Skip to content

Writing a Syntax Highlighting File

Thursday, 24 March 2005 | Christoph Cullmann

Note: Please refer to the Kate Handbook for the most recent version of how to write syntax highlighting files.

Hint: If you want to write a syntax highlighting file, the XML Completion plugin might be of great help.

This section is an overview of the Highlight Definition XML format in KDE4. Based on a small example it will describe the main components and their meaning and usage. The next section will go into detail with the highlight detection rules.

Main sections of Kate Highlight Definition files

The formal definition, aka the DTD is stored in the file language.dtd which should be installed on your system in the folder $KDEDIR/share/apps/katepart/syntax. If $KDEDIR is unset look up the folder by using kde4-config --prefix.

An Example

A highlighting file contains a header that sets the XML version and the doctype:

<?xml version="1.0" encoding="UTF-8"?> <!DOCTYPE language SYSTEM "language.dtd"> 

The root of the definition file is the element language. Available attributes are:

Required attributes:

  • name sets the name of the language. It appears in the menus and dialogs.
  • section specifies the category.
  • extensions defines file extensions, like “.cpp;.h”

Optional attributes:

  • mimetype associates files Mime Type based.
  • version specifies the current version of the definition file.
  • kateversion specifies the latest supported version of Kate.
  • casesensitive defines, whether the keywords are case sensitive or not. NOTE: not implemented yet.
  • priority is necessary if another highlight definition file uses the same extensions. The higher priority will win.
  • author contains the name of the author and his email-address.
  • license contains the license, usually LGPL, Artistic, GPL and others. It’s important to specify a license, as the kate team needs some legal backing for the distribution of the files.
  • hidden defines, whether the name should appear in Kate’s menus.

So the next line may look like this:

<language name="C++" version="1.00" kateversion="2.4" section="Sources" extensions="*.cpp;*.h" > 

Next comes the highlighting element, which contains the optional element list and the required elements contexts and itemDatas.

list elements contain a list of keywords. In this case the keywords are class and const. You can add as many lists as you need. The contexts element contains all contexts. The first context is by default the start of the highlighting. There are two rules in the context Normal Text, which match the list of keywords with the name somename and a rule that detects a quote and switches the context to string. To learn more about rules read the next chapter.
The third part is the itemDatas element. It contains all color and font styles needed by the contexts and rules.
In this example, the itemData Normal Text, String and Keyword are used.

 <highlighting> <list name="somename"> <item> class </item> <item> const </item> </list> <contexts> <context attribute="Normal Text" lineEndContext="#pop" name="Normal Text" > <keyword attribute="Keyword" context="#stay" String="somename" /> <DetectChar attribute="String" context="string" char="&quot;" /> </context> <context attribute="String" lineEndContext="#stay" name="string" > <DetectChar attribute="String" context="#pop" char="&quot;" /> </context> </contexts> <itemDatas> <itemData name="Normal Text" defStyleNum="dsNormal" /> <itemData name="Keyword" defStyleNum="dsKeyword" /> <itemData name="String" defStyleNum="dsString" /> </itemDatas> </highlighting> 

The last part of a highlight definition is the optional general section. It may contain information about keywords, code folding, comments and indentation.

The comment section defines with what string a single line comment is introduced. You also can define a multiline comments using multiLine with the additional attribute end. This is used if the user presses the corresponding shortcut for comment/uncomment.
The keywords section defines whether keyword lists are case sensitive or not. Other attributes will be explained later.

 <general> <comments> <comment name="singleLine" start="#"/> </comments> <keywords casesensitive="1"/> </general> </language> 

The Sections in Detail

This part will describe all available attributes for contexts,
itemDatas, keywords, comments, code folding and indentation.

The element context belongs into the group contexts. A context itself defines context specific rules like what should happen if the highlight system reaches the end of a line. Available attributes are:

  • name the context name/identifier. Rules will use this name to specify the context to switch to if the rule matches.
  • attribute the default item data that is used if no rules match in the current context.
  • lineEndContext defines the context the highlight system switches to if it reaches the end of a line. This may either be a name of another context, #stay to not switch the context (i.e. do nothing) or #pop which will cause to leave this context. It is possible to use for example #pop#pop#pop to pop three times.
  • lineBeginContext defines the context if a begin of a line is encountered. Default: #stay.
  • fallthrough defines if the highlight system switches to the context specified in fallthroughContext if no rule matches. Default: false.
  • fallthroughContext specifies the next context if no rule matches.
  • dynamic if true, the context remembers strings/placeholders saved by dynamic rules. This is needed for HERE documents for example. Default: false.

