<?xml version="1.0"?>
<!DOCTYPE refentry PUBLIC "-//OASIS//DTD DocBook XML V4.3//EN"
"http://www.oasis-open.org/docbook/xml/4.3/docbookx.dtd"
[
<!ENTITY % local.common.attrib "xmlns:xi CDATA #FIXED 'http://www.w3.org/2003/XInclude'">
<!ENTITY % gtkdocentities SYSTEM "xml/gtkdocentities.ent">
%gtkdocentities;
]>
<refentry id="lang-tutorial" revision="5 May 2005">
<refmeta>
<refentrytitle>Language Definition v2.0 Tutorial</refentrytitle>
</refmeta>
<refnamediv>
<refname>Language Definition v2.0 Tutorial</refname>
<refpurpose>
Guide to the GtkSourceView language definition file format
</refpurpose>
</refnamediv>
<refsect1 id="tutorial">
<title>A language definition for the C language</title>
<note>
<para>
The version 2 here refers to the language definition file format,
not to the version of GtkSourceView. This tutorial is suitable
for GtkSourceView 2 and 3.
</para>
</note>
<para>
To describe the syntax of a language GtkSourceView uses an XML format which
defines nested context to be highlighted. Each context roughly corresponds
to a portion of the syntax which has to be highlighted (e.g. keywords,
strings, comments), and can contain nested contexts (e.g. escaped
characters.)
</para>
<para>
In this tutorial we will analyze a simple example to highlight a subset of
C, based on the full C language definition.
</para>
<para>
Like every well formed XML document, the language description starts with a
XML declaration:
</para>
<programlisting>
<?xml version="1.0" encoding="UTF-8"?>
</programlisting>
<para>
After the usual preamble, the main tag is the <code><language></code> element:
</para>
<programlisting>
<language id="c" _name="C" version="2.0" _section="Source">
</programlisting>
<para>
The attribute <code>id</code> is used in external references and defines a standard
way to refer to this language definition, while the <code>name</code> attribute is
the name presented to the user (it is translatable using gettext prepending a
<code>_</code>.)
</para>
<para>
The attribute <code>section</code>, also translatable, tells the category where
this language should be grouped when it is presented to the user. Currently
available categories in GtkSourceView are "Source", "Script", "Markup",
"Scientific" and "Other".
</para>
<para>
The attribute <code>version</code> specifies the version of the xml syntax
used in your language definition file, so it should always be <code>2.0</code>.
</para>
<para>
The <code><language></code> element contains three sections:
<code><metadata></code>, <code><styles></code> and
<code><definitions></code>
</para>
<programlisting>
<metadata>
</programlisting>
<para>
The <code><metadata></code> element is optional and provides a collection
of properties which specify arbitrary information about the language definition
file itself. It is particularly important to specify the conventional
<code>mimetypes</code> and <code>globs</code> properties that
GtkSourceView uses to automatically detect which syntax highlighting to use
for a given file. They respectively contain a semi-colon separated list of
mimetypes and filename extensions.
</para>
<programlisting>
<metadata>
<property name="mimetypes">text/x-c;text/x-csrc</property>
<property name="globs">*.c</property>
</metadata>
</programlisting>
<programlisting>
<styles>
</programlisting>
<para>
This element contains every association between the styles used in the
description and the defaults stored internally in GtkSourceView.
For each style there is a <code><style></code> element:
</para>
<programlisting>
<style id="comment" _name="Comment" map-to="def:comment"/>
</programlisting>
<para>
This defines a <code>comment</code> style, which inherits the font
properties from the defaults style <code>def:comment</code>.
The <code>name</code> attribute contains the translatable name for this
style, which is the name to show to the user.
</para>
<para>
For each style used in the language definition there is a corresponding
<code><style></code> element; every style can be used in different
contexts, so they will share the same appearance.
</para>
<programlisting>
<style id="string" _name="String" map-to="def:string"/>
<style id="escaped-character" _name="Escaped Character" map-to="def:special-char"/>
<style id="preprocessor" _name="Preprocessor" map-to="def:preprocessor"/>
<style id="included-file" _name="Included File" map-to="def:string"/>
<style id="char" _name="Character" map-to="def:character"/>
<style id="keyword" _name="Keyword" map-to="def:keyword"/>
<style id="type" _name="Data Type" map-to="def:type"/>
</programlisting>
<para>
Following the <code><styles></code> element there is the
<code><definitions></code> element, which contains the
description proper of the syntax:
</para>
<programlisting>
<definitions>
</programlisting>
<para>
Here we should define a main context, the one we enter at the beginning of
the file: to do so we use the <code><context></code> tag, with an
<code>id</code> equal to the <code>id</code> of the
<code><language></code> element:
</para>
<programlisting>
<context id="c">
</programlisting>
<para>
The element <code><include></code> contains the list of sub-contexts
for the current context: as we are in the main context we should put here
the top level contexts for the C language:
</para>
<programlisting>
<include>
</programlisting>
<para>
The first context defined is the one for single-line C style comments: they
start with a double slash <code>//</code> and end at the end of the line:
</para>
<programlisting>
<context id="comment" style-ref="comment">
<start>\/\/</start>
<end>$</end>
</context>
</programlisting>
<para>
The <code><start></code> element contains the regular expression telling
the highlighting engine to enter in the defined context, until the terminating
regular expression contained in the <code><end></code> element is found.
