easier for documentation writers, a quoted
	identifier is only interpreted as such if the quotes surround
	a lexically valid Haskell identifier.  This means, for
	example, that it normally isn't necessary to escape the single
	quote when used as an apostrophe:</para>

<programlisting>
-- | I don't have to escape my apostrophes; great, isn't it?
</programlisting>

	<para>For compatibility with other systems, the following
	alternative form of markup is accepted<footnote><para>
	We chose not to use this as the primary markup for
	identifiers because strictly speaking the <literal>`</literal>
	character should not be used as a left quote, it is a grave accent.</para>
	  </footnote>: <literal>`T'</literal>.</para>
      </section>

      <section>
	<title>Emphasis and Monospaced text</title>

	<para>Emphasis may be added by surrounding text with
	<literal>/.../</literal>.</para>

	<para>Monospaced (or typewriter) text is indicated by
	surrounding it with <literal>@...@</literal>.  Other markup is
	valid inside a monospaced span: for example
	<literal>@'f'&nbsp;a&nbsp;b@</literal> will hyperlink the
	identifier <literal>f</literal> inside the code fragment.</para>
      </section>

      <section>
	<title>Linking to modules</title>

	<para>Linking to a module is done by surrounding the module
	name with double quotes:</para>

<programlisting>
-- | This is a reference to the "Foo" module.
</programlisting>

      </section>

      <section>
	<title>Itemized and Enumerated lists</title>

	<para>A bulleted item is represented by preceding a paragraph
	with either <quote><literal>*</literal></quote> or
	<quote><literal>-</literal></quote>.  A sequence of bulleted
	paragraphs is rendered as an itemized list in the generated
	documentation, eg.:</para>

<programlisting>
-- | This is a bulleted list:
--
--     * first item
--
--     * second item
</programlisting>

	<para>An enumerated list is similar, except each paragraph
	must be preceded by either
	<quote><literal>(<replaceable>n</replaceable>)</literal></quote>
	or
	<quote><literal><replaceable>n</replaceable>.</literal></quote>
	where <replaceable>n</replaceable> is any integer.  e.g.</para>

<programlisting>
-- | This is an enumerated list:
--
--     (1) first item
--
--     2. second item
</programlisting>
      </section>

      <section>
	<title>URLs</title>

	<para>A URL can be included in a documentation comment by
	surrounding it in angle brackets:
	<literal>&lt;...&gt;</literal>.  If the output format supports
	it, the URL will be turned into a hyperlink when
	rendered.</para>
      </section>
    </section>
  </chapter>
</book>