The element itemData is in the group itemDatas. It defines the font style and colors. So it is possible to define your own styles and colors, however we recommend to stick to the default styles if possible so that the user will always see the same colors used in different languages. Though, sometimes there is no other way and it is necessary to change color and font attributes. The attributes name and defStyleNum are required, the other optional. Available attributes are:

  • name sets the name of the itemData. Contexts and rules will use this name in their attribute.
  • attribute to reference an itemData.
  • defStyleNum defines which default style to use. Available default styles are explained in detail later.
  • color defines a color. Valid formats are ‘#rrggbb’ or ‘#rgb’.
  • selColor defines the selection color.
  • italic if true, the text will be italic.
  • bold if true, the text will be bold.
  • underline if true, the text will be underlined.
  • strikeout if true, the text will be stroked out.
  • spellChecking if true, the text will be spell checked, otherwise it will be ignored during spell check.

The element keywords in the group general defines keyword properties. Available attributes are:

  • casesensitive may be true or false. If true, all keywords are matched case sensitive. Default: true.
  • weakDeliminator is a list of characters that do not act as word delimiters. For example the dot (.) is a word delimiter. Assume a keyword in a list contains a dot, it will only match if you specify the dot as a weak delimiter.
  • additionalDeliminator defines additional delimiters.
  • wordWrapDeliminator defines characters after which a line wrap may occur. Default delimiters and word wrap delimiters are the characters .():!+,-<=>%&_/;?[]^{|}~_*, space (‘ ‘) and tabulator (\t).

The element comment in the group comments defines comment properties which are used
for Tools > Comment and Tools > Uncomment.
Available attributes are:

  • name is either singleLine or multiLine.
  • If you choose singleLine the optinal attribute position is available. Default for this attribute is to insert the single line comment string in column 0. If you want it to appear after the whitespaces you have to set it to afterwhitespace, like: position="afterwhitespace".
  • If you choose multiLine the attributes end and region are required.
  • start defines the string used to start a comment. In C++ this is /*.
  • end defines the string used to close a comment. In C++ this is */.
  • region should be the name of the the foldable multiline comment. Assume you have beginRegion=”Comment”endRegion=”Comment” in your rules, you should use region=”Comment”. This way uncomment works even if you do not select all the text of the multiline comment. The cursor only must be in the multiline comment.

The element folding in the group general defines code folding properties. Available attributes are:

  • indentationsensitive if true, the code folding markers will be added indentation based, like in the scripting language Python. Usually you do not need to set it, as it defaults to false.

The element indentation in the group general defines which indenter will be used, however we strongly recommend to omit this element, as the indenter usually will be set by either defining a File Type or by adding a mode line to the text file. If you specify an indenter though, you will force a specific indentation on the user, which he might not like at all.
Available attributes are:

  • mode is the name of the indenter. Available indenters right now are: none, normal, cstyle, haskell, lilypond, lisp, python, ruby and xml.

Available Default Styles

Default styles are predefined font and color styles. For convenience Kate provides several default styles, in detail:

  • dsNormal, used for normal text.
  • dsKeyword, used for keywords.
  • dsDataType, used for data types.
  • dsDecVal, used for decimal values.
  • dsBaseN, used for values with a base other than 10.
  • dsFloat, used for float values.
  • dsChar, used for a character.
  • dsString, used for strings.
  • dsComment, used for comments.
  • dsOthers, used for ‘other’ things.
  • dsAlert, used for warning messages.
  • dsFunction, used for function calls.
  • dsRegionMarker, used for region markers.
  • dsError, used for error highlighting and wrong syntax.

Highlight Detection Rules

This section describes the syntax detection rules.

Each rule can match zero or more characters at the beginning of the string they are tested against. If the rule matches, the matching characters are assigned the style or attribute defined by the rule, and a rule may ask that the current context is switched.