</para>
<para>
Those regular expressions are PCRE regular expressions in the form
<code>/regex/options</code> (see the documentation of PCRE for details.) If
there are no options to be specified and you don't need to match the spaces at
the start and at the end of the regular expression, you can omit the slashes,
putting here only <code>regex</code>.
</para>
<para>
The possible options are:
</para>
<itemizedlist>
<listitem><para>
<code>i</code>: case insensitive;
</para></listitem>
<listitem><para>
<code>x</code>: extended (spaces are ignored and it is possible to put comments
starting with <code>#</code> and ending at the end of the line);
</para></listitem>
<listitem><para>
<code>s</code>: the metacharacter <code>.</code> matches the <code>\n</code>.
</para></listitem>
</itemizedlist>
<para>
You can set the default options using the <code><default-regex-options</code> tag
before the <code><definitions></code> element. To disable a group of options,
instead, you have to preced them with a hyphen (<code>-</code>).
[FIXME: add an example]
</para>
<para>
In GtkSourceView are available also some extensions to the standard perl
style regular expressions:
</para>
<itemizedlist>
<listitem><para>
<code>\%[</code> and <code>\%]</code> are custom word boundaries, which can
be redefined with the <code><keyword-char-class></code> tag (in
contrast with <code>\b</code>);
</para></listitem>
<listitem><para>
<code>\%{id}</code> will include the regular expression defined in the
<code><define-regex></code> tag with the same id, useful if you have
common portions of regular expressions used in different contexts;
</para></listitem>
<listitem><para>
<code>\%{subpattern@start}</code> can be used only inside the
<code><end></code> tag and will be substituted with the
string matched in the corresponding
sub-pattern (can be a number or a name if named sub-patterns are
used) in the preceding <code><start></code> element. For an example
see the implementation of here-documents in the <filename>sh.lang</filename>
language description distribuited with GtkSourceView.
</para></listitem>
</itemizedlist>
<para>
The next context is for C-style strings. They start and end with a double
quote but they can contain escaped double quotes, so we sould make sure
we don't end the string prematurely:
</para>
<programlisting>
<context id="string" end-at-line-end="true" style-ref="string">
</programlisting>
<para>
The <code>end-at-line-end</code> attribute tells the engine that the current context
should be forced to terminate at the end of the line, even if the ending
regular expression is not found, and that an error should be displayed.
</para>
<programlisting>
<start>"</start>
<end>"</end>
<include>
</programlisting>
<para>
To implement the escape handling we include a <code>escape</code> context:
</para>
<programlisting>
<context id="escape" style-ref="escaped-character">
<match>\\.</match>
</context>
</programlisting>
<para>
This is a simple context matching a single regular expression, contained in
the <code><match></code> element. This context will extend its parent, causing the
ending regular expression of the <code>"string"</code> context to not match the escaped
double quote.
</para>
<programlisting>
</include>
</context>
</programlisting>
<para>
Multiline C-style comment can span over multiple lines and cannot be
escaped, but to make things more interesting we want to highlight every
internet address contained:
</para>
<programlisting>
<context id="comment-multiline" style-ref="comment">
<start>\/\*</start>
<end>\*\/</end>
<include>
<context id="net-address" style-ref="net-address" extend-parent="false">
</programlisting>
<para>
In this case, the child should be terminated if the end of the parent is
found, so we use <code>false</code> in the <code>extend-parent</code> attribute.
</para>
<programlisting>
<match>http:\/\/[^\s]*</match>
</context>
</include>
</context>
</programlisting>
<para>
For instance in the following comment the string <code>http://www.gnome.org*/</code>
matches the <code>net-address</code> context but it contains the end of the parent
context (<code>*/</code>.) As <code>extend-parent</code> is false,
only <code>http://www.gnome.org</code> is
highlighted as an address and <code>*/</code> is correctly recognized as the end of
the comment.
</para>
<programlisting>
/* This is a comment http://www.gnome.org */
</programlisting>
<para>
Character constants in C are delimited by single quotes (<code>'</code>) and can
contain escaped characters:
</para>
<programlisting>
<context id="char" end-at-line-end="true" style-ref="string">
<start>'</start>
<end>'</end>
<include>
<context ref="escape"/>
</programlisting>
<para>
The <code>ref</code> attribute is used when we want to reuse a previously defined
context. Here we reuse the <code>escape</code> context defined in the <code>string</code>
context, without repeating its definition.
