jStyleParser Manual

Radek Burget and Karel Piwko

Table of Contents


jStyleParser is a Java library for parsing CSS style sheets and assigning styles to the HTML or XML document elements according to the W3C CSS 2.1 specification and a subset of the CSS 3 specification. It allows parsing the individual CSS files as well as computing the efficient style of the DOM elements.

Style Sheet Parsing

This functionality may be used for parsing individual style sheets obtained from a remote file (URL), local file or a string. Three static methods are defined in the CSSFactory class for this purpose:

During the parsing process, the parser automatically imports all the style sheets referenced using the @import rules. See the Media section for further reference about how to limit this behavior to certain media only or disable it completely.

For obtaining the imported style sheets referenced by their URLs, it is possible to provide a custom implementation of a NetworkProcessor that simply provides obtaining an InputStream from the given URL. The default built-in NetworkProcessor is based on the standard Java URLConnection infrastructure.

The result of the parsing is a StyleSheet object that is generally a collection of rules discovered in the style sheet. It can be also used as the input for the DOM Analyzer as described below in the DOM Analysis section.

Parsed style sheet processing

When the DOM Analyzer feature is not used, the parsed style sheed may be manually traversed in order to obtain the contained rules and declarations. The most important data structures are the following:

A simple example of the style sheet processing is available in the test.ParserDemo example.

Obtaining the style sheets used in an HTML document

The CSSFactory provides a method that analyzes an HTML or XML document represented by a DOM and it parses all the referenced style sheets:

It parses all the style sheets referenced from the document that correspond to the specified media and it returns a single style sheet structure that contains all the relevant CSS rules. The base URL is used for the possible relative URLs used in the style sheets. The encoding specifies a default encoding that is used when the encoding is not specified within the style sheet. Finally, media specifies the current media features that should be used for evaluating the media queries used in the style sheet as described in the Media section.

DOM Analysis

The purpose of the DOM analysis is to apply the relevant style sheets to a particular HTML or XML document represented by a DOM in order to obtain the efficient styles of the individual document elements. The result of the analysis is a mapping between the DOM elements and the corresponding CSS declarations. This mapping can be used either for displaying the HTML document or for performing some further analysis on the document structure.

Analyzing a Style Sheet

When a StyleSheet instance is obtained from the parser, it can be passed to an Analyzer.

The Analyzer basically provides the following method:

Additionally, a DirectAnalyzer analyzer class is provided for cases when it is not necessary to evaluate the whole DOM. It computes a style for individual DOM nodes without creating the whole map. It is suitable for obtaining the style of individual elements without computing the style for the whole DOM tree. However, in larger scale, the performance of the individual computation is significantly worse.

Simplified and Direct Usage Method

To provide simpler approach while parsing an (X)HTML document, CSSFactory provides the following method:

It automatically downloads and parses all the internal and external style sheets referenced from a DOM for the given media and runs the style mapping to the given DOM. It creates and assigns a NodeData to each element in the DOM document doc for the given medium media. While searching for externally stored CSS style sheets, base URL base is used.

There exist several variants of the assignDOM() method in the CSSFactory class that allow to specify a custom MatchCondition for matching pseudoclasses (as discussed in the pseudoclasses section) and/or a custom NetworkProcessor for obtaining the imported style sheets as discussed in the parsing section.

Retrieving the Style of DOM Elements

When the analyzing part is done for the style sheet, the computed mapping between DOM elements and the NodeData structures representing their styles is available as a StyleMap structure. This structure extends the standard Java Map structure and the style of a particular DOM element may be therefore obtained using the following method:

The NodeData structure provides two basic methods:

For both these methods, there are equivalent ones defined with automatic inclusion of inherited properties/terms.

Example of enum values for the CSSProperty max-height: length, percentage, NONE, INHERIT

Value INHERIT is present for all properties, length and percentage determine type of token which is about to be retrieved to get exact information about style

For determining the type of the CSSProperty, compiler inference is used. Strictly speaking that means that the type of L-value(expression at the left side of equal-sign) is used to determine type to which the result is casted. This could lead in ClassCastException in cases when the user uses invalid combination of L-value type and property name.
When there is no L-value, the supertype (that is CSSProperty) is used to cast the resulting expression. This is always valid cast.

The following example shows how to obtain the value of a top margin of an element:

//get the element style
StyleMap map = CSSFactory.assignDOM(doc, encoding, base, medium, true);
NodeData style = map.get(element);
//get the type of the assigned value
CSSProperty.Margin mm = style.getProperty("margin-top");
System.out.println("margin-top=" + mm);
//if a length is specified, obtain the exact value
if (mm == Margin.length)
    TermLength mtop = style.getValue(TermLength.class, "margin-top");
    System.out.println("value=" + mtop);

Obtaining the Style of Pseudo-Elements

CSS specification allows the use of several pseudo-elements for addressing specific parts of the existing DOM elements. The style of the pseudo-elements may be accessed using the following method of the StyleMap obtained for the DOM tree:

The pseudo-elements are specified by a Selector.PseudoDeclaration pseudo-element value and may be one of the following: BEFORE, AFTER, FIRST_LETTER or FIRST_LINE. The remaining values of Selector.PseudoDeclaration correspond to pseudo-classes that must be treated differently as described in the following section.

Applying Pseudo-Classes

jStyle parser supports a subset of the available CSS3 pseudo-classes: the structural pseudo-classes and the dynamic pseudo-classes.