A rule looks like this:

The attribute identifies the style to use for matched characters by name, and the context identifies the context to use from here.

The context can be identified by:

  • An identifier, which is the name/identifier of another context.
  • An order telling the engine to stay in the current context (#stay), or to pop back to a previous context used in the string (#pop). To go back more steps, the #pop keyword can be repeated: #pop#pop#pop

Some rules can have child rules which are then evaluated only if the parent rule matched. The entire matched string will be given the attribute defined by the parent rule. A rule with child rules looks like this:

<RuleName (attributes)> <ChildRuleName (attributes) /> ... </RuleName> 

Rule specific attributes vary and are described in the following sections.

Common attributes

All rules have the following attributes in common and are available whenever a (common attributes) appears. All following attributes are optional.

  • attribute maps to a defined itemData. Default: the attribute from the destination context
  • context specifies the context to which the highlighting system switches if the rule matches. Default: #stay
  • beginRegion starts a code folding block. Default: unset.
  • endRegion closes a code folding block. Default: unset.
  • lookAhead, if true, the highlighting system will not process the matches length. Default: false.
  • firstNonSpace, if true, the rule only matches, if the string is the first non-whitespace in the line. Default: false.
  • column defines the column. The rule only matches, if the current column matches the given one. Default: unset.

Dynamic rules

Some rules allow the optional attribute dynamic of type boolean that defaults to false. If dynamic is true, a rule can use placeholders representing the text matched by a regular expression rule that switched to the current context in its string or char attributes. In a string, the placeholder %N (where N is a number) will be replaced with the corresponding capture N from the calling regular expression. In a char the placeholer must be a number N and it will be replaced with the first character of the corresponding capture N from the calling regular expression. Whenever a rule allows this attribute it will contain a (dynamic).

  • dynamic may be either true or false. Default: false.

The Rules in Detail


Detect a single specific character. Commonly used for example to find the ends of quoted strings.

<DetectChar char="(character)" (common attributes) (dynamic) /> 

The char attribute defines the character to match.


Detect two specific characters in a defined order.

<Detect2Chars char="(character)" char1="(character)" (common attributes) (dynamic) /> 

The char attribute defines the first character to match, char1 the second.


Detect one character of a set of specified characters.

<AnyChar String="(string)" (common attributes) /> 

The String attribute defines the set of characters.


Detect an exact string.

<StringDetect String="(string)" [insensitive="true|false"] (common attributes) (dynamic) /> 

The String attribute defines the string to match. The insensitive attribute defaults to false and is passed to the string comparison function. If the value is true insensitive comparing is used.

WordDetect (KDE >= 4.5, Kate >= 3.5)

Detect an exact string but additionally require word boundaries like a dot (.) or a whitespace on the beginning and the end of the word. You can think of \b\b in terms of a regular expression.

<WordDetect String="(string)" [insensitive="true|false"] (common attributes) (dynamic) /> 

The String attribute defines the string to match. The insensitive attribute defaults to false and is passed to the string comparison function. If the value is true insensitive comparing is used.


Matches against a regular expression.

<RegExpr String="(string)" [insensitive="true|false"] [minimal="true|false"] (common attributes) (dynamic) /> 
  • The String attribute defines the regular expression.
  • insensitive defaults to false and is passed to the regular expression engine.
  • minimal defaults to false and is passed to the regular expression engine.

Because the rules are always matched against the beginning of the current string, a regular expression starting with a caret (^) indicates that the rule should only be matched against the start of a line.


Detect a keyword from a specified list.

<keyword String="(list name)" (common attributes) /> 

The String attribute identifies the keyword list by name. A list with that name must exist.


Detect an integer number.

<Int (common attributes) (dynamic) /> 

This rule has no specific attributes. Child rules are typically used to detect combinations of L and U after the number, indicating the integer type in program code. Actually all rules are allowed as child rules, though, the DTD only allowes the child rule StringDetect.
The following example matches integer numbers follows by the character ‘L’.

<Int attribute="Decimal" context="#stay" > <StringDetect attribute="Decimal" context="#stay" String="L" insensitive="true"/> </Int> 


Detect a floating point number.