</para>
<programlisting>
</include>
</context>
</programlisting>
<para>
Using <code>ref</code> it is also possible to refer to contexts defined in other
languages, preceding the id of the context with the id of the containing
language, separating them with a colon:
</para>
<programlisting>
<context ref="def:decimal"/>
<context ref="def:float"/>
</programlisting>
<para>
The definitions for decimal and float constants are in a external file,
with id <code>def</code>, which is not associated with any language but contains
reusable contexts wich every language definition can import.
</para>
<para>
The <code>def</code> language file contains an <code>in-comment</code> context that can contain
addresses and tags such as FIXME and TODO, so we can write a new version of
our <code>comment-multiline</code> context that uses the definitions from <code>def.lang</code>.
</para>
<programlisting>
<context id="comment-multiline" style-ref="comment">
<start>\/\*</start>
<end>\*\/</end>
<include>
<context ref="def:in-comment"/>
</programlisting>
<programlisting>
</include>
</context>
</programlisting>
<para>
Keywords can be grouped in a context using a list of <code><keyword></code>
elements:
</para>
<programlisting>
<context id="keywords" style-ref="keyword">
<keyword>if</keyword>
<keyword>else</keyword>
<keyword>for</keyword>
<keyword>while</keyword>
<keyword>return</keyword>
<keyword>break</keyword>
<keyword>switch</keyword>
<keyword>case</keyword>
<keyword>default</keyword>
<keyword>do</keyword>
<keyword>continue</keyword>
<keyword>goto</keyword>
<keyword>sizeof</keyword>
</context>
</programlisting>
<para>
Keywords with different meaning can be grouped in different context, making
possible to highlight them differently:
</para>
<programlisting>
<context id="types" style-ref="type">
<keyword>char</keyword>
<keyword>const</keyword>
<keyword>double</keyword>
<keyword>enum</keyword>
<keyword>float</keyword>
<keyword>int</keyword>
<keyword>long</keyword>
<keyword>short</keyword>
<keyword>signed</keyword>
<keyword>static</keyword>
<keyword>struct</keyword>
<keyword>typedef</keyword>
<keyword>union</keyword>
<keyword>unsigned</keyword>
<keyword>void</keyword>
</context>
</programlisting>
<para>
You can also set a prefix (or a suffix) common to every keyword using the
<code><prefix></code> and <code><suffix></code> tags:
</para>
<programlisting>
<context id="preprocessor" style-ref="preprocessor">
<prefix>^#</prefix>
</programlisting>
<para>
If not specified, <code><prefix></code> and <code><suffix></code>
are set to, respectively, <code>\%[</code> and
<code>\%]</code>.
</para>
<programlisting>
<keyword>define</keyword>
<keyword>undef</keyword>
</programlisting>
<para>
Keep in mind that every keyword is a regular expression:
</para>
<programlisting>
<keyword>if(n?def)?</keyword>
<keyword>else</keyword>
<keyword>elif</keyword>
<keyword>endif</keyword>
</context>
</programlisting>
<para>
In C, there is a common pratice to use <code>#if 0</code> to express multi-line
nesting comments. To make things easier to the user, we want to highlight
these pseudo-comments as comments:
</para>
<programlisting>
<context id="if0-comment" style-ref="comment">
<start>^#if 0\b</start>
<end>^#(endif|else|elif)\b</end>
<include>
</programlisting>
<para>
As <code>#if 0</code> comments are nesting, we should consider that inside a comment
we can find other <code>#if</code>s with the corresponding <code>#endif</code>s, avoiding
the termination of the comment on the wrong <code>#endif</code>. To do so we use a
nested context, that will extend the parent on every nested
<code>#if</code>/<code>#endif</code>:
</para>
<programlisting>
<context id="if-in-if0">
<start>^#if(n?def)?\b</start>
<end>^#endif\b</end>
<include>
</programlisting>
<para>
Nested contexts can be recursive:
</para>
<programlisting>
<context ref="if-in-if0"/>
</include>
</context>
</include>
</context>
</programlisting>
<para>
Because contexts defined before have higher priority, <code>if0-comment</code> will
never be matched. To make things work we should move it before the
<code>preprocessor</code> context, thus giving <code>if0-comment</code> a higher priority.
</para>
<para>
For the <code>#include</code> preprocessor directive it could be useful to highlight
differently the included file:
</para>
<programlisting>
<context id="include" style-ref="preprocessor">
<match>^#include (".*"|&lt;.*&gt;)</match>
<include>
</programlisting>
<para>
To do this we use grouping sub-patterns in the regular expression,
associating them with a context with the <code>sub-pattern</code> attribute:
</para>
<programlisting>
<context id="included-file" sub-pattern="1"
style-ref="included-file"/>
</programlisting>
<para>
In the <code>sub-pattern</code> attribute we could use:
</para>
<itemizedlist>
<listitem><para>
0: the whole regular expression;
</para></listitem>
<listitem><para>
1: the first sub-pattern (a sub-espression enclosed in
parenthesis);
</para></listitem>
<listitem><para>
2: the second;
</para></listitem>
<listitem><para>
...
</para></listitem>
<listitem><para>
<code>name</code>: a named sub-pattern with name <code>name</code> (see the PCRE
documentation.)
</para></listitem>
</itemizedlist>
<para>
We could also use a <code>where</code> attribute with value
<code>start</code> or <code>end</code> to
specify the regular expression the context is referring, when we have both
the <code><start></code> and <code><end></code> element.
</para>
<programlisting>
</include>
</context>
</programlisting>
<para>
Having defined a good subset of the C syntax we close every tag still open:
</para>
<programlisting>
</include>
</context>
</definitions>
</language>
</programlisting>
</refsect1>
<refsect1>
<title>The full language definition</title>
<para>
This is the full language definition for the subset of C taken in consideration
for this tutorial:
</para>
<programlisting>
<?xml version="1.0" encoding="UTF-8"?>
<language id="c" _name="C" version="2.0" _section="Source">
<metadata>
<property name="mimetypes">text/x-c;text/x-csrc</property>
<property name="globs">*.c</property>
</metadata>
<styles>
<style id="comment" _name="Comment" map-to="def:comment"/>
<style id="string" _name="String" map-to="def:string"/>
<style id="escaped-character" _name="Escaped Character" map-to="def:special-char"/>
<style id="preprocessor" _name="Preprocessor" map-to="def:preprocessor"/>
<style id="included-file" _name="Included File" map-to="def:string"/>
<style id="char" _name="Character" map-to="def:character"/>
<style id="keyword" _name="Keyword" map-to="def:keyword"/>
<style id="type" _name="Data Type" map-to="def:type"/>
</styles>
<definitions>
<context id="c">
<include>
<context id="comment" style-ref="comment">
<start>\/\/</start>
<end>$</end>
</context>
<context id="string" end-at-line-end="true" style-ref="string">
<start>"</start>
<end>"</end>
<include>
<context id="escape" style-ref="escaped-character">
<match>\\.</match>
</context>
</include>
</context>
<context id="comment-multiline" style-ref="comment">
<start>\/\*</start>
<end>\*\/</end>
<include>
<context ref="def:in-comment"/>
</include>
</context>
<context id="char" end-at-line-end="true" style-ref="string">
<start>'</start>
<end>'</end>
<include>
<context ref="escape"/>
</include>
</context>
<context ref="def:decimal"/>
<context ref="def:float"/>
<context id="keywords" style-ref="keyword">
<keyword>if</keyword>
<keyword>else</keyword>
<keyword>for</keyword>
<keyword>while</keyword>
<keyword>return</keyword>
<keyword>break</keyword>
<keyword>switch</keyword>
<keyword>case</keyword>
<keyword>default</keyword>
<keyword>do</keyword>
<keyword>continue</keyword>
<keyword>goto</keyword>
<keyword>sizeof</keyword>
</context>
<context id="types" style-ref="type">
<keyword>char</keyword>
<keyword>const</keyword>
<keyword>double</keyword>
<keyword>enum</keyword>
<keyword>float</keyword>
<keyword>int</keyword>
<keyword>long</keyword>
<keyword>short</keyword>
<keyword>signed</keyword>
<keyword>static</keyword>
<keyword>struct</keyword>
<keyword>typedef</keyword>
<keyword>union</keyword>
<keyword>unsigned</keyword>
<keyword>void</keyword>
</context>
<context id="if0-comment" style-ref="comment">
<start>^#if 0\b</start>
<end>^#(endif|else|elif)\b</end>
<include>
<context id="if-in-if0">
<start>^#if(n?def)?\b</start>
<end>^#endif\b</end>
<include>
<context ref="if-in-if0"/>
</include>
</context>
</include>
</context>
<context id="preprocessor" style-ref="preprocessor">
<prefix>^#</prefix>
<keyword>define</keyword>
<keyword>undef</keyword>
<keyword>if(n?def)?</keyword>
<keyword>else</keyword>
<keyword>elif</keyword>
<keyword>endif</keyword>
</context>
<context id="include" style-ref="preprocessor">
<match>^#include (".*"|&lt;.*&gt;)</match>
<include>
<context id="included-file"
sub-pattern="1"
style-ref="included-file"/>
</include>
</context>
</include>
</context>
</definitions>
</language>
</programlisting>
</refsect1>
</refentry>