www-lib-markdown/php/markdown.php

<?php
declare(strict_types=1);

/**
 * Static utilities.
 */
class MDUtils {
	// Modified from https://urlregex.com/ to remove capture groups. Matches fully qualified URLs only.
	public static $baseURLRegex = '(?:(?:(?:[a-z]{3,9}:(?:\\/\\/)?)(?:[\\-;:&=\\+\\$,\\w]+@)?[a-z0-9\\.\\-]+|(?:www\\.|[\\-;:&=\\+\\$,\\w]+@)[a-z0-9\\.\\-]+)(?:(?:\\/[\\+~%\\/\\.\\w\\-_]*)?\\??(?:[\\-\\+=&;%@\\.\\w_]*)#?(?:[\\.\\!\\/\\\\\\w]*))?)';
	// Modified from https://emailregex.com/ to remove capture groups.
	public static $baseEmailRegex = '(?:(?:[^<>()\\[\\]\\\\.,;:\\s@"]+(?:\\.[^<>()\\[\\]\\\\.,;:\\s@"]+)*)|(?:".+"))@(?:(?:\\[[0-9]{1,3}\\.[0-9]{1,3}\\.[0-9]{1,3}\\.[0-9]{1,3}])|(?:(?:[a-z\\-0-9]+\\.)+[a-z]{2,}))';

	/**
	 * Encodes characters as HTML numeric entities to make it marginally more
	 * difficult for web scrapers to grab sensitive info. If `text` starts with
	 * `mailto:` only the email address following it will be obfuscated.
	 */
	public static function escapeObfuscated(string $text): string {
		if (str_starts_with($text, 'mailto:')) {
			return 'mailto:' . self::escapeObfuscated(mb_substr($text, 7));
		}
		$html = '';
		$l = mb_strlen($text);
		for ($p = 0; $p < $l; $p++) {
			$cp = mb_ord(mb_substr($text, $p, 1));
			$html .= "&#{{$cp}}";
		}
		return $html;
	}

	/**
	 * Removes illegal characters from an HTML attribute name.
	 */
	public static function scrubAttributeName(string $name): string {
		return mb_ereg_replace('[\\t\\n\\f \\/>"\'=]+', '', $name);
	}

	/**
	 * Strips one or more leading indents from a line or lines of markdown. An
	 * indent is defined as 4 spaces or one tab. Incomplete indents (i.e. 1-3
	 * spaces) are treated like one indent level.
	 *
	 * @param string|string[] $line
	 * @param int $levels
	 * @return string|string[]
	 */
	public static function stripIndent(string|array &$line, int $levels=1): string|array {
		$regex = "^(?: {1,4}|\\t){{$levels}}";
		return is_array($line) ? array_map(fn(string $l): string => mb_ereg_replace($regex, '', $l), $line) : mb_ereg_replace($regex, '', $line);
	}

	/**
	 * Counts the number of indent levels in a line of text. Partial indents
	 * (1 to 3 spaces) are counted as one indent level unless `fullIndentsOnly`
	 * is `true`.
	 */
	public static function countIndents(string &$line, bool $fullIndentsOnly=false): int {
		// normalize indents to tabs
		$t = mb_ereg_replace($fullIndentsOnly ? "(?: {4}|\\t)" : "(?: {1,4}|\\t)", "\t", $line);
		// remove content after indent
		$t = mb_ereg_replace("^(\\t*)(.*?)$", "\\1", $t);
		// count tabs
		return mb_strlen($t);
	}

	/**
	 * Returns a copy of an array without any whitespace-only lines at the end.
	 *
	 * @param string[] $lines
	 * @return string[]
	 */
	public static function withoutTrailingBlankLines(array $lines): array {
		$stripped = $lines;
		while (sizeof($stripped) > 0 && mb_strlen(trim($stripped[sizeof($stripped) - 1])) == 0) {
			array_pop($stripped);
		}
		return $stripped;
	}

	/**
	 * Tests if an array of lines contains at least one blank. A blank line
	 * can contain whitespace.
	 *
	 * @param string[] $lines
	 */
	public static function containsBlankLine(array $lines): bool {
		foreach ($lines as $line) {
			if (mb_strlen(trim($line)) == 0) return true;
		}
		return false;
	}

	public static function equalAssocArrays(array &$a, array &$b) {
		return empty(array_diff_assoc($a, $b));
	}
}

/**
 * Token type enum for `MDToken`.
 */
enum MDTokenType {
	case Text;
	/**
	 * Only used for the leading and trailing whitespace around a run of text,
	 * not every single whitespace character.
	 */
	case Whitespace;

	case Underscore;
	case Asterisk;
	case Slash;
	case Tilde;
	case Bang;
	case Backtick;
	case Equal;
	case Caret;

	case Label; // content=label
	case URL; // content=URL, extra=title
	case Email; // content=email address, extra=title
	case SimpleLink; // content=URL
	case SimpleEmail; // content=email address
	case Footnote; // content=symbol
	case Modifier; // modifier=MDTagModifier

	case HTMLTag; // tag=MDHTMLTag

	/** Wildcard for `MDToken::findFirstTokens` */
	case META_AnyNonWhitespace;
	/** Wildcard for `MDToken::findFirstTokens` */
	case META_OptionalWhitespace;
}

/**
 * Search results from `MDToken.findFirstTokens`.
 */
class MDTokenMatch {
	/** @var MDToken{} */
	public array $tokens;
	public int $index;

	/**
	 * @param MDToken[] $tokens
	 * @param int $index
	 */
	public function __construct(array $tokens, int $index) {
		$this->tokens = $tokens;
		$this->index = $index;
	}
}

/**
 * Search results from `MDToken.findPairedTokens`.
 */
class MDPairedTokenMatch {
	/** @var MDToken[] */
	public array $startTokens;
	/** @var MDToken[] */
	public array $contentTokens;
	/** @var MDToken[] */
	public array $endTokens;
	public int $startIndex;
	public int $contentIndex;
	public int $endIndex;
	public int $totalLength;

	public function __construct(array $startTokens, array $contentTokens,
			array $endTokens, int $startIndex, int $contentIndex, int $endIndex,
			int $totalLength) {
		$this->startTokens = $startTokens;
		$this->contentTokens = $contentTokens;
		$this->endTokens = $endTokens;
		$this->startIndex = $startIndex;
		$this->contentIndex = $contentIndex;
		$this->endIndex = $endIndex;
		$this->totalLength = $totalLength;
	}
}

/**
 * One lexical unit in inline markdown syntax parsing.
 */
class MDToken {
	/**
	 * The original verbatim token string. Required as a plaintext fallback if
	 * the token remains unresolved.
	 */
	public string $original;
	public MDTokenType $type;
	public ?string $content = null;
	public ?string $extra = null;
	public ?MDHTMLTag $tag = null;
	public ?MDTagModifier $modifier = null;

	/**
	 * Creates a token.
	 *
	 * @param string $original  verbatim token string
	 * @param MDTokenType $type  token type
	 * @param string|MDTagModifier|MDHTMLTag|null $content  primary content of
	 *   the token
	 * @param string|null $extra  additional content
	 */
	public function __construct(string $original, MDTokenType $type,
			string|MDTagModifier|MDHTMLTag|null $content=null,
			?string $extra=null) {
		$this->original = $original;
		$this->type = $type;
		if ($content instanceof MDTagModifier) {
			$this->modifier = $content;
		} elseif ($content instanceof MDHTMLTag) {
			$this->tag = $content;
		} else {
			$this->content = $content;
		}
		$this->extra = $extra;
	}

	public function __toString(): string {
		$classname = get_class($this);
		return "({$classname} type={$this->type} content={$this->content})";
	}

	/**
	 * Attempts to parse a label token from the beginning of `line`. A label is
	 * of the form `[content]`. If found, returns an array:
	 * - `0`: the entire label including brackets
	 * - `1`: the content of the label
	 *
	 * @param string $line
	 * @return ?string[] match groups or null if not found
	 */
	public static function tokenizeLabel(string $line): ?array {
		if (!str_starts_with($line, '[')) return null;
		$parenCount = 0;
		$bracketCount = 0;
		$l = mb_strlen($line);
		for ($p = 1; $p < $l; $p++) {
			$ch = mb_substr($line, $p, 1);
			if ($ch == '\\') {
				$p++;
			} elseif ($ch == '(') {
				$parenCount++;
			} elseif ($ch == ')') {
				$parenCount--;
				if ($parenCount < 0) return null;
			} elseif ($ch == '[') {
				$bracketCount++;
			} elseif ($ch == ']') {
				if ($bracketCount > 0) {
					$bracketCount--;
				} else {
					return [ mb_substr($line, 0, $p + 1), mb_substr($line, 1, $p - 1) ];
				}
			}
		}
		return null;
	}

	private static $urlWithTitleRegex = '^\\((\\S+?)\\s+"(.*?)"\\)';  // 1=URL, 2=title
	private static $urlRegex = '^\\((\\S+?)\\)';  // 1=URL

	/**
	 * Attempts to parse a URL token from the beginning of `line`. A URL token
	 * is of the form `(url)` or `(url "title")`. If found, returns an array:
	 * - `0`: the entire URL token including parentheses
	 * - `1`: the URL
	 * - `2`: the optional title, or `null`
	 *
	 * @param string $line
	 * @return ?array token tuple
	 */
	public static function tokenizeURL(string $line): ?array {
		$groups = [];
		if (mb_eregi(self::$urlWithTitleRegex, $line, $groups)) {
			if (self::tokenizeEmail($line)) return null; // make sure it's not better described as an email address
			return $groups;
		}
		if (mb_eregi(self::$urlRegex, $line, $groups)) {
			if (self::tokenizeEmail($line)) return null;
			return [ $groups[0], $groups[1], null ];
		}
		return null;
	}

	/**
	 * Attempts to parse an email address from the beginning of `line`. An
	 * email address is of the form `(user@example.com)` or
	 * `(user@example.com "link title")`. If found, returns an array:
	 * - `0`: the entire token including parentheses
	 * - `1`: the email address
	 * - `2`: the optional link title, or `null`
	 *
	 * @param string $line
	 * @return ?string[] token tuple
	 */
	public static function tokenizeEmail(string $line): ?array {
		$groups;
		if (mb_eregi("^\\(\\s*(" . MDUtils::$baseEmailRegex . ")\\s+\"(.*?)\"\\s*\\)",
				$line, $groups)) {
			return $groups;
		}
		if (mb_eregi("^\\(\\s*(" . MDUtils::$baseEmailRegex . ")\\s*\\)", $line, $groups)) {
			return [ $groups[0], $groups[1], null ];
		}
		return null;
	}