The structural pseudo-classes (such as :first-child) are supported and evaluated automatically. Their defined style is automatically included in the resulting style assigned to the appropriate DOM elements in the resulting StyleMap.

The dynamic pseudo-classes (such as :hover) are more complicated. Any element may belong dynamically to several pseudo-classes that influence the resulting style of the element itself but also the style of its child elements. Therefore, before the DOM style is evaluated as described in previous sections, the current pseudo-classes must be assigned to the individual elements in order to compute the resulting styles properly.

The default behavior of jStyleParser corresponds to the standard static HTML file displaying behavior:

For specifying other pseudo-classes for different elements, a special MatchCondition mechanism may be used. Generally, a MatchCondition specifies an additional condition applied when matching specific parts of the CSS selectors. Its default implementation MatchConditionImpl implements the default behavior described above. For implementing a better behavior, a configurable MatchConditionOnElements implementation is prepared. It allows do assign a set of pseudo-classes directly to given DOM elements or to specified element names. It usage is demonstrated on the following code:

//obtain the elements e1 and e2 that should be assigned the style. e.g.:
Element e1 = document.getElementById('element1');
Element e2 = ... //or any other way of obtaining a DOM Element

//Create the match condition. Preserve the default behavior:
//  all <a> links are assigned the :link class
MatchConditionOnElements cond = new MatchConditionOnElements("a", PseudoDeclaration.LINK);

//assign pseudo-classes to the selected elements
cond.addMatch(e1, PseudoDeclaration.HOVER);
cond.addMatch(e2, PseudoDeclaration.VISITED);

//register the match condition so that it is used by jStyleParser

//evaluate the DOM styles as normally
StyleMap decl = CSSFactory.assignDOM(doc, null, base, "screen", true);

When the pseudo-class assignment changes, the match condition must be reconfigured and the DOM style must be recomputed.

Alternatively, it is possible to pass the MatchCondition directly to the DOM analyzer instead of using the global registerDefaultMatchCondition() function. When the Analyzer is used for evaluating the DOM (as discussed here), the match condition may be registered using its registerMatchCondition function. When using the similified method (using CSSFactory), the match condition may be passed as an additional argument to the CSSFactory.assignDOM() method.


The style sheets are typically evaluated for some particular media. In jStyleParser the type and the features of the media being used are specified using the MediaSpec structure. It specifies the media type such as "screen" or "print" and the features such as display area size, device size, colors, etc. A simple media specification may be created as follows:

//create a 'screen' media type with default feature values
MediaSpec media = new MediaSpec("screen");

//set the display area size in pixeld
media.setDimensions(1600, 1200);

The created media specification may be used for several parsing and analysis steps as described in the following sections.

Automatic loading of imported style sheets

By default, the parser automatically loads and recursively parses all the style sheets imported using the @import rules. The optional media queries used in the @import rules are evaluated later during the DOM analysis. However, the automatic download may be limited to certain media only by using the setAutoImportMedia(cz.vutbr.web.css.MediaSpec media) method. The following code shows several typical configuration.

//import only the style sheets valid for "screen" media type with default values of all the features
CSSFactory.setAutoImportMedia(new MediaSpec("screen"));

//import all the style sheets valid for "screen" media type regardless on the feature values
CSSFactory.setAutoImportMedia(new MediaSpecType("screen"));

//import all style sheets (the default behavior)
CSSFactory.setAutoImportMedia(new MediaSpecAll());

//do not import any style sheets automatically
CSSFactory.setAutoImportMedia(new MediaSpecNone());

Note that this specification only affects the behavior of the @import rules during parsing. The media used for assigning the style to the DOM elements are specified in the further step.

DOM style analysis

The media specification is used in the following steps of the DOM analysis:

The default behavior of the parser is to download all the imported style sheets and to apply the media during the DOM analysis. However, the above methods allow to optimize this process by downloading only the style sheets that are actually used.

Internal Structure of the Library

The code is divided into following packages:

  1. cz.vutbr.web.css,
  2. cz.vutbr.web.csskit,
  3. cz.vutbr.web.domassign and
  4. test.

The first package provides in general the contracts used in other applications, the second one contains its implementation. The third package is about assigning CSS rules to HTML elements. The last one contains test units.

Package cz.vutbr.web.css

In addition to several implementation interfaces, this package provides a general entry point of the jStyleParser library. It is the CSSFactory class. By default, the interfaces use an implementation provided by cz.vutbr.csskit package.

Another remarkable class in this package is the CSSProperty interface, which provides a base for CSS properties. By implementing this interface, new CSS properties can be added.

Package cz.vutbr.web.csskit

This package provides a default implementation of cz.vutbr.web.css. This can be changed by registering other implementation by calling the appropriate methods of CSSFactory.

Internally, it uses ANTLR to parse CSS input into structures defined by contracts in the package cz.vutbr.web.css.

Package cz.vutbr.web.domassign

This package provides among others an Analyzer class, which is able sort the CSS declarations, to classify them according to a CSS medium and finally, to assign them to the DOM elements.

The transformation core is implemented in the DeclarationTransformer class.

Extending Current Version

An extra work was done to simplify the implementation of new CSS standards. The implementation is defined by interface contracts. To implement another CSS parsing level, additional work must be done:

Extending Performance

During the implementation, some additional storage methods have been tested for NodeData storage considering time and spatial complexity. Current implementation is a compromise, specialized to lower memory usage. All parts of the library can be changed to use a different implementation by changing the CSSFactory. Other factories that can be replaced are factory for creating terms (TermFactory) and factory for creating CSS rule parts (RuleFactory).