<Float (common attributes) /> 

This rule has no specific attributes. AnyChar is allowed as a child rules and typically used to detect combinations, see rule Int for reference.


Detect an octal point number representation.

<HlCOct (common attributes) /> 

This rule has no specific attributes.


Detect a hexadecimal number representation.

<HlCHex (common attributes) /> 

This rule has no specific attributes.


Detect an escaped character.

<HlCStringChar (common attributes) /> 

This rule has no specific attributes.

It matches literal representations of characters commonly used in program code, for example \n (newline) or \t (tabulator).

The following characters will match if they follow a backslash (***): **abefnrtv”’?*. Additionally, escaped hexadecimal numbers like for example \xff and escaped octal numbers, for example \033 will match.


Detect an C character.

<HlCChar (common attributes) /> 

This rule has no specific attributes.

It matches C characters enclosed in a tick (Example: ‘c’). So in the ticks may be a simple character or an escaped character. See HlCStringChar for matched escaped character sequences.


Detect a string with defined start and end characters.

<RangeDetect char="(character)" char1="(character)" (common attributes) /> 

char defines the character starting the range, char1 the character ending the range. Usefull to detect for example small quoted strings and the like, but note that since the highlighting engine works on one line at a time, this will not find strings spanning over a line break.


Matches a backslash (‘&#8217;) at the end of a line.

<LineContinue (common attributes) /> 

This rule has no specific attributes.
This rule is useful for switching context at end of line, if the last character is a backslash (‘****’). This is needed for example in C/C++ to continue macros or strings.


Include rules from another context or language/file.

<IncludeRules context="contextlink" [includeAttrib="true|false"] /> 

The context attribute defines which context to include.
If it a simple string it includes all defined rules into the current context, example:

<IncludeRules context="anotherContext" /> 

If the string begins with ## the highlight system will look for another language definition with the given name, example:

<IncludeRules context="##C++" /> 

If includeAttrib attribute is true, change the destination attribute to the one of the source. This is required to make for example commenting work, if text matched by the included context is a different highlight than the host context.


Detect whitespaces.

<DetectSpaces (common attributes) /> 

This rule has no specific attributes.
Use this rule if you know that there can several whitespaces ahead, for example in the beginning of indented lines. This rule will skip all whitespace at once, instead of testing multiple rules and skipping one at the time due to no match.


Detect identifier strings (as a regular expression: [a-zA-Z_][a-zA-Z0-9_]*).

<DetectIdentifier (common attributes) /> 

This rule has no specific attributes.
Use this rule to skip a string of word characters at once, rather than testing with multiple rules and skipping one at the time due to no match.

Tips & Tricks

Once you have understood how the context switching works it will be easy to write highlight definitions. Though you should carefully check what rule you choose in what situation. Regular expressions are very mighty, but they are slow compared to the other rules. So you may consider the following tips.

  • If you only match two characters use Detect2Chars instead of StringDetect. The same applies to DetectChar.

  • Regular expressions are easy to use but often there is another much faster way to achieve the same result. Consider you only want to match the character **#**if it is the first character in the line. A regular expression based solution would look like this:

    You can achieve the same much faster in using:

    <DetectChar attribute="Macro" context="macro" char="#" firstNonSpace="true" /> 

    If you want to match the regular expression ‘^#’ you can still use DetectChar with the attribute column=”0″. The attribute column counts character based, so a tabulator still is only one character.

    • You can switch contexts without processing characters. Assume that you want to switch context when you meet the string /*, but need to process that string in the next context. The below rule will match, and the lookAheadattribute will cause the highlighter to keep the matched string for the next context.

    • Use DetectSpaces if you know that many whitespaces occur.

    • Use DetectIdentifier instead of the regular expression ‘[a-zA-Z_]\w*’.

    • Use default styles whenever you can. This way the user will find a familiar environment.

    • Look into other XML-files to see how other people implement tricky rules.

    • You can validate every XML file by using the command xmllint --dtdvalid language.dtd mySyntax.xml.

    • If you repeat complex regular expression very often you can use ENTITIES. Example: ]>

      Now you can use &myref; instead of the regular expression.