	/**
	 * Searches an array of `MDToken` for the given pattern of `MDTokenType`s.
	 * If found, returns a `MDTokenMatch`, otherwise `null`.
	 *
	 * Special token types `META_AnyNonWhitespace` and `META_OptionalWhitespace`
	 * are special supported token types. Note that `META_OptionalWhitespace`
	 * may give a result with a variable number of tokens.
	 *
	 * @param (MDToken|MDNode)[] $tokensToSearch - mixed array of `MDToken` and
	 *   `MDNode` elements
	 * @param MDTokenType[] $pattern - contiguous run of token types to find
	 * @param int $startIndex - token index to begin searching (defaults to 0)
	 * @return ?MDTokenMatch match object, or `null` if not found
	 */
	public static function findFirstTokens(array $tokensToSearch, array $pattern, int $startIndex=0): ?MDTokenMatch {
		$matched = [];
		for ($t = $startIndex; $t < sizeof($tokensToSearch); $t++) {
			$matchedAll = true;
			$matched = [];
			$patternOffset = 0;
			for ($p = 0; $p < mb_strlen($pattern); $p++) {
				$t0 = $t + $p + $patternOffset;
				if ($t0 >= sizeof($tokensToSearch)) return null;
				$token = $tokensToSearch[$t0];
				$elem = $pattern[$p];
				if ($elem == MDTokenType::META_OptionalWhitespace) {
					if ($token instanceof MDToken && $token->type == MDTokenType::Whitespace) {
						array_push($matched, $token);
					} else {
						$patternOffset--;
					}
				} elseif ($elem == MDTokenType::META_AnyNonWhitespace) {
					if ($token instanceof MDToken && $token->type == MDTokenType::Whitespace) {
						$matchedAll = false;
						break;
					}
					array_push($matched, $token);
				} else {
					if (!($token instanceof MDToken) || $token->type != $elem) {
						$matchedAll = false;
						break;
					}
					array_push($matched, $token);
				}
			}
			if ($matchedAll) {
				return new MDTokenMatch($matched, $t);
			}
		}
		return null;
	}

	/**
	 * Searches an array of MDToken for a given starting pattern and ending
	 * pattern and returns match info about both and the tokens in between.
	 *
	 * If `contentValidator` is specified, it will be called with the content
	 * tokens of a potential match. If the validator returns `true`, the result
	 * will be accepted and returned by this method. If the validator returns
	 * `false`, this method will keep looking for another matching pair. If no
	 * validator is given the first match will be returned regardless of content.
	 *
	 * If a match is found, a `MDPairedTokenMatch` is returned with details
	 * of the opening tokens, closing tokens, and content tokens between. Otherwise
	 * `null` is returned.
	 *
	 * @param MDToken[] $tokensToSearch - array of `MDToken` to search in
	 * @param MDTokenType[] $startPattern - array of `MDTokenType` to find first
	 * @param MDTokenType[] $endPattern - array of `MDTokenType` to find positioned after `startPattern`
	 * @param ?callable $contentValidator - optional validator function. If provided, will be passed an array of inner `MDToken`, and the function can return `true` to accept the contents or `false` to keep searching
	 * @param number $startIndex - token index where searching should begin
	 * @return ?MDPairedTokenMatch match, or `null`
	 */
	public static function findPairedTokens(array $tokensToSearch,
			array $startPattern, array $endPattern, ?callable $contentValidator=null,
			int $startIndex=0): ?MDPairedTokenMatch {
		for ($s = $startIndex; $s < sizeof($tokensToSearch); $s++) {
			$startMatch = findFirstTokens($tokensToSearch, $startPattern, $s);
			if ($startMatch === null) return null;
			$endStart = $startMatch->index + sizeof($startMatch->tokens);
			while ($endStart < sizeof($tokensToSearch)) {
				$endMatch = findFirstTokens($tokensToSearch, $endPattern, $endStart);
				if ($endMatch === null) break;
				$contentStart = $startMatch->index + sizeof($startMatch->tokens);
				$contentLength = $endMatch->index - $contentStart;
				$contents = array_slice($tokensToSearch, $contentStart, $contentLength);
				if (sizeof($contents) > 0 && ($contentValidator === null || $contentValidator($contents))) {
					return new MDPairedTokenMatch($startMatch->tokens,
						$contents,
						$endMatch->tokens,
						$startMatch->index,
						$startMatch->index + sizeof($startMatch->tokens),
						$endMatch->index,
						$endMatch->index + sizeof($endMatch->tokens) - $startMatch->index);
				} else {
					// Contents rejected. Try next end match.
					$endStart = $endMatch->index + 1;
				}
			}
			// No end matches. Increment start match.
			$s = $startMatch->index;
		}
		return null;
	}

	public function equals($other) {
		if (!($other instanceof MDToken)) return false;
		if ($other->original !== $this->original) return false;
		if ($other->type != $this->type) return false;
		if ($other->content !== $this->content) return false;
		if ($other->extra !== $this->extra) return false;
		if ($other->tag !== $this->tag) return false;
		if ($other->modifier != $this->modifier) return false;
		return true;
	}
}

/**
 * Parsing and rendering state. Passed around throughout the parsing process.
 *
 * States are hierarchical. A sub-state can be created by calling `.copy()` with
 * a new array of lines. The sub-state points back to its parent state. This
 * is done to parse inner content of a syntax as its own standalone document.
 *
 * If a custom `MDReader` implementation wants to store data in this object,
 * always do so on `state.root` to ensure it's stored on the original state,
 * not a child state. Otherwise data may be lost when the sub-state is discarded.
 */
class MDState {
	/**
	 * Ascends the parent chain to the root `MDState` instance. This should be
	 * used when referencing most stored fields except `lines` and `p`.
	 */
	public function root(): MDState {
		return $this->parent ? $this->parent->root() : $this;
	}

	/**
	 * Lines of the markdown document. The current line index is pointed to by `p`.
	 *
	 * @var string[]
	 */
	public array $lines;

	/**
	 * The current line in `lines`.
	 */
	public function currentLine(): ?string {
		return ($this->p < sizeof($this->lines)) ? $this->lines[$this->p] : null;
	}

	/**
	 * Current line pointer into array `lines`.
	 */
	public int $p = 0;

	/**
	 * General storage for anything readers need to track during the parsing
	 * process.
	 */
	public array $userInfo = [];

	private ?MDState $parent = null;

	/**
	 * Array of `MDReader`s sorted by block reading priority.
	 * @var MDReader[]
	 */
	public array $readersByBlockPriority = [];

	/**
	 * Array of `MDReader`s sorted by tokenization priority.
	 * @var MDReader[]
	 */
	public array $readersByTokenPriority = [];

	/**
	 * Array of tuples of `pass:number` and `MDReader` sorted by substitution
	 * priority.
	 * @var array[]
	 */
	public array $readersBySubstitutePriority = [];

	/**
	 * Prefix to include in any generated `id` attributes on HTML elements.
	 * Useful for keeping elements unique in multiple parsed documents in the
	 * same HTML page.
	 */
	public string $elementIdPrefix = '';

	/**
	 * Filter for removing unapproved HTML tags, attributes, and values.
	 */
	public MDHTMLFilter $tagFilter;

	/**
	 * @param string[] $lines - lines of markdown text
	 */
	public function __construct(array $lines) {
		$this->lines = $lines;
	}

	/**
	 * Creates a copy of this state with new lines. Useful for parsing nested
	 * content.
	 *
	 * @param string[] $lines
	 * @return MDState copied sub-state
	 */
	public function copy(array $lines) {
		$cp = new MDState($lines);
		$cp->parent = $this;
		return $cp;
	}

	/**
	 * Tests if there are at least `minCount` lines available to read. If `p`
	 * is not provided it will be relative to `this.p`.
	 */
	public function hasLines(int $minCount, ?int $p=null): bool {
		$relativeTo = ($p === null) ? $this->p : $p;
		return $relativeTo + $minCount <= sizeof($this->lines);
	}

	/**
	 * Reads and returns an array of blocks from the current line pointer.
	 *
	 * @return MDBlockNode[]  parsed blocks
	 */
	public function readBlocks(): array {
		$blocks = [];
		while ($this->hasLines(1)) {
			$block = $this->readNextBlock();
			if ($block) {
				array_push($blocks, $block);
			} else {
				break;
			}
		}
		return $blocks;
	}

	/**
	 * Creates a simple `MDBlockNode` if no other registered blocks match.
	 */
	private function readFallbackBlock(): ?MDBlockNode {
		if ($this->p >= sizeof($this->lines)) return null;
		$lines = MDUtils::withoutTrailingBlankLines(array_slice($this->lines, $this->p));
		if (sizeof($lines) == 0) return null;
		$this->p = sizeof($this->lines);
		return new MDBlockNode($this->inlineMarkdownToNode(implode("\n", $lines)));
	}

	/**
	 * Attempts to read one block from the current line pointer. The pointer
	 * will be positioned just after the end of the block.
	 */
	private function readNextBlock(): ?MDBlockNode {
		while ($this->hasLines(1) && mb_strlen(trim($this->lines[$this->p])) == 0) {
			$this->p++;
		}
		if (!$this->hasLines(1)) return null;
		foreach ($this->root()->readersByBlockPriority as $reader) {
			$startP = $this->p;
			$block = $reader->readBlock($this);
			if ($block) {
				if ($this->p == $startP) {
					$readerClassName = get_class($reader);
					$blockClassName = get_class($block);
					throw new Error("{$readerClassName} returned an " .
						"{$blockClassName} without incrementing MDState.p. " .
						"This could lead to an infinite loop.");
				}
				return $block;
			}
		}
		$fallback = $this->readFallbackBlock();
		return $fallback;
	}

	/**
	 * @param string $line
	 * @return MDToken[]
	 */
	private function inlineMarkdownToTokens(string $line): array {
		if ($this->parent) return $this->parent->inlineMarkdownToTokens($line);

		$tokens = [];
		$text = '';
		$expectLiteral = false;

		/**
		 * Flushes accumulated content in `text` to `tokens`.
		 */
		$endText = function() use (&$tokens, &$text) {
			if (mb_strlen($text) == 0) return;
			$textGroups = [];
			if (mb_eregi('^(\s+)(.*?)$', $text, $textGroups)) {
				array_push($tokens, new MDToken($textGroups[1], MDTokenType::Whitespace, $textGroups[1]));
				$text = $textGroups[2];
			}
			if (mb_eregi('^(.*?)(\s+)$', $text, $textGroups)) {
				array_push($tokens, new MDToken($textGroups[1], MDTokenType::Text, $textGroups[1]));
				array_push($tokens, new MDToken($textGroups[2], MDTokenType::Whitespace, $textGroups[2]));
			} else {
				array_push($tokens, new MDToken($text, MDTokenType::Text, $text));
			}
			$text = '';
		};

		for ($p = 0; $p < mb_strlen($line); $p++) {
			$ch = mb_substr($line, $p, 1);
			$remainder = mb_substr($line, $p);
			if ($expectLiteral) {
				$text .= $ch;
				$expectLiteral = false;
				continue;
			}
			if ($ch == '\\') {
				$expectLiteral = true;
				continue;
			}
			$found = false;
			foreach ($this->root()->readersByTokenPriority as $reader) {
				$token = $reader->readToken($this, $remainder);
				if ($token === null) continue;
				$endText();
				array_push($tokens, $token);
				if ($token->original == null || mb_strlen($token->original) == 0) {
					$readerClassName = get_class($reader);
					throw new Error(`{$readerClassName} returned a token with an empty .original. This would cause an infinite loop.`);
				}
				$p += mb_strlen($token->original) - 1;
				$found = true;
				break;
			}
			if (!$found) {
				$text .= $ch;
			}
		}
		$endText();
		return $tokens;
	}

	/**
	 * Converts a line of markdown to an `MDInlineNode`.
	 *
	 * @param string|string[] $line
	 * @return MDInlineNode
	 */
	public function inlineMarkdownToNode(string|array $line): MDInlineNode {
		$nodes = $this->inlineMarkdownToNodes($line);
		return (sizeof($nodes) == 1) ? $nodes[0] : new MDInlineNode($nodes);
	}

	/**
	 * Converts a line of markdown to an array of `MDInlineNode`s.
	 *
	 * @param string|string[] $line
	 * @return MDInlineNode[]
	 */
	public function inlineMarkdownToNodes(string|array $line): array {
		$tokens = $this->inlineMarkdownToTokens(is_array($line) ? implode("\n", $line) : $line);
		return $this->tokensToNodes($tokens);
	}

	/**
	 * Converts a mixed array of `MDToken` and `MDInlineNode` elements into an array
	 * of only `MDInlineNode` via repeated `MDReader` substition.
	 *
	 * @param (MDToken|MDInlineNode)[] $tokens
	 * @return MDInlineNode[]
	 */
	public function tokensToNodes(array $tokens): array {
		$nodes = $tokens;

		// Perform repeated substitutions, converting sequences of tokens into
		// nodes, until no more substitutions can be made.
		$anyChanges = false;
		do {
			$anyChanges = false;
			foreach ($this->root()->readersBySubstitutePriority as $readerTuple) {
				/** @var int */
				$pass = $readerTuple[0];
				/** @var MDReader */
				$reader = $readerTuple[1];
				$changed = $reader->substituteTokens($this, $pass, $nodes);
				if (!$changed) continue;
				$anyChanges = true;
				break;
			}
		} while ($anyChanges);

		// Convert any remaining tokens to text nodes. Also apply any inline
		// CSS modifiers.
		$lastNode = null;
		$me = $this;
		$nodes = array_map(function($node) use ($lastNode, $me) {
			if ($node instanceof MDToken) {
				/** @var MDToken */
				$token = $node;
				if ($token->type == MDTokenType::Modifier && $lastNode) {
					$me->root()->tagFilter->scrubModifier($token->modifier);
					$token->modifier->applyTo($lastNode);
					$lastNode = null;
					return new MDTextNode('');
				}
				$lastNode = null;
				return new MDTextNode($token->original);
			} elseif ($node instanceof MDNode) {
				$lastNode = ($node instanceof MDTextNode) ? null : $node;
				return $node;
			} else {
				$nodeClassName = get_class($node);
				throw new Error("Unexpected node type {$nodeClassName}");
			}
		}, $nodes);

		return $nodes;
	}

	/**
	 * Mapping of reference symbols to URLs. Used by `MDReferencedLinkReader`
	 * and `MDReferencedImageReader`.
	 * @var array symbol -> URL
	 */
	private array $referenceToURL = [];

	/**
	 * Mapping of reference symbols to titles. Used by `MDReferencedLinkReader`
	 * and `MDReferencedImageReader`.
	 * @var array symbol -> title string
	 */
	private array $referenceToTitle = [];

	/**
	 * Defines a URL by reference symbol.
	 */
	public function defineURL(string $reference, string $url, ?string $title=null) {
		$this->root()->referenceToURL[mb_strtolower($reference)] = $url;
		if ($title !== null) $this->root()->referenceToTitle[mb_strtolower($reference)] = $title;
	}

	/**
	 * Returns the URL associated with a reference symbol.
	 */
	public function urlForReference(string $reference): ?string {
		return $this->root()->referenceToURL[mb_strtolower($reference)] ?? null;
	}

	/**
	 * Returns the link title associated with a reference symbol.
	 */
	public function urlTitleForReference(string $reference): ?string {
		return $this->root()->referenceToTitle[mb_strtolower($reference)] ?? null;
	}
}

/**
 * Defines a set of allowable HTML tags, attributes, and CSS.
 */
class MDHTMLFilter {
	/**
	 * Mapping of permitted lowercase tag names to objects containing allowable
	 * attributes for those tags. Does not need to include those attributes
	 * defined in `allowableGlobalAttributes`.
	 *
	 * Values are objects with allowable lowercase attribute names mapped to
	 * allowable value patterns. A `*` means any value is acceptable. Multiple
	 * allowable values can be joined together with `|`. These special symbols
	 * represent certain kinds of values and can be used in combination or in
	 * place of literal values.
	 *
	 * - `{classlist}`: A list of legal CSS classnames, separated by spaces
	 * - `{int}`: An integer
	 * - `{none}`: No value (an attribute with no `=` or value, like `checked`)
	 * - `{style}`: One or more CSS declarations, separated by semicolons (simple
	 *   `key: value;` syntax only)
	 * - `{url}`: A URL
	 * @type {object}
	 */
	public array $allowableTags = [
		'address' => [
			'cite' => '{url}',
		],
		'h1' => [],
		'h2' => [],
		'h3' => [],
		'h4' => [],
		'h5' => [],
		'h6' => [],
		'blockquote' => [],
		'dl' => [],
		'dt' => [],
		'dd' => [],
		'div' => [],
		'hr' => [],
		'ul' => [],
		'ol' => [
			'start' => '{int}',
			'type' => 'a|A|i|I|1',
		],
		'li' => [
			'value' => '{int}',
		],
		'p' => [],
		'pre' => [],
		'table' => [],
		'thead' => [],
		'tbody' => [],
		'tfoot' => [],
		'tr' => [],
		'td' => [],
		'th' => [],
		'a' => [
			'href' => '{url}',
			'target' => '*',
		],
		'abbr' => [],
		'b' => [],
		'br' => [],
		'cite' => [],
		'code' => [],
		'data' => [
			'value' => '*',
		],
		'dfn' => [],
		'em' => [],
		'i' => [],
		'kbd' => [],
		'mark' => [],
		'q' => [
			'cite' => '{url}',
		],
		's' => [],
		'samp' => [],
		'small' => [],
		'span' => [],
		'strong' => [],
		'sub' => [],
		'sup' => [],
		'time' => [
			'datetime' => '*',
		],
		'u' => [],
		'var' => [],
		'wbr' => [],
		'img' => [
			'alt' => '*',
			'href' => '{url}',
		],
		'figure' => [],
		'figcaption' => [],
		'del' => [],
		'ins' => [],
		'details' => [],
		'summary' => [],
	];

	/**
	 * Mapping of allowable lowercase global attributes to their permitted
	 * values. Uses same value pattern syntax as described in `allowableTags`.
	 * @type {object}
	 */
	public array $allowableGlobalAttributes = [
		'class' => '{classlist}',
		'data-*' => '*',
		'dir' => 'ltr|rtl|auto',
		'id' => '*',
		'lang' => '*',
		'style' => '{style}',
		'title' => '*',
		'translate' => 'yes|no|{none}',
	];

	/**
	 * Mapping of allowable CSS style names to their allowable value patterns.
	 * Multiple values can be delimited with `|` characters. Limited support
	 * so far.
	 *
	 * Recognized special values:
	 * - `{color}`: A hex or named color
	 *
	 * @type {object}
	 */
	public array $allowableStyleKeys = [
		'background-color' => '{color}',
		'color' => '{color}',
	];

	/**
	 * Scrubs all forbidden attributes from an HTML tag. Assumes the tag name
	 * itself has already been whitelisted.
	 *
	 * @param {MDHTMLTag} tag - HTML tag
	 */
	public function scrubTag(MDHTMLTag $tag) {
		foreach ($tag->attributes as $name => $value) {
			if (!$this->isValidAttributeName($tag->tagName, $name)) {
				unset($tag->attributes[$name]);
			}
			if (!$this->isValidAttributeValue($tag->tagName, $name, $value)) {
				unset($tag->attributes[$name]);
			}
		}
	}

	/**
	 * Scrubs all forbidden attributes from an HTML modifier.
	 *
	 * @param MDTagModifier $modifier
	 * @param ?string $tagName  HTML tag name, if known, otherwise only
	 *   global attributes will be permitted
	 */
	public function scrubModifier(MDHTMLModifier $modifier, ?string $tagName) {
		if (sizeof($modifier->cssClasses) > 0) {
			$classList = implode(' ', $modifier->cssClasses);
			if (!$this->isValidAttributeValue($tagName, 'class', $classList)) {
				$modifier->cssClasses = [];
			}
		}
		if ($modifier->cssId !== null) {
			if (!$this->isValidAttributeValue($tagName, 'id', $modifier->cssId)) {
				$modifier->cssId = null;
			}
		}
		if (!$this->isValidAttributeName($tagName, 'style')) {
			$modifier->cssStyles = [];
		} else {
			foreach ($modifier->cssStyles as $key => $val) {
				if (!$this->isValidStyleValue($key, $val)) {
					unset($modifier->cssStyles[$key]);
				}
			}
		}
		foreach ($modifier->attributes as $key => $val) {
			if (!$this->isValidAttributeValue($tagName, $key, $val)) {
				unset($modifier->attributes[$key]);
			}
		}
	}

	/**
	 * Tests if an HTML tag name is permitted.
	 */
	public function isValidTagName(string $tagName): bool {
		return ($this->allowableTags[mb_strtolower($tagName)] ?? null) !== null;
	}

	/**
	 * Tests if an HTML attribute name is permitted.
	 */
	public function isValidAttributeName(?string $tagName, string $attributeName): bool {
		$lcAttributeName = mb_strtolower($attributeName);
		if (($this->allowableGlobalAttributes[$lcAttributeName] ?? null) !== null) {
			return true;
		}
		foreach ($this->allowableGlobalAttributes as $pattern => $valuePattern) {
			if (!str_ends_with($pattern, '*')) continue;
			$patternPrefix = mb_substr($pattern, 0, mb_strlen($pattern) - 1);
			if (str_starts_with($lcAttributeName, $patternPrefix)) {
				return true;
			}
		}
		if ($tagName === null) return false;
		$lcTagName = mb_strtolower($tagName);
		$tagAttributes = $this->allowableTags[$lcTagName];
		if ($tagAttributes !== null) {
			return ($tagAttributes[$lcAttributeName] ?? null) !== null;
		}
		return false;
	}

	/**
	 * Tests if an attribute value is allowable.
	 */
	public function isValidAttributeValue(?string $tagName, string $attributeName, $attributeValue): bool {
		$lcAttributeName = mb_strtolower($attributeName);
		$globalPattern = $this->allowableGlobalAttributes[$lcAttributeName] ?? null;
		if ($globalPattern !== null) {
			return $this->attributeValueMatchesPattern($attributeValue, $globalPattern);
		}
		foreach ($this->allowableGlobalAttributes as $namePattern => $valuePattern) {
			if (str_ends_with($namePattern, '*') && str_starts_with($lcAttributeName, mb_substr($namePattern, 0, mb_strlen($namePattern) - 1))) {
				return $this->attributeValueMatchesPattern($attributeValue, $valuePattern);
			}
		}
		if ($tagName === null) return false;
		$lcTagName = mb_strtolower($tagName);
		$tagAttributes = $this->allowableTags[$lcTagName] ?? null;
		if ($tagAttributes === null) return false;
		$valuePattern = $tagAttributes[$lcAttributeName] ?? null;
		if ($valuePattern === null) return false;
		return $this->attributeValueMatchesPattern($attributeValue, $valuePattern);
	}

	private static string $permissiveURLRegex = '^\\S+$';
	private static string $integerRegex = '^[\\-]?\\d+$';
	private static string $classListRegex = '^-?[_a-zA-Z]+[_a-zA-Z0-9-]*(?:\\s+-?[_a-zA-Z]+[_a-zA-Z0-9-]*)*$';

	private function attributeValueMatchesPattern(string|bool $value, string $pattern): bool {
		$options = explode('|', $pattern);
		foreach ($options as $option) {
			switch ($option) {
				case '*':
					return true;
				case '{classlist}':
					if (mb_eregi(self::$classListRegex, $value)) return true;
					break;
				case '{int}':
					if (mb_eregi(self::$integerRegex, $value)) return true;
					break;
				case '{none}':
					if ($value === true) return true;
					break;
				case '{style}':
					if ($this->isValidStyleDeclaration($value)) return true;
					break;
				case '{url}':
					if (mb_eregi(self::$permissiveURLRegex, $value)) return true;
					break;
				default:
					if ($value === $option) return true;
					break;
			}
		}
		return false;
	}

	/**
	 * Tests if a string of one or more style `key: value;` declarations is
	 * fully allowable.
	 */
	public function isValidStyleDeclaration(string $styles): bool {
		$settings = explode(';', $styles);
		foreach ($settings as $setting) {
			if (mb_strlen(trim($setting)) == 0) continue;
			$parts = explode(':', $setting);
			if (sizeof($parts) != 2) return false;
			$name = trim($parts[0]);
			if (!$this->isValidStyleKey($name)) return false;
			$value = trim($parts[1]);
			if (!$this->isValidStyleValue($name, $value)) return false;
		}
		return true;
	}

	/**
	 * Tests if a CSS style key is allowable.
	 */
	public function isValidStyleKey(string $key): bool {
		return ($this->allowableStyleKeys[$key] ?? null) !== null;
	}

	/**
	 * Tests if a CSS style value is allowable.
	 */
	public function isValidStyleValue(string $key, string $value): bool {
		$pattern = $this->allowableStyleKeys[$key] ?? null;
		if ($pattern === null) return false;
		$options = explode('|', $pattern);
		foreach ($options as $option) {
			switch ($option) {
				case '{color}':
					if ($this->isValidCSSColor($value)) return true;
				default:
					if ($value === $option) return true;
			}
		}
		return false;
	}

	private static string $styleColorRegex = '^#[0-9a-f]{3}(?:[0-9a-f]{3})?$|^[a-zA-Z]+$';

	private function isValidCSSColor(string $value): bool {
		return mb_eregi(self::$styleColorRegex, $value);
	}
}

/**
 * Represents a single HTML tag. Paired tags are represented separately.
 */
class MDHTMLTag {
	/**
	 * Verbatim string of the original parsed tag. Not modified. Should be
	 * considered unsafe for inclusion in the final document. Use `toString()`
	 * instead.
	 */
	public string $original;
	public string $tagName;
	public bool $isCloser;
	/**
	 * Map of attribute names to value strings.
	 */
	public array $attributes;

	/**
	 * @param string $original
	 * @param string $tagName
	 * @param bool $isCloser
	 * @param array $attributes
	 */
	public function __construct(string $original, string $tagName, bool $isCloser,
			array $attributes) {
		$this->original = $original;
		$this->tagName = $tagName;
		$this->isCloser = $isCloser;
		$this->attributes = $attributes;
	}

	public function __toString(): string {
		if ($this->isCloser) {
			return "</{$this->tagName}>";
		}
		$html = '<';
		$html .= $this->tagName;
		foreach ($this->attributes as $key => $value) {
			$safeName = MDUtils::scrubAttributeName($key);
			if ($value === true) {
				$html .= " {$safeName}";
			} else {
				$escapedValue = htmlentities("{$value}");
				$html .= " {$safeName}=\"{$escapedValue}\"";
			}
		}
		$html .= '>';
		return $html;
	}

	public function equals($other): bool {
		if (!($other instanceof MDHTMLTag)) return false;
		if ($other->tagName != $this->tagName) return false;
		if ($other->isCloser != $this->isCloser) return false;
		return MDUtils::equal($other->attributes, $this->attributes);
	}

	private static string $htmlTagNameFirstRegex = '[a-z]';
	private static string $htmlTagNameMedialRegex = '[a-z0-9]';
	private static string $htmlAttributeNameFirstRegex = '[a-z]';
	private static string $htmlAttributeNameMedialRegex = '[a-z0-9-]';
	private static string $whitespaceCharRegex = '\\s';

	/**
	 * Checks the start of the given string for presence of an HTML tag.
	 */
	public static function fromLineStart(string $line): ?MDHTMLTag {
		$expectOpenBracket = 0;
		$expectCloserOrName = 1;
		$expectName = 2;
		$expectAttributeNameOrEnd = 3;
		$expectEqualsOrAttributeOrEnd = 4;
		$expectAttributeValue = 5;
		$expectCloseBracket = 6;

		$isCloser = false;
		$tagName = '';
		$attributeName = '';
		$attributeValue = '';
		$attributeQuote = null;
		$attributes = [];
		$fullTag = null;
		$endAttribute = function(bool $unescape=false) use (&$attributes, &$attributeName, &$attributeValue, &$attributeQuote) {
			if (mb_strlen($attributeName) > 0) {
				if (mb_strlen($attributeValue) > 0 || $attributeQuote !== null) {
					$attributes[$attributeName] = $unescape ? html_entity_decode($attributeValue, ENT_QUOTES | ENT_SUBSTITUTE | ENT_HTML401, 'UTF-8') : $attributeValue;
				} else {
					$attributes[$attributeName] = true;
				}
			}
			$attributeName = '';
			$attributeValue = '';
			$attributeQuote = null;
		};

		$expect = $expectOpenBracket;
		for ($p = 0; $p < mb_strlen($line) && $fullTag === null; $p++) {
			$ch = mb_substr($line, $p, 1);
			$isWhitespace = mb_eregi(self::$whitespaceCharRegex, $ch);
			switch ($expect) {
				case $expectOpenBracket:
					if ($ch != '<') return null;
					$expect = $expectCloserOrName;
					break;
				case $expectCloserOrName:
					if ($ch == '/') {
						$isCloser = true;
					} else {
						$p--;
					}
					$expect = $expectName;
					break;
				case $expectName:
					if (mb_strlen($tagName) == 0) {
						if (!mb_eregi(self::$htmlTagNameFirstRegex, $ch)) return null;
						$tagName .= $ch;
					} else {
						if (mb_eregi(self::$htmlTagNameMedialRegex, $ch)) {
							$tagName .= $ch;
						} else {
							$p--;
							$expect = ($isCloser) ? $expectCloseBracket : $expectAttributeNameOrEnd;
						}
					}
					break;
				case $expectAttributeNameOrEnd:
					if (mb_strlen($attributeName) == 0) {
						if ($isWhitespace) {
							// skip whitespace
						} elseif ($ch == '/') {
							$expect = $expectCloseBracket;
						} elseif ($ch == '>') {
							$fullTag = mb_substr($line, 0, $p + 1);
							break;
						} elseif (mb_eregi(self::$htmlAttributeNameFirstRegex, $ch)) {
							$attributeName .= $ch;
						} else {
							return null;
						}
					} elseif ($isWhitespace) {
						$expect = $expectEqualsOrAttributeOrEnd;
					} elseif ($ch == '/') {
						$endAttribute();
						$expect = $expectCloseBracket;
					} elseif ($ch == '>') {
						$endAttribute();
						$fullTag = mb_substr($line, 0, $p + 1);
						break;
					} elseif ($ch == '=') {
						$expect = $expectAttributeValue;
					} elseif (mb_eregi(self::$htmlAttributeNameMedialRegex, $ch)) {
						$attributeName .= $ch;
					} else {
						return null;
					}
					break;
				case $expectEqualsOrAttributeOrEnd:
					if ($ch == '=') {
						$expect = $expectAttributeValue;
					} elseif ($isWhitespace) {
						// skip whitespace
					} elseif ($ch == '/') {
						$expect = $expectCloseBracket;
					} elseif ($ch == '>') {
						$fullTag = mb_substr($line, 0, $p + 1);
						break;
					} elseif (mb_eregi(self::$htmlAttributeNameFirstRegex, $ch)) {
						$endAttribute();
						$expect = $expectAttributeNameOrEnd;
						$p--;
					}
					break;
				case $expectAttributeValue:
					if (mb_strlen($attributeValue) == 0) {
						if ($attributeQuote === null) {
							if ($isWhitespace) {
								// skip whitespace
							} elseif ($ch == '"' || $ch == "'") {
								$attributeQuote = $ch;
							} else {
								$attributeQuote = ''; // explicitly unquoted
								$p--;
							}
						} else {
							if ($ch === $attributeQuote) {
								// Empty string
								$endAttribute($attributeQuote != '');
								$expect = $expectAttributeNameOrEnd;
							} elseif ($attributeQuote === '' && ($ch == '/' || $ch == '>')) {
								return null;
							} else {
								$attributeValue .= $ch;
							}
						}
					} else {
						if ($ch === $attributeQuote) {
							$endAttribute($attributeQuote != '');
							$expect = $expectAttributeNameOrEnd;
						} elseif ($attributeQuote === '' && $isWhitespace) {
							$endAttribute();
							$expect = $expectAttributeNameOrEnd;
						} else {
							$attributeValue .= $ch;
						}
					}
					break;
				case $expectCloseBracket:
					if ($isWhitespace) {
						// ignore whitespace
					} elseif ($ch == '>') {
						$fullTag = mb_substr($line, 0, $p + 1);
						break;
					}
					break;
			}
		}
		if ($fullTag === null) return null;
		$endAttribute();
		return new MDHTMLTag($fullTag, $tagName, $isCloser, $attributes);
	}
}

/**
 * Represents HTML modifications to a node, such as CSS classes to add or
 * additional attributes. See `MDHTMLFilter.scrubModifier()` to remove disallowed
 * values.
 */
class MDTagModifier {
	/**
	 * Verbatim markdown syntax. Unmodified by changes to other properties.
	 */
	public string $original;
	/** @var string[] */
	public array $cssClasses = [];
	public ?string $cssId = null;
	public array $cssStyles = [];
	public array $attributes = [];

	private static $baseClassRegex = '\\.([a-z_\\-][a-z0-9_\\-]*?)';
	private static $baseIdRegex = '#([a-z_\\-][a-z0-9_\\-]*?)';
	private static $baseAttributeRegex = '([a-z0-9]+?)=([^\\s\\}]+?)';
	private static $baseRegex = '\\{([^}]+?)}';
	private static $leadingClassRegex = '^\\{([^}]+?)}';
	private static $trailingClassRegex = '^(.*?)\\s*\\{([^}]+?)}\\s*$';
	private static $classRegex = '^\\.([a-z_\\-][a-z0-9_\\-]*?)$';  // 1=classname
	private static $idRegex = '^#([a-z_\\-][a-z0-9_\\-]*?)$';  // 1=id
	private static $attributeRegex = '^([a-z0-9]+?)=([^\\s\\}]+?)$';  // 1=attribute name, 2=attribute value

	public function applyTo(MDNode $node) {
		if ($node instanceof MDNode) {
			foreach ($this->cssClasses as $cssClass) {
				$node->addClass($cssClass);
			}
			if ($this->cssId) $node->cssId = $this->cssId;
			foreach ($this->attributes as $name => $value) {
				$node->attributes[$name] = $value;
			}
			foreach ($this->cssStyles as $name => $value) {
				$node->cssStyles[$name] = $value;
			}
		}
	}

	/**
	 * Adds a CSS class. If already present it will not be duplicated.
	 */
	public function addClass(string $cssClass): bool {
		if (array_search($cssClass, $this->cssClasses) !== false) return false;
		array_push($this->cssClasses, $cssClass);
		return true;
	}

	/**
	 * Removes a CSS class.
	 */
	public function removeClass(string $cssClass): bool {
		$beforeLength = sizeof($this->cssClasses);
		$this->cssClasses = array_diff($this->cssClasses, [ $cssClass ]);
		return sizeof($this->cssClasses) != beforeLength;
	}

	public function equals($other): bool {
		if (!($other instanceof MDTagModifier)) return false;
		if (!MDUtils::equal($other->cssClasses, $this->cssClasses)) return false;
		if ($other->cssId !== $this->cssId) return false;
		if (!MDUtils::equal($other->attributes, $this->attributes)) return false;
		return true;
	}

	public function __toString(): string {
		return $this->original;
	}

	private static function styleToObject(string $styleValue): array {
		$pairs = explode(';', $styleValue);
		$styles = [];
		foreach ($pairs as $pair) {
			$keyAndValue = explode(':', $pair);
			if (sizeof($keyAndValue) != 2) continue;
			$styles[$keyAndValue[0]] = $keyAndValue[1];
		}
		return $styles;
	}

	private static function fromContents(string $contents): ?MDTagModifier {
		$modifierTokens = mb_split('\\s+', $contents);
		$mod = new MDTagModifier();
		$mod->original = "{{$contents}}";
		foreach ($modifierTokens as $token) {
			if (trim($token) == '') continue;
			if (mb_eregi(self::$classRegex, $token, $groups)) {
				$mod->addClass($groups[1]);
			} elseif (mb_eregi(self::$idRegex, $token, $groups)) {
				$mod->cssId = $groups[1];
			} elseif (mb_eregi(self::$attributeRegex, $token, $groups)) {
				if ($groups[1] == 'style') {
					$mod->cssStyles = self::styleToObject($groups[2]);
				} else {
					$mod->attributes[$groups[1]] = $groups[2];
				}
			} else {
				return null;
			}
		}
		return $mod;
	}

	/**
	 * Extracts block modifier from end of a line. Always returns a 2-element
	 * tuple array:
	 * - `0`: the line without the modifier
	 * - `1`: an `MDTagModifier` if found or `null` if not
	 *
	 * @param string $line
	 * @param ?MDState $state
	 * @return array tuple with remaining line and `MDTagModifier` or `null`
	 */
	public static function fromLine(string $line, ?MDState $state): array {
		if ($state) {
			$found = false;
			foreach ($state->root()->readersByBlockPriority as $reader) {
				if ($reader instanceof MDModifierReader) {
					$found = true;
					break;
				}
			}
			if (!$found) return [ $line, null ];
		}
		if (!mb_eregi(self::$trailingClassRegex, $line, $groups)) return [ $line, null ];
		$bareLine = $groups[1];
		$mod = self::fromContents($groups[2]);
		return [ $bareLine, $mod ];
	}

	/**
	 * Attempts to extract modifier from head of string.
	 */
	public static function fromStart(string $line): ?MDTagModifier {
		if (!mb_eregi(self::$leadingClassRegex, $line, $groups)) return null;
		return self::fromContents($groups[1]);
	}

	/**
	 * Discards any modifiers from a line and returns what remains.
	 */
	public static function strip(string $line): string {
		if (!mb_eregi(self::$trailingClassRegex, $line, $groups)) return $line;
		return $groups[1];
	}
}


// -- Readers ---------------------------------------------------------------


/**
 * Base class for readers of various markdown syntax. A `Markdown` instance can
 * be created with any combination of subclasses of these to customize the
 * flavor of markdown parsed.
 * 
 * @see {@link custom.md} for details on subclassing
 */
class MDReader {
	/**
	 * Called before processing begins. `state.lines` is populated and the
	 * line pointer `state.p` will be at `0`.
	 *
	 * Default implementation does nothing.
	 */
	public function preProcess(MDState $state) {}

	/**
	 * Attempts to read an `MDBlockNode` subclass at the current line pointer
	 * `state.p`. Only matches if the block pattern starts at the line pointer,
	 * not elsewhere in the `state.lines` array. If a block is found, `state.p`
	 * should be incremented to the next line _after_ the block structure and
	 * a `MDBlockNode` subclass instance is returned. If no block is found,
	 * returns `null`.
	 *
	 * Default implementation always returns `null`.
	 */
	public function readBlock(MDState $state): ?MDBlockNode { return null; }

	/**
	 * Attempts to read an inline token from the beginning of `line`. Only the
	 * start of the given `line` is considered. If a matching token is found, an
	 * `MDToken` is returned. Otherwise `null` is returned.
	 *
	 * Default implementation always returns `null`.
	 */
	public function readToken(MDState $state, string $line): ?MDToken { return null; }

	/**
	 * Attempts to find a pattern anywhere in `tokens` and perform a _single_
	 * in-place substitution with one or more `MDNode` subclass instances.
	 * If a substitution is performed, must return `true`, otherwise `false`.
	 *
	 * Default implementation always returns `false`.
	 *
	 * @param MDState $state
	 * @param int $pass  what substitution pass this is, starting with 1
	 * @param (MDToken|MDInlineNode)[] $tokens  mixed array of `MDToken` and `MDInlineNode` elements
	 * @return bool  `true` if a substitution was performed, `false` if not
	 */
	public function substituteTokens(MDState $state, int $pass, array &$tokens): bool { return false; }

	/**
	 * Called after all parsing has completed. An array `blocks` is passed of
	 * all the top-level `MDBlockNode` elements in the document which this
	 * method can traverse or alter in-place via `.splice` operations if
	 * necessary.
	 *
	 * `MDNode.visitChildren` is useful for recursively looking for certain
	 * `MDNode` instances. `MDNode.replaceNodes` is useful for swapping in
	 * replacements.
	 *
	 * Default implementation does nothing.
	 *
	 * @param MDState $state
	 * @param MDBlockNode[] $blocks
	 */
	public function postProcess(MDState $state, array &$blocks) {}

	/**
	 * Can be overridden to influence ordering of this reader with respect to
	 * another during the block parsing phase. Return `-1` to be ordered before
	 * the given reader, `1` to be ordered after it, or `0` for no preference.
	 * Only return non-`0` values to resolve specific conflicts.
	 *
	 * Default implementation always returns `0` (no preference).
	 *
	 * @param MDReader $other
	 * @return int  a negative, positive, or 0 value to be ordered before,
	 *   after, or anwhere relative to `other`, respectively
	 */
	public function compareBlockOrdering(MDReader $other): int {
		return 0;
	}

	/**
	 * Can be overridden to influence ordering of this reader with respect to
	 * another during the tokenizing phase. Return `-1` to be ordered before
	 * the given reader, `1` to be ordered after it, or `0` for no preference.
	 * Only return non-`0` values to resolve specific conflicts.
	 *
	 * Default implementation always returns `0` (no preference).
	 *
	 * @param MDReader $other
	 * @return int  a negative, positive, or 0 value to be ordered before,
	 *   after, or anwhere relative to `other`, respectively
	 */
	public function compareTokenizeOrdering(MDReader $other): int {
		return 0;
	}

	/**
	 * Can be overridden to influence ordering of this reader with respect to
	 * another during the substitution phase. Return `-1` to be ordered before
	 * the given reader, `1` to be ordered after it, or `0` for no preference.
	 * Only return non-`0` values to resolve specific conflicts.
	 *
	 * Readers are sorted within each substitution pass. All pass 1 readers are
	 * processed first, then all pass 2 readers, etc. The number of passes this
	 * reader participates in is dictated by `substitionPassCount`.
	 *
	 * Default implementation always returns `0` (no preference).
	 *
	 * @param MDReader $other
	 * @param int $pass  substitution pass, with numbering starting at `1`
	 * @return int  a negative, positive, or 0 value to be ordered before,
	 *   after, or anwhere relative to `other`, respectively
	 */
	public function compareSubstituteOrdering(MDReader $other, int $pass): int {
		return 0;
	}

	/**
	 * How many substitution passes this reader requires. Substitution allows
	 * all pass 1 readers to process first, then all pass 2 readers, etc.
	 */
	public function substitutionPassCount(): int { return 1; }

	/**
	 * For sorting readers with ordering preferences. The `compare` methods
	 * don't have the properties of normal sorting compares so need to sort
	 * differently.
	 *
	 * @param MDReader[] $arr  array to sort
	 * @param callable $compareFn  comparison function, taking two array element
	 *   arguments and returning -1, 0, or 1 for a < b, a == b, and a > b,
	 *   respectively
	 * @param callable $idFn  function for returning a unique hashable id for
	 *   the array element
	 * @return MDReader[]  sorted array
	 */
	private static function kahnTopologicalSort(array $arr, callable $compareFn, callable $idFn): array {
		$graph = [];
		$inDegrees = [];
		$valuesById = [];

		// Build the graph and compute in-degrees
		foreach ($arr as $index => $elem) {
			$id = $idFn($elem);
			$graph[$id] = [];
			$inDegrees[$id] = 0;
			$valuesById[$id] = $elem;
		}

		for ($i = 0; $i < sizeof($arr); $i++) {
			$elemA = $arr[$i];
			$idA = $idFn($elemA);
			for ($j = 0; $j < sizeof($arr); $j++) {
				if ($i === $j) continue;
				$elemB = $arr[$j];
				$idB = $idFn($elemB);
				$comparisonResult = $compareFn($elemA, $elemB);
				if ($comparisonResult < 0) {
					array_push($graph[$idA], $idB);
					$inDegrees[$idB]++;
				} elseif ($comparisonResult > 0) {
					array_push($graph[$idB], $idA);
					$inDegrees[$idA]++;
				}
			}
		}

		// Initialize the queue with zero-inDegree nodes
		$queue = [];
		foreach ($inDegrees as $elemId => $degree) {
			if ($degree === 0) {
				array_push($queue, $elemId);
			}
		}

		// Process the queue and build the topological order list
		$sorted = [];
		while (sizeof($queue) > 0) {
			$elemId = array_shift($queue);
			array_push($sorted, $valuesById[$elemId]);
			unset($valuesById[$elemId]);

			foreach ($graph[$elemId] as $neighbor) {
				$inDegrees[$neighbor]--;
				if ($inDegrees[$neighbor] === 0) {
					array_push($queue, $neighbor);
				}
			}
		}
		// Anything left over can go at the end. No ordering dependencies.
		foreach ($valuesById as $elemId => $value) {
			array_push($sorted, $value);
		}

		return $sorted;
	}

	/**
	 * Returns a sorted array of readers by their block priority preferences.
	 *
	 * @param MDReader[] $readers
	 * @return MDReader[]  sorted readers
	 */
	public static function sortReaderForBlocks(array &$readers): array {
		$sorted = $readers;
		return self::kahnTopologicalSort($sorted, function(MDReader $a, MDReader $b): int {
			return $a->compareBlockOrdering($b);
		}, fn($elem) => get_class($elem));
	}

	/**
	 * Returns a sorted array of readers by their tokenization priority preferences.
	 *
	 * @param MDReader[] $readers
	 * @return MDReader[]  sorted readers
	 */
	public static function sortReadersForTokenizing(array &$readers): array {
		$sorted = $readers;
		return self::kahnTopologicalSort($sorted, function(MDReader $a, MDReader $b): int {
			return $a->compareTokenizeOrdering($b);
		}, fn($elem) => get_class($elem));
	}

	/**
	 * Returns a sorted array of tuples (arrays) containing the substitution
	 * pass number and reader instance, sorted by their substitution priority
	 * preferences.
	 *
	 * For readers with `substitutionPassCount` > `1`, the same reader will
	 * appear multiple times in the resulting array, one per pass.
	 *
	 * @param MDReader[] $readers
	 * @return MDReader[]  sorted array of tuples with the pass number and
	 *   reader instance in each
	 */
	public static function sortReadersForSubstitution(array &$readers): array {
		$tuples = [];
		$maxPass = 1;
		foreach ($readers as $reader) {
			$passCount = $reader->substitutionPassCount();
			for ($pass = 1; $pass <= $passCount; $pass++) {
				array_push($tuples, [[ $pass, $reader ]]);
			}
			$maxPass = max($maxPass, $pass);
		}
		$result = [];
		for ($pass = 1; $pass <= $maxPass; $pass++) {
			$readersThisPass = array_filter($tuples, fn($tup) => $tup[0] == $pass);
			$passResult = self::kahnTopologicalSort($readersThisPass, function(array $a, array $b) use ($pass): int {
				$aReader = $a[1];
				$bReader = $b[1];
				return $aReader->compareSubstituteOrdering($bReader, $pass);
			}, fn($elem) => get_class($elem[1]));
			$result = array_merge($result, $passResult);
		}
		return $result;
	}
}

/**
 * Reads markdown blocks for headings denoted with the underline syntax.
 *
 * Supports `MDTagModifier` suffixes.
 */
class MDUnderlinedHeadingReader extends MDReader {
	public function readBlock(MDState $state): ?MDBlockNode {
		$p = $state->p;
		if (!$state->hasLines(2)) return null;
		$modifier;
		$contentLine = trim($state->lines[$p++]);
		[$contentLine, $modifier] = MDTagModifier::fromLine($contentLine, $state);
		$underLine = trim($state->lines[$p++]);
		if ($contentLine == '') return null;
		if (mb_eregi('^=+$', $underLine)) {
			$state->p = $p;
			$block = new MDHeadingNode(1, $state->inlineMarkdownToNodes($contentLine));
			if ($modifier) $modifier->applyTo($block);
			return $block;
		}
		if (mb_eregi('^\-+$', $underLine)) {
			$state->p = $p;
			$block = new MDHeadingNode(2, $state->inlineMarkdownToNodes($contentLine));
			if ($modifier) $modifier->applyTo($block);
			return $block;
		}
		return null;
	}
}

/**
 * Reads markdown blocks for headings denoted with hash marks. Heading levels 1
 * to 6 are supported.
 *
 * Supports `MDTagModifier` suffixes.
 */
class MDHashHeadingReader extends MDReader {
	private static $hashHeadingRegex = '^(#{1,6})\\s*([^#].*?)\\s*\\#*\\s*$';  // 1=hashes, 2=content

	public function readBlock(MDState $state): ?MDBlockNode {
		$p = $state->p;
		$line = $state->lines[$p++];
		$modifier;
		[$line, $modifier] = MDTagModifier::fromLine($line, $state);
		if (!mb_eregi(self::$hashHeadingRegex, $line, $groups)) return null;
		$state->p = $p;
		$level = mb_strlen($groups[1]);
		$content = $groups[2];
		$block = new MDHeadingNode($level, $state->inlineMarkdownToNodes($content));
		if ($modifier) $modifier->applyTo($block);
		return $block;
	}
}

/**
 * Reads subtext blocks. Subtext is smaller, fainter text for things like
 * disclaimers or sources.
 *
 * Supports `MDTagModifier` suffixes.
 */
class MDSubtextReader extends MDReader {
	private static $subtextRegex = '^\\-#\\s*(.*?)\\s*$';  // 1=content

	public function readBlock(MDState $state): ?MDBlockNode {
		$p = $state->p;
		$line = $state->lines[$p++];
		$modifier;
		[$line, $modifier] = MDTagModifier::fromLine($line, $state);
		if (!mb_eregi(self::$subtextRegex, $line, $groups)) return null;
		$state->p = $p;
		$content = $groups[1];
		$block = new MDSubtextNode($state->inlineMarkdownToNodes($content));
		if ($modifier) $modifier->applyTo($block);
		return $block;
	}

	public function compareBlockOrdering(MDReader $other): int {
		if ($other instanceof MDUnorderedListReader) {
			return -1;
		}
		return 0;
	}
}

/**
 * Reads markdown blocks for blockquoted text.
 */
class MDBlockQuoteReader extends MDReader {
	public function readBlock(MDState $state): ?MDBlockNode {
		$blockquoteLines = [];
		$p = $state->p;
		while ($p < sizeof($state->lines)) {
			$line = $state->lines[$p++];
			if (str_starts_with($line, ">")) {
				array_push($blockquoteLines, $line);
			} else {
				break;
			}
		}
		if (sizeof($blockquoteLines) == 0) return null;
		$contentLines = array_map(fn($line) => mb_eregi_replace('^ {0,3}\\t?', '', mb_substr($line, 1)), $blockquoteLines);
		$substate = $state->copy($contentLines);
		$quotedBlocks = $substate->readBlocks();
		$state->p = $p;
		return new MDBlockquoteNode($quotedBlocks);
	}
}

/**
 * Internal abstract base class for ordered and unordered lists.
 */
class _MDListReader extends MDReader {
	private static function readItemLines(MDState $state, int $firstLineStartPos): array {
		$p = $state->p;
		$lines = [];
		$seenBlankLine = false;
		$stripTrailingBlankLines = true;
		while ($state->hasLines(1, $p)) {
			$isFirstLine = ($p == $state->p);
			$line = $state->lines[$p++];
			if ($isFirstLine) {
				$line = mb_substr($line, $firstLineStartPos);
			}
			if (mb_eregi('^(?:\\*|\\+|\\-|\\d+\\.)\\s+', $line)) {
				// Found next list item
				$stripTrailingBlankLines = false; // because this signals extra spacing intended
				break;
			}
			$isBlankLine = trim($line) == '';
			$isIndented = mb_eregi('^\\s+\\S', $line);
			if ($isBlankLine) {
				$seenBlankLine = true;
			} elseif (!$isIndented && $seenBlankLine) {
				// Post-list content
				break;
			}
			array_push($lines, $line);
		}
		$lines = MDUtils::withoutTrailingBlankLines($lines);
		return MDUtils::stripIndent($lines);
	}

	protected function readListItemContent(MDState $state, int $firstLineStartPos): MDBlockNode {
		$itemLines = $this->readItemLines($state, $firstLineStartPos);
		$state->p += max(sizeof($itemLines), 1);

		if (sizeof($itemLines) == 1) {
			return $state->inlineMarkdownToNode($itemLines[0]);
		}

		$hasBlankLines = sizeof(array_filter($itemLines, fn($line) => trim($line) == '')) > 0;
		if ($hasBlankLines) {
			$substate = $state->copy($itemLines);
			$blocks = $substate->readBlocks();
			return (sizeof($blocks) == 1) ? $blocks[0] : new MDNode($blocks);
		}

		// Multiline content with no blank lines. Search for new block
		// boundaries without the benefit of a blank line to demarcate it.
		for ($p = 1; $p < sizeof($itemLines); $p++) {
			$line = $itemLines[$p];
			if (mb_eregi('^(?:\\*|\\-|\\+|\\d+\\.)\\s+', $line)) {
				// Nested list found
				$firstBlock = $state->inlineMarkdownToNode(implode("\n", array_slice($itemLines, 0, $p)));
				$substate = $state->copy(array_slice($itemLines, $p));
				$blocks = $substate->readBlocks();
				return array_merge([ $firstBlock, $blocks ]);
			}
		}

		// Ok, give up and just do a standard block read
		{
			$substate = $state->copy($itemLines);
			$blocks = $substate->readBlocks();
			return (sizeof($blocks) == 1) ? $blocks[0] : new MDNode($blocks);
		}
	}

	public function readBlock(MDState $state): ?MDBlockNode {
		$className = get_class($this);
		throw new Error("Abstract readBlock must be overridden in {$className}");
	}
}

/**
 * Block reader for unordered (bulleted) lists.
 */
class MDUnorderedListReader extends _MDListReader {
	private static string $unorderedListRegex = '^([\\*\\+\\-]\\s+)(.*)$';  // 1=bullet, 2=content

	private function readUnorderedListItem(MDState $state): ?MDListItemNode {
		$p = $state->p;
		$line = $state->lines[$p];
		if (!mb_eregi(self::$unorderedListRegex, $line, $groups)) return null;
		$firstLineOffset = mb_strlen($groups[1]);
		return new MDListItemNode($this->readListItemContent($state, $firstLineOffset));
	}

	public function readBlock(MDState $state): ?MDBlockNode {
		$items = [];
		$item = null;
		do {
			$item = $this->readUnorderedListItem($state);
			if ($item) array_push($items, $item);
		} while ($item);
		if (sizeof($items) == 0) return null;
		return new MDUnorderedListNode($items);
	}
}

/**
 * Block reader for ordered (numbered) lists. The number of the first item is
 * used to begin counting. The subsequent items increase by 1, regardless of
 * their value.
 */
class MDOrderedListReader extends _MDListReader {
	private static string $orderedListRegex = '^(\\d+)(\\.\\s+)(.*)$';  // 1=number, 2=dot, 3=content

	private function readOrderedListItem(MDState $state): ?MDListItemNode {
		$p = $state->p;
		$line = $state->lines[$p];
		if (!mb_eregi(self::$orderedListRegex, $line, $groups)) return null;
		$ordinal = intval($groups[1]);
		$firstLineOffset = mb_strlen($groups[1]) + mb_strlen($groups[2]);
		return new MDListItemNode($this->readListItemContent($state, $firstLineOffset), $ordinal);
	}

	public function readBlock(MDState $state): ?MDBlockNode {
		$items = [];
		$item = null;
		do {
			$item = $this->readOrderedListItem($state);
			if ($item) array_push($items, $item);
		} while ($item);
		if (sizeof($items) == 0) return null;
		return new MDOrderedListNode($items, $items[0]->ordinal);
	}
}

/**
 * Block reader for code blocks denoted by pairs of triple tickmarks. If
 * a programming language name, _xyz_, immediately follows the backticks, a
 * `language-xyz` CSS class will be added to the resulting `<code>`
 * element.
 *
 * Supports `MDTagModifier` suffix.
 */
class MDFencedCodeBlockReader extends MDReader {
	public function readBlock(MDState $state): ?MDBlockNode {
		if (!$state->hasLines(2)) return null;
		$p = $state->p;
		$openFenceLine = $state->lines[$p++];
		[$openFenceLine, $modifier] = MDTagModifier::fromLine($openFenceLine, $state);
		if (!mb_eregi('```\s*([a-z0-9]*)\s*$', $openFenceLine, $groups)) return null;
		$language = mb_strlen($groups[1]) > 0 ? $groups[1] : null;
		$codeLines = [];
		while ($state->hasLines(1, $p)) {
			$line = $state->lines[$p++];
			if (trim($line) == '```') {
				$state->p = $p;
				$block = new MDCodeBlockNode(implode("\n", $codeLines), $language);
				if ($modifier) $modifier->applyTo($block);
				return $block;
			}
			array_push($codeLines, $line);
		}
		return null;
	}
}

/**
 * Block reader for code blocks denoted by indenting text.
 */
class MDIndentedCodeBlockReader extends MDReader {
	public function readBlock(MDState $state): ?MDBlockNode {
		$p = $state->p;
		$codeLines = [];
		while ($state->hasLines(1, $p)) {
			$line = $state->lines[$p++];
			if (MDUtils::countIndents($line, true) < 1) {
				$p--;
				break;
			}
			array_push($codeLines, MDUtils::stripIndent($line));
		}
		if (sizeof($codeLines) == 0) return null;
		$state->p = $p;
		return new MDCodeBlockNode(implode("\n", $codeLines));
	}
}

/**
 * Block reader for horizontal rules. Composed of three or more hypens or
 * asterisks on a line by themselves, with or without intermediate whitespace.
 */
class MDHorizontalRuleReader extends MDReader {
	private static string $horizontalRuleRegex = '^\\s*(?:\\-(?:\\s*\\-){2,}|\\*(?:\\s*\\*){2,})\\s*$';

	public function readBlock(MDState $state): ?MDBlockNode {
		$p = $state->p;
		$line = $state->lines[$p++];
		[$line, $modifier] = MDTagModifier::fromLine($line, $state);
		if (mb_eregi(self::$horizontalRuleRegex, $line)) {
			$state->p = $p;
			$block = new MDHorizontalRuleNode();
			if ($modifier) $modifier->applyTo($block);
			return $block;
		}
		return null;
	}

	public function compareBlockOrdering(MDReader $other): int {
		if ($other instanceof MDUnorderedListReader) {
			return -1;
		}
		return 0;
	}
}

/**
 * Block reader for tables.
 *
 * Supports `MDTagModifier` suffix.
 */
class MDTableReader extends MDReader {
	private function readTableRow(MDState $state, bool $isHeader): ?MDTableRowNode {
		if (!$state->hasLines(1)) return null;
		$p = $state->p;
		$line = MDTagModifier::strip(trim($state->lines[$p++]));
		if (!mb_eregi('.*\\|.*', $line)) return null;
		if (str_starts_with($line, '|')) $line = mb_substr($line, 1);
		if (str_ends_with($line, '|')) $line = mb_substr($line, 0, mb_strlen($line) - 1);
		$cellTokens = explode('|', $line);
		$cells = array_map(function($token) use ($isHeader) {
			$content = $state->inlineMarkdownToNode(trim($token));
			return $isHeader ? new MDTableHeaderCellNode($content) : new MDTableCellNode($content);
		}, $cellTokens);
		$state->p = $p;
		return new MDTableRowNode($cells);
	}

	/**
	 * @param string $line
	 * @return string[]
	 */
	private function parseColumnAlignments(string $line): array {
		$line = trim($line);
		if (str_starts_with($line, '|')) $line = mb_substr($line, 1);
		if (str_ends_with($line, '|')) $line = mb_substr($line, 0, mb_strlen($line) - 1);
		return array_map(function($token) {
			if (str_starts_with($token, ':')) {
				if (str_ends_with($token, ':')) {
					return 'center';
				}
				return 'left';
			} elseif (str_ends_with($token, ':')) {
				return 'right';
			}
			return null;
		}, mb_split('\\s*\\|\\s*', $line));
	}

	private static string $tableDividerRegex = '^\\s*[|]?\\s*(?:[:]?-+[:]?)(?:\\s*\\|\\s*[:]?-+[:]?)*\\s*[|]?\\s*$';

	public function readBlock(MDState $state): ?MDBlockNode {
		if (!$state->hasLines(2)) return null;
		$startP = $state->p;
		$firstLine = $state->lines[$startP];
		$modifier = MDTagModifier::fromLine($firstLine, $state)[1];
		$headerRow = $this->readTableRow($state, true);
		if ($headerRow === null) {
			$state->p = $startP;
			return null;
		}
		$dividerLine = $state->lines[$state->p++];
		if (!mb_eregi(self::$tableDividerRegex, $dividerLine, $dividerGroups)) {
			$state->p = $startP;
			return null;
		}
		$columnAlignments = $this->parseColumnAlignments($dividerLine);
		$bodyRows = [];
		while ($state->hasLines(1)) {
			$row = $this->readTableRow($state, false);
			if ($row === null) break;
			array_push($bodyRows, $row);
		}
		$table = new MDTableNode($headerRow, $bodyRows);
		$table->columnAlignments = $columnAlignments;
		if ($modifier) $modifier->applyTo($table);
		return $table;
	}
}

/**
 * Block reader for definition lists. Definitions go directly under terms starting
 * with a colon.
 */
class MDDefinitionListReader extends MDReader {
	public function readBlock(MDState $state): ?MDBlockNode {
		$p = $state->p;
		$groups;
		$termCount = 0;
		$definitionCount = 0;
		$defLines = [];
		while ($state->hasLines(1, $p)) {
			$line = $state->lines[$p++];
			if (trim($line) === '') {
				break;
			}
			if (mb_eregi('^\\s+', $line)) {
				if (sizeof($defLines) == 0) return null;
				$defLines[sizeof($defLines) - 1] .= "\n" . $line;
			} elseif (mb_eregi('^:\\s+', $line)) {
				array_push($defLines, $line);
				$definitionCount++;
			} else {
				array_push($defLines, $line);
				$termCount++;
			}
		}
		if ($termCount == 0 || $definitionCount == 0) return null;
		$blocks = array_map(function($line) {
			if (mb_eregi('^:\\s+(.*?)$', $line)) {
				return new MDDefinitionListDefinitionNode($state->inlineMarkdownToNodes($groups[1]));
			} else {
				return new MDDefinitionListTermNode($state->inlineMarkdownToNodes($line));
			}
		}, $defLines);
		$state->p = $p;
		return new MDDefinitionListNode($blocks);
	}
}

/**
 * Block reader for defining footnote contents. Footnotes can be defined anywhere
 * in the document but will always be rendered at the end of a page or end of
 * the document.
 */
class MDFootnoteReader extends MDReader {
	private static string $footnoteWithTitleRegex = '^\\[\\^([^\\s\\[\\]]+?)\\s+"(.*?)"\\]';  // 1=symbol, 2=title
	private static string $footnoteRegex = '^\\[\\^([^\\s\\[\\]]+?)\\]';  // 1=symbol

	/**
	 * @param MDState $state
	 * @param string $symbol
	 * @param MDNode[] $footnote
	 */
	private function defineFootnote(MDState $state, string $symbol, array $footnote) {
		$footnotes = $state->root()->userInfo['footnotes'] ?? [];
		$footnotes[$symbol] = $footnote;
		$state->root()->userInfo['footnotes'] = $footnotes;
	}

	private function registerUniqueInstance(MDState $state, string $symbol, int $unique) {
		$footnoteInstances = $state->root()->userInfo['footnoteInstances'];
		$instances = $footnoteInstances[$symbol] ?? [];
		array_push($instances, $unique);
		$footnoteInstances[$symbol] = $instances;
	}

	private function idForFootnoteSymbol(MDState $state, string $symbol): int {
		$footnoteIds = $state->root()->userInfo['footnoteIds'];
		$existing = $footnoteIds[$symbol];
		if ($existing) return $existing;
		$nextFootnoteId = $state->root()->userInfo['nextFootnoteId'];
		$id = $nextFootnoteId++;
		$footnoteIds[$symbol] = $id;
		$state->root()->userInfo['nextFootnoteId'] = $nextFootnoteId;
		return $id;
	}

	public function preProcess(MDState $state) {
		$state->root()->userInfo['footnoteInstances'] = [];
		$state->root()->userInfo['footnotes'] = [];
		$state->root()->userInfo['footnoteIds'] = [];
		$state->root()->userInfo['nextFootnoteId'] = 1;
	}

	public function readBlock(MDState $state): ?MDBlockNode {
		$p = $state->p;
		if (!mb_eregi('^\\s*\\[\\^\\s*([^\\]]+)\\s*\\]:\\s+(.*)\\s*$', $state->lines[$p++], $groups)) return null;
		$symbol = $groups[1];
		$def = $groups[2];
		while ($state->hasLines(1, $p)) {
			$line = $state->lines[$p++];
			if (mb_eregi('^\\s+', $line)) {
				$def .= "\n" . $line;
			} else {
				$p--;
				break;
			}
		}
		$content = $state->inlineMarkdownToNodes($def);
		$this->defineFootnote($state, $symbol, $content);
		$state->p = $p;
		return new MDNode(); // empty
	}

	public function readToken(MDState $state, string $line): ?MDToken {
		$groups;
		if (mb_eregi(self::$footnoteWithTitleRegex, $line, $groups)) {
			return new MDToken($groups[0], MDTokenType::Footnote, $groups[1], $groups[2]);
		}
		if (mb_eregi(self::$footnoteRegex, $line, $groups)) {
			return new MDToken($groups[0], MDTokenType::Footnote, $groups[1]);
		}
		return null;
	}

	public function substituteTokens(MDState $state, int $pass, array &$tokens): bool {
		if ($match = self::findFirstTokens($tokens, [ MDTokenType::Footnote ])) {
			$symbol = $match->tokens[0]->content;
			array_splice($tokens, $match->index, 1, new MDFootnoteNode($symbol));
			return true;
		}
		return false;
	}

	/**
	 * @param {MDState} state
	 * @param {MDBlockNode[]} blocks
	 */
	public function postProcess(MDState $state, array &$blocks) {
		$nextOccurrenceId = 1;
		foreach ($blocks as $block) {
			$block->visitChildren(function($node) use (&$nextOccurrenceId) {
				if (!($node instanceof MDFootnoteNode)) return;
				$node->footnoteId = $this->idForFootnoteSymbol($state, $node->symbol);
				$node->occurrenceId = $nextOccurrenceId++;
				$node->displaySymbol = strval($node->footnoteId);
				$this->$registerUniqueInstance($state, $node->symbol, $node->occurrenceId);
			});
		}
		if (sizeof($state->userInfo['footnotes']) == 0) return;
		array_push($blocks, new MDFootnoteListNode());
	}

	public function compareBlockOrdering(MDReader $other): int {
		if ($other instanceof MDLinkReader || $other instanceof MDImageReader) {
			return -1;
		}
		return 0;
	}

	public function compareTokenizeOrdering(MDReader $other): int {
		if ($other instanceof MDLinkReader || $other instanceof MDImageReader) {
			return -1;
		}
		return 0;
	}

	public function compareSubstituteOrdering(MDReader $other, int $pass): int {
		if ($other instanceof MDLinkReader || $other instanceof MDImageReader) {
			return -1;
		}
		return 0;
	}
}

/**
 * Block reader for abbreviation definitions. Anywhere the abbreviation appears
 * in plain text will have its definition available when hovering over it.
 * Definitions can appear anywhere in the document. Their content should only
 * contain simple text, not markdown.
 */
class MDAbbreviationReader extends MDReader {
	private function defineAbbreviation(MDState $state, string $abbreviation, string $definition) {
		$state->root()->abbreviations[$abbreviation] = $definition;
		$regex = "\\b(" . preg_quote($abbreviation) . ")\\b";
		$state->root()->abbreviationRegexes[$abbreviation] = $regex;
	}

	public function preProcess(MDState $state) {
		$state->root()->userInfo['abbreviations'] = [];
		$state->root()->userInfo['abbreviationRegexes'] = [];
	}

	public function readBlock(MDState $state): ?MDBlockNode {
		$p = $state->p;
		$line = $state->lines[$p++];
		if (!mb_eregi('^\\s*\\*\\[([^\\]]+?)\\]:\\s+(.*?)\\s*$', $line, $groups)) return null;
		$abbrev = $groups[1];
		$def = $groups[2];
		$this->defineAbbreviation($state, $abbrev, $def);
		$state->p = $p;
		return new MDNode(); // empty
	}

	/**
	 * @param MDState $state
	 * @param MDNode[] $blocks
	 */
	public function postProcess(MDState $state, array &$blocks) {
		$abbreviations = $state->root()->userInfo['abbreviations'];
		$regexes = $state->root()->userInfo['abbreviationRegexes'];
		MDNode::replaceNodes($state, $blocks, function($original) use ($abbreviations, $regexes) {
			if (!($original instanceof MDTextNode)) return null;
			$changed = false;
			$elems = [ $original->text ]; // mix of strings and MDNodes
			for ($i = 0; $i < sizeof($elems); $i++) {
				$text = $elems[$i];
				if (!is_string($text)) continue;
				foreach ($abbreviations as $abbreviation) {
					$index = strpos($text, $abbreviation);
					if ($index === false) break;
					$prefix = substr($text, 0, $index);
					$suffix = substr($text, $index + strlen($abbreviation));
					$definition = $abbreviations[$abbreviation];
					array_splice($elems, $i, 1, [ $prefix, new MDAbbreviationNode($abbreviation, $definition), $suffix ]);
					$i = -1; // start over
					$changed = true;
					break;
				}
			}
			if (!$changed) return null;
			$nodes = array_map(fn($elem) => is_string($elem) ? new MDTextNode($elem) : $elem);
			return new MDNode($nodes);
		});
	}
}

/**
 * Block reader for simple paragraphs. Paragraphs are separated by a blank (or
 * whitespace-only) line. This reader is prioritized after every other reader
 * since there is no distinguishing syntax.
 */
class MDParagraphReader extends MDReader {
	public function readBlock(MDState $state): ?MDBlockNode {
		$paragraphLines = [];
		$p = $state->p;
		while ($state->hasLines(1, $p)) {
			$line = $state->lines[$p++];
			if (trim($line) === '') {
				break;
			}
			array_push($paragraphLines, $line);
		}
		if ($state->p == 0 && $p >= sizeof($state->lines)) {
			// If it's the entire document don't wrap it in a paragraph
			return null;
		}
		if (sizeof($paragraphLines) > 0) {
			$state->p = $p;
			$content = implode("\n", $paragraphLines);
			return new MDParagraphNode($state->inlineMarkdownToNodes($content));
		}
		return null;
	}

	public function compareBlockOrdering(MDReader $other): int {
		return 1; // always dead last
	}
}

/**
 * Abstract base class for readers that look for one or two delimiting tokens
 * on either side of some content. E.g. `**strong**`.
 */
class MDSimplePairInlineReader extends MDReader {
	// Passes:
	//   1. Syntaxes with two delimiting tokens, interior tokens of the same
	//      kind must be even in number
	//   2. Syntaxes with one delimiting token, interior tokens of the same
	//      kind must be even in number
	//   3. Syntaxes with two delimiting tokens, any tokens inside
	//   4. Syntaxes with one delimiting token, any tokens inside
	public function substitutionPassCount(): int { return 4; }

	/**
	 * Attempts a substitution of a matched pair of delimiting token types.
	 * If successful, the substitution is performed on `tokens` and `true` is
	 * returned, otherwise `false` is returned and the array is untouched.
	 *
	 * If `this.substitutionPassCount` is greater than 1, the first pass
	 * will reject matches with the delimiting character inside the content
	 * tokens. If the reader uses a single pass or a subsequent pass is performed
	 * with multiple pass any contents will be accepted.
	 *
	 * @param MDState $state
	 * @param int $pass  pass number, starting with `1`
	 * @param (MDToken|MDNode)[] $tokens  tokens/nodes to perform substitution on
	 * @param string $nodeClass  class of the node to return if matched
	 * @param MDTokenType $delimiter  delimiting token
	 * @param int $count  how many times the token is repeated to form the delimiter
	 * @param bool $plaintext  whether to invoke `nodeClass` with a verbatim
	 *   content string instead of parsed `MDNode`s
	 * @return bool  `true` if substitution was performed, `false` if not
	 */
	public function attemptPair(MDState $state, int $pass, array &$tokens, string $nodeClass, MDTokenType $delimiter, int $count=1, bool $plaintext=false): bool {
		// We do four passes. #1: doubles without inner tokens, #2: singles
		// without inner tokens, #3: doubles with paired inner tokens,
		// #4: singles with paired inner tokens
		if ($count == 1 && $pass != 2 && $pass != 4) return false;
		if ($count > 1 && $pass != 1 && $pass != 3) return false;
		$delimiters = [];
		array_fill(0, $count, $delimiter);
		$isFirstOfMultiplePasses = $this->substitutionPassCount() > 1 && $pass == 1;
		$match = MDToken::findPairedTokens($tokens, $delimiters, $delimiters, function($content) {
			$firstType = $content[0] instanceof MDToken ? $content[0]->type : null;
			$lastType = $content[sizeof($content) - 1] instanceof MDToken ? $content[sizeof($content) - 1]->type : null;
			if ($firstType == MDTokenType::Whitespace) return false;
			if ($lastType == MDTokenType::Whitespace) return false;
			foreach ($content as $token) {
				// Don't allow nesting
				if (get_class($token) == $nodeClass) return false;
			}
			if ($isFirstOfMultiplePasses) {
				$innerCount = 0;
				foreach ($content as $token) {
					if ($token instanceof MDToken && $token->type == $delimiter) $innerCount++;
				}
				if (($innerCount % 2) != 0) return false;
			}
			return true;
		});
		if ($match === null) return false;
		$content = ($plaintext)
			? implode('', array_map(fn($token) => $token->original, $match->contentTokens))
			: $state->tokensToNodes($match->contentTokens);
		$ref = new ReflectionClass($nodeClass);
		$node = $ref->newInstanceArgs([ $content ]);
		array_splice($tokens, $match->startIndex, $match->totalLength, [ $node ]);
		return true;
	}
}

/**
 * Reader for emphasis syntax. Denoted with a single underscore on either side of
 * some text (preferred) or a single asterisk on either side.
 */
class MDEmphasisReader extends MDSimplePairInlineReader {
	public function readToken(MDState $state, string $line): ?MDToken {
		if (str_starts_with($line, '_')) return new MDToken('_', MDTokenType::Underscore);
		if (str_starts_with($line, '*')) return new MDToken('*', MDTokenType::Asterisk);
		return null;
	}

	public function substituteTokens(MDState $state, int $pass, array &$tokens): bool {
		if ($this->attemptPair($state, $pass, $tokens, 'MDEmphasisNode', MDTokenType::Underscore)) return true;
		if ($this->attemptPair($state, $pass, $tokens, 'MDEmphasisNode', MDTokenType::Asterisk)) return true;
		return false;
	}

	public function compareSubstituteOrdering(MDReader $other, int $pass): int {
		if ($other instanceof MDStrongReader) {
			return 1;
		}
		return 0;
	}
}

/**
 * Reader for strong syntax. Denoted with two asterisks on either side of some
 * text (preferred) or two underscores on either side. Note that if
 