The wikis

Reader\MediaWiki\MediaWikiReader, Reader\DokuWiki\DokuWikiReader and Reader\Creole\CreoleReader read the syntax of three wikis into a UJO document, which phil/ujo renders as HTML. Each reads the native syntax of its wiki — headings, emphasis, links to pages and to other sites, lists, tables, preformatted text, text kept as written, images, rules — and the extensions every wiki of PHIL shares, written as the wiki writes its own: tags in MediaWiki and DokuWiki, <note>, placeholders in Creole, <<note>>; templates and magic words in MediaWiki, {{TOC}}, __TOC__; macros in DokuWiki, ~~TOC~~. Together they give each wiki every feature of the Markdown reader. Formats puts each syntax beside those of the other formats.

$reading = new MediaWikiReader()->read("== Rules ==\nRoll '''two''' dice.<ref>Six-sided.</ref>");

$reading->document->blocks[0];   // a `header` block, level 2, its anchor `rules`
$reading->document->blocks[2];   // the `footnote` block, `Six-sided.`

new MediaWikiReader(true) keeps raw HTML; DefaultCompiler asks for it at the level all only. A second argument replaces the phil/ujo normalizer the document is checked by; a third, a Reader\Wiki\WikiLinks, says where pages and files are found.

Pages and files

A link to a page, [[Main Page#History]], goes to the path of the page, its spaces underscores, its other unsafe characters encoded, and its section an anchor made as the readers make the anchors of headings: /Main_Page#history. A file, [[File:map.png]] or {{map.png}}, goes to the path of the media. A web address is kept as written.

new MediaWikiReader(false, null, new WikiLinks(pages: '/wiki/', media: 'https://cdn.example.com/'));

Both paths start with / by default, so that a name holding : — a namespace, Help:Contents — is never read as a scheme. DokuWiki writes its page names in lower case, as DokuWiki does.

The extension tags

A tag is written <name>, <name flag key="value"> or <name/> in MediaWiki and DokuWiki, <<name …>> in Creole, its name in any case, and closed by </name> or <</name>>. A value holding spaces is quoted.

Inline

A tag not known, or left open, is text. A tag holding a tag of its own name closes at the first closing tag.

Blocks

A container opens with its tag at the start of a line, and holds the blocks up to its closing tag — on the same line or a line after it — or up to the closing tag of a container around it, or to the end. <note>a</note> is a note holding a paragraph.

Every container takes the attributes id — the anchor tune of the block it makes —, class — a container around it —, align, width, height and wrap="left" or wrap="right". Containers nest as deep as UJO allows: the tag of a container deeper is left out, its blocks read as those of the container around it. A closing tag that closes nothing open is left out.

MediaWiki

A template is named in any case, its spaces and underscores left out: {{Page break}} is {{pagebreak}}. A template not known is text.

DokuWiki

Creole

Text that is not well formed

Any text is a document, and the readers never throw. A mark of emphasis closes the element of its kind still open, and those opened inside it and left open are text, so that the HTML is always well nested; a mark never closed is text. A link, an image or a template is read within 4 KB and on one line; one not closed there is text. A container never closed holds the rest of the text.

Linear time

Each line is read once, the tags of containers set on lines of their own by one pass first. A delimiter missing from a position is missing from every position after it, which the inline reader remembers; content is read eight levels deep at most, deeper as text; a link, an image or a template is searched within 4 KB. The open containers are counted by name, and a row keeps sixty-four cells, the text of the others joined to the last. Sources of 100 KB written to be slow are each read in a fraction of a second.