Index of /sandbox/html4strict

[ICO]NameLast modifiedSizeDescription
[PARENTDIR]Parent Directory   -  
[DIR]xhtml11/ 2018-06-17 08:52 -  
[DIR]tools/ 2013-03-22 08:14 -  
[DIR]test/ 2018-06-17 08:52 -  
[TXT]rst2xhtml11.py 2015-08-18 08:37 1.2K 
[TXT]rst2xhtml.py 2020-11-06 19:44 1.2K 
[DIR]html4strict/ 2013-03-22 08:14 -  
[DIR]docs/ 2012-01-28 09:45 -  
[DIR]data/ 2018-06-17 08:52 -  
The XHTML 1.1 Writer

The XHTML 1.1 Writer

Author:

Günter Milde

Date:
2016-10-20

Abstract

A HTML writer, generating XHTML 1.1 for styling with CSS 2.1.

1 Introduction

1.1 State of the art

Docutils' default HTML Writer, docutils.writers.html4css1 generates output that conforms to the HTML 4.01 Transitional DTD and to the Extensible HTML version 1.0 Transitional DTD (almost strict).

Almost, as it contains some deprecated constructs and "a minimum of formatting information" in order to ensure correct display with deficient but (at the time of creation) widespread browsers (mainly IE6).

A new HTML5 writer with most features of this writer may become part of Docutils 0.13.

1.2 Objective

Goals of the xhtml11 writer:

  • Strict standards compliance.

  • Generate good looking, readable, and accessible documents.

  • Clear distinction of content and layout:

    • Clean HTML output without "hard-coded" visual markup,

    • extended configurability by CSS style sheets.

  • Graceful Degradation

  • Best viewed with any (CSS2-conforming) HTML browser. [1]

  • Support scientific documents (numbering tables and figures, formal tables, ...). Cf. [markschenk].

1.3 Audience

This writer is for you, if you

  • care much about standards compliance,

  • care less about the rendering in non-compliant browsers,

  • want extended CSS configurability.

2 Usage

The rst2xhtml.py front end reads standalone reStructuredText source files and produces clean XHTML 1.1 output. A CSS 2 stylesheet is required for proper rendering; a complete sample stylesheet is provided.

Reader:

Standalone

Parser:

reStructuredText

Writer:

xhtml (xhtml11)

The front end can be called from the command line (when it is installed in the BINARY PATH):

rst2xhtml.py [options] [<source> [<destination>]]

The full usage text can be obtained with the --help option.

The front end rst2xhtml.py is also an example of programmatic use.

3 Implementation

The writer module subclasses the html_plain.Writer and html_plain.HTMLTranslator classes and adds compatibility to the strict requirements of XHTML 1.1:

The math-output configuration setting defaults to "MathML".

The xhtml11.css style sheet extends the standard layout for CSS2-conforming HTML browsers.

3.1 Changes to the html4css1 writer

3.1.1 Document type

  • The output conforms to the XHTML version 1.1 DTD:

    <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.1//EN"
 '"http://www.w3.org/TR/xhtml11/DTD/xhtml11.dtd">

    or XHTML + MathML:

    <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.1 plus MathML 2.0//EN"
"http://www.w3.org/Math/DTD/mathml2/xhtml-math11-f.dtd">

3.1.2 Docinfo and field lists based on definition lists (instead of tables)

  • Reduced loading time for documents with long field lists.

  • Enables CSS styling for:

    • label width (obsoleting the --field-name-limit option)

    • handling of long labels: truncate, wrap, ...

    • label separator (default: ':')

    • compact vs. open list

3.1.3 Class arguments for docinfo items

Items in the docinfo list are passed class arguments specifying their type to enable customizing the docinfo layout.

The default style sheet contains example definitions: author and date are typeset centered and without label, if they occur as first docinfo fields.

3.1.4 Footnotes and citations

  • Typeset as CSS-styled definition lists.

  • Collect adjacent footnotes/citations in one list.

3.1.5 Counter for enumerated lists

A CSS counter for enumerated lists replaces the deprecated "start" attribute.

  • Enables CSS styling for:

    • label style (including nested numbers),

    • label separator.

The complicated part was to find out a correct CSS rule-set to replicate the standard behaviour with hanging indent (list-style: "outside"). There is a W3C example to number nested list items, however, the result is similar to 'list-style: inside': subsequent lines start below the label instead of a hanging indent.

Most Internet resources come to the conclusion that "there’s no straightforward replacement mechanism" [tekkie], "the solution is buried so deep in CSS2 that there's no point in trying to do it in CSS for the foreseeable future" [webjunction], or "the main point to note is that there is no direct mapping from the previous behaviour to CSS" [codelair]. Taming Lists did give valuable advise but no working complete solution.

The common advise is "Use 'HTML 4.01 Transitional' and keep the START attribute". [highdots], especially, since "There are arguments over whether the start attribute is presentational or not, and indeed HTML5 has declared that it is no longer deprecated in the current working drafts (at the time of writing)" [dev.opera].

However, a reasonable replacement of 'outside'-styled ordered lists with CSS is possible:

  • The ordered list defines/resets the counter, the automatic numbering is suppressed:

    ol {
  counter-reset: item;
  list-style-type: none  ! important;
}

  • The label is defined as "before" pseudo element. The content consists of the counter and a separator (by default a trailing dot):

    ol > li:before {
  counter-increment: item;
  content: counter(item) ".";
}

  • The label is right aligned in a box. Both the label and the list content (which Docutils puts in a paragraph node) must be displayed as "inline-block" so that they line up:

    ol > li:before {
  display: inline-block;
  vertical-align: top;
  width: 2em;
  padding-right: 0.5em;
  text-align: right;
}

ol > li > p { display: inline-block; }

    However, subsequent paragraphs are to be set as nested block elements:

    ol > li > p + p {
  display: block;
  margin-top: 0em;
}

  • The hanging indent is realized via a negative "textindent" which must be reset for the list content to prevent over-striking:

    ol > li { text-indent: -2.5em; }
ol > li > p { text-indent: 0em; }

The resulting list can be customized to a large extend

  • Different label types and separators are possible, e.g.:

    ol.lowerroman  > li:before {
  content: "(" counter(item, lower-roman) ")";
}

  • nested counters (1, 1.1, 1.1.1, etc):

    ol.nested > li:before {
  content: counters(item, ".") ". ";
}

  • chapter/section prefix, continued lists, ...

3.1.6 Inline literal role with pre-wrap

In contrast to the html4css1 writer, runs of whitespace are not replaced by &nbsp; entities (cf. bug #1938891).

Whitespace-handling and wrapping are configured with the CSS property white-space: pre-wrap:

Whitespace is preserved by the browser. Text will wrap when necessary, and on line breaks

However, most browsers wrap on non-word chars, too, if set to wrap at white-space. Text like "--an-option" or the regular expression [+]?(\d+(\.\d*)?|\.\d+) may be broken at the wrong place! The setting white-space: pre; prevents this, but also prevents wrapping at white space, contrary to the specification

In order to allow line-wrap at whitespace only, words-with-non-word-chars are wrapped in <span>s with class "pre".

  • simpler HTML code

  • White-space handling in inline literals configurable with the CSS stylesheet. Possible values: normal, nowrap, pre, pre-wrap, pre-line.

3.1.7 Table styling with CSS

  • No hard-coded border setting.

  • Pre-defined table styles selected by class arguments "borderless" and "booktabs" matching the interpretation in the latex2e writer.

3.1.8 SimpleListChecker also checks field-lists and docinfo

Unified test if a list is compactable:

  • cleaner code

  • also works for nesting field-list in enumeration/bullet-list and vice versa

  • also test docinfo, as a field may contain more than one paragraph

3.1.9 Docutils-generated section numbers in a <span>

Instead of hard-coded formatting with trailing &nbsp;, section numbers in section headings and the toc are placed in spans with class='sectnum' and separated from the heading by a CSS rule.

3.1.10 Omit redundant class arguments

Do not mark the first child with 'class="first"' and the last child with 'class="last"' in definitions, table cells, field bodies, option descriptions, and list items. Use the :first-child and :last-child selectors instad.

3.1.11 Language attribute name changed to 'xml:lang'

The name of the language attribute changed from 'lang' in XHTML 1.0 to 'xml:lang' in XHTML 1.1. Documents using 'lang' do not validate.

3.1.12 Do not omit <p> tags

The HTML4CSS1 writer does this to "produce visually compact lists (less vertical whitespace)". This writer relies on CSS rules for"visual compactness".

3.2 TODO

  • The first list in the test 2.3. Enumerated Lists should be compact.

  • Hanging indent for numbered section headings and ToC entries.

  • search stylesheets along standard path if enclosed in <> (like the RST syntax for include files).

  • Validate output with "critical" cases not covered by the functional test (e.g. headings with level > 6).

  • Move widely supported constructs to the html4css1 writer.

  • Number sections with CSS if sectnum_xform is False.

  • Footnotes and Citations (for footnotes see http://www.archiva.net/footnote/index.htm and http://www.xmlplease.com/footnotes

4 References

../html4trans

is a similar sandbox project, a HTML writer producing XHTML that contains enough formatting information to be viewed without a cascading style sheet by a lightweight html browser (e.g. Dillo or the console browser elinks).

XHTML 1.1

XHTML™ 1.1 - Module-based XHTML - Second Edition, W3C Recommendation, 23 November 2010. http://www.w3.org/TR/xhtml11/