jfiveparse: a java html5 parser

jfiveparse pass all the non-scripted tests for the tokenizer and tree construction from the html5lib-tests suite.

It provides both fragment and full document parsing. It can parse directly from a String or by streaming through a Reader (note: the encoding must be known, currently the parser does not implement an autodetect feature).

Version 1.1.4 and older require Java 11. Version 2.x.x require Java 17.

Javadoc@javadoc.io.

Features

• a html 5 parser with a compact footprint with 0 dependencies (jar is around 175kb)
• css selector parser and matcher (that works on both the "jfiveparse" types or org.w3c.dom.* nodes)
• a typesafe selector builder and matcher
• a converter from the internal "jfiveparse" Document to org.w3c.dom.Document

License

jfiveparse is licensed under the Apache License Version 2.0.

Download

Stable

maven:

<dependency>
    <groupId>ch.digitalfondue.jfiveparse</groupId>
    <artifactId>jfiveparse</artifactId>
    <version>1.1.4</version>
</dependency>

gradle:

compile 'ch.digitalfondue.jfiveparse:jfiveparse:1.1.4'

Milestone releases

maven:

<dependency>
    <groupId>ch.digitalfondue.jfiveparse</groupId>
    <artifactId>jfiveparse</artifactId>
    <version>2.0.0-M3</version>
</dependency>

gradle:

compile 'ch.digitalfondue.jfiveparse:jfiveparse:2.0.0-M3'

Use:

If you use it as a module, remember to add requires ch.digitalfondue.jfiveparse; in your module-info. If you are using the W3CDom class (and the various inner classes), you may also need to require the java.xml module, as it's an optional dependency.

Examples:

Parse, select and print all title from HN

package ch.digitalfondue.jfiveparse.example;

import ch.digitalfondue.jfiveparse.*;

import java.io.IOException;
import java.io.InputStreamReader;
import java.io.Reader;
import java.net.URI;
import java.nio.charset.StandardCharsets;

public class LoadHNTitle {

    public static void main(String[] args) throws IOException {
        try (Reader reader = new InputStreamReader(URI.create("https://news.ycombinator.com/").toURL().openStream(), StandardCharsets.UTF_8)) {
            NodeMatcher<Node> matcher = Selector.parseSelector("td.title > span.titleline > a");
            JFiveParse.parse(reader).getAllNodesMatchingAsStream(matcher)
                    .map(Element.class::cast)
                    .forEach(a -> System.out.printf("%s [%s]\n", a.getTextContent(), a.getAttribute("href")));
        }
    }
}

Type safe selectors:

See https://github.com/digitalfondue/jfiveparse/blob/master/src/test/java/ch/digitalfondue/jfiveparse/NodeMatchersTest.java

Other examples

See directory: https://github.com/digitalfondue/jfiveparse/tree/master/src/test/java/ch/digitalfondue/jfiveparse/example

Convert to the java DOM representation

If you need to generate a org.w3c.dom.Document from the ch.digitalfondue.jfiveparse.Document representation, there is a static method in the helper class: W3CDom.toW3CDocument.

Notes:

Specs/Doc:

• html by whatwg: https://html.spec.whatwg.org/multipage/syntax.html (of interest: tokenization, tree-construction)
• entities: https://html.spec.whatwg.org/entities.json

Template element handling

The template element is a "normal" element, so the child nodes are not placed inside a documentFragment. This will be fixed.

Special parsing options

The parser can be customized to allow some non-standard behaviour, you can see the following tests: https://github.com/digitalfondue/jfiveparse/blob/master/src/test/java/ch/digitalfondue/jfiveparse/OptionParseTest.java

• DISABLE_IGNORE_TOKEN_IN_BODY_START_TAG : allow to have for example "tr" tag without the containing table/tbody.
• INTERPRET_SELF_CLOSING_ANYTHING_ELSE : When encountering unknown self-closing tag, they will be interpreted as it is and not as open tag only, thus creating a non-intuitive DOM.

Entities

The &ntities; are by default (and by specification) parsed and interpreted. This behavior can be disabled by:

• passing the enum "Option.DONT_TRANSFORM_ENTITIES" to the Parser
• when calling Node.get{Inner,Outer}HTML(), pass the enum "Option.DONT_TRANSFORM_ENTITIES" for disabling the escaping. It's possible that something will be wrong in the generated document.

Preserving as much as possible the original document when serializing

By default, when parsing/serializing, the following transformations will be applied:

• entities will be interpreted and converted
• all the attribute values will be double-quoted
• tag and attribute names will be lower-case
• the "/" character used in self-closing tag will be ignored
• some whitespace will be ignored

Currently, jfiveparse can preserve the entities, the attribute quoting type and the case and the tag name case.

If you require to preserve as much as possible the document when serializing back in a string, pass the following parameters:

• pass the enum "Option.DONT_TRANSFORM_ENTITIES" to the Parser
• when calling Node.get{Inner,Outer}HTML(), pass the enums:
  • Option.DONT_TRANSFORM_ENTITIES
  • Option.PRINT_ORIGINAL_ATTRIBUTE_QUOTE
  • Option.PRINT_ORIGINAL_ATTRIBUTES_CASE
  • Option.PRINT_ORIGINAL_TAG_CASE

Uppercase handling in the tokenizer

Note: this is a deviation from the specification in term of implementation of the tokenizer, but globally, the end result is correct, as the attributes and tag names are then converted to lower case.

In the tokenizer, instead of applying the toLowerCase function on each character, the transformation is done in a single call in the TreeConstructor (see setTagName). This is used for saving the original case of the attributes and tag names.

TODO:

• additional doc
• expand the typesafe matcher api
• keep track of lines, eventually chars too
• profile
  • various optimizations...
  • TokenizerRCDataStates.handleRCDataState could be optimized
    • (textarea related)

mvn clean test jacoco:report