Changelog

All notable changes to phil/markup will be documented in this file.

Updates should follow the Keep a CHANGELOG principles.

0.1.0-alpha - 2026-11-11

Added

• MarkupFormat (markdown, html, bbcode, mediawiki, dokuwiki, creole, text, ujo), nameOf() the name of a case or of a format an add-on brings, and Level (basic, forum, all).
• MarkedUpText: a source, the name of its format and its purified HTML, a Doctrine embeddable; doctrine/dbal and doctrine/orm optional, in suggest, only an application that stores it needing them.
• Exception\MarkupException; Exception\InvalidSource, thrown for a UJO source that is not valid; Exception\UnknownFormat, for a format no reader knows and a name an add-on cannot take.
• MarkupCompiler and HtmlPurifier, the two ports, a format given as a MarkupFormat or by its name; LevelAwareCompiler, a compiler told the level its HTML will be purified at; Markup, which compiles — at the level, for a compiler that knows levels — then purifies.
• Lexer\Cursor, the analyser every reader shares: a cursor that lexes forward without regular expressions — comesNext(), consume(), consumeWhile(), consumeUntil(), consumeThrough() —, and moves back when a rule does not match; Lexer\Slugger, the unique anchors of a document's headings, in constant time per heading.
• Reader\MarkupReader and Reader\Reading: a source read into a checked UJO document of phil/ujo, with its metadata and its losses, Reader\Loss and Reader\LossReason — the element not kept, its path and why —; the readers given no normalizer share one, built once, that accepts raw blocks.
• Contract\ExtractedElementStore, the port where a reader stores, as files, the elements it extracts: an inline SVG drawing.
• Reader\Markdown\MarkdownReader, reading in two passes, blocks then inline content by one rule per token: CommonMark — headings, paragraphs, quotes, lists, code, rules, links and images, references, emphasis, escapes, entities, HTML —, the text of an image its alternative text and its title, written after its address or in its reference, its caption; the GitHub extensions — tables, strikethrough, autolinks, task lists —; footnotes; attributes {.class #id} and {class,#id,property:value} as tunes or a container; definition lists; heading anchors; [TOC]; ==highlight==; ruby, {漢字|かんじ}; YAML front matter; ::: blocks; raw HTML kept for all or escaped below; pages and columns, {{ }} blocks and the spans alone on their line mapped to the UJO types of the same role — boxed notes, read-aloud text, stat blocks, tables of contents, quotes, credits, class tables, spell lists, covers, page numbers, running footers, watermarks, banners, decorations. Any text is a document, read in linear time: emphasis is paired on a linked list of nodes and a paragraph of many lines is built in place. A Reader\Markdown\Contract\MarkdownDialect — BaseDialect to extend — adds its inline rules (Rule\InlineRule, Rule\DelimiterRule('^') for exponents) and block rules (Contract\BlockRule) after the built-in ones, turns off a built-in syntax (MarkdownSyntax), and reads its own metadata, [label]: value lines, definition marks and paragraph marks; what it keeps during one reading goes into Context::$extra.
• Reader\BBCode\BBCodeReader, on the same analyser, in linear time however the tags are nested, left open or closed without being opened: the tags of forums — bold, italic, underline, strikethrough, exponents, highlight, inline code, links, emails, images with their alternative text and their layout, quotes with their author, code with its language, lists ordered, counted or checked, tables with colspan, rowspan and align, headings, colors, alignment, spoilers, [ruby=reading] — and extensions in their style for every feature of the Markdown reader — [footnote], [dl], [toc], attributes id, class, align, width, height, wrap, [note], [descriptive], [warning], [alert], [wide], [block], [columns] and [column], [pagebreak], [columnbreak], [statblock], [credit], [classtable], [spacer], [pagenumber], [footer], [watermark], [banner], [decoration], [nbsp], [icon], [hspace], [var], [metadata], [style] —; raw HTML in [html] kept for all or escaped below.
• Reader\MediaWiki\MediaWikiReader, Reader\DokuWiki\DokuWikiReader and Reader\Creole\CreoleReader, built on one reader the three wikis share, Reader\Wiki, in linear time however marks, links, templates and containers are left open, crossed or nested: the native syntax of each wiki — headings, emphasis, links to pages and to other sites, images and files placed and sized, with their caption — and MediaWiki's alt= apart from it, Creole's text as the alternative text —, lists, task lists, definitions, indents, tables — MediaWiki's caption, spans and alignment, DokuWiki's || and ::: spans and alignment by spaces —, preformatted text, text kept as written, line breaks, rules; MediaWiki's entities, magic words, comments and templates, DokuWiki's footnotes ((…)), media placed by their spaces and macros ~~…~~, Creole 1.0 and its additions — and extension tags they share, <note> in MediaWiki and DokuWiki and <<note>> in Creole, for every feature of the Markdown reader: footnotes, <ref name> referenced again, spoilers, colors, icons, spaces, variables, notes, read-aloud text, warnings, alerts and DokuWiki's note plugin, quotes with their author, toggles, blocks across the columns, classes of the look, columns side by side or flowing, stat blocks, credits, class tables, alignment, running footers, watermarks, banners, tables of contents, page and column breaks, page styles, spacers, page numbers, decorations, code with its language, metadata and style, attributes id, class, align, width, height, wrap; raw HTML in <html> kept for all or escaped below. Reader\Wiki\WikiLinks says where the pages and the media a wiki text names are found.
• Reader\Html\HtmlReader, HTML read into UJO through the HTML5 parser of PHP: paragraphs, headings, lists and checklists, quotes, figures, images — their alt the alternative text, their <figcaption> or else their title the caption —, code, rules, tables, definition lists, details, alignment and the inline elements UJO keeps; the HTML phil/ujo renders, its ujo-* classes and footnotes read back into their blocks; the styles a paste from a word processor writes read as bold, italic, underline and strikethrough; frames of known services as embed blocks, videos and sounds with their tracks, formulas in MathML, download links alone in their paragraph as attached files, abbreviations, languages as tunes and <span lang>, ruby, tables with their caption, spans, header cells and alignments, a cell's lists as lines with their marks, data-* attributes and comments as the data and note tunes, links of the schemes phil/ujo allows and anchors the page names, inline SVG drawings purified by phil/svg-sanitizer and stored by an ExtractedElementStore; scripts, styles and forms left out; what is not kept listed in the reading's losses; the title, the language and the named <meta> of a page as its metadata; ext-dom and ext-libxml required.
• Reader\UjoReader: UJO JSON read strictly by phil/ujo.
• Reader\Text\TextReader: plain text read into UJO, a paragraph for each run of lines, line breaks kept, every character escaped and the addresses made links.
• Adapter\DefaultCompiler: Markdown, BBCode, MediaWiki, DokuWiki, Creole and UJO read into UJO and rendered by phil/ujo; HTML as written; plain text by Adapter\PlainText, addresses made links by Adapter\Autolink; reader() gives the reader of every format into UJO, HTML's and plain text's included; the formats of add-ons, each a Reader\Contract\ReaderProvider given by name, read and rendered the same way; formats() the name of each.
• Adapter\SanitizerPurifier: a policy of symfony/html-sanitizer per level, taking the hosts images may come from, since an image of another site, which every reader's browser would call, is tracking at the forum and all levels; every class dropped but those of code, kept by Adapter\CodeClasses: language-… on code and hl-… on span; links of the schemes of phil/ujo, and MathML Core at all.
• The documentation in doc/, indexed by the README, with the correspondence of the features of every format.
• CONTRIBUTING.md, SECURITY.md and CODE_OF_CONDUCT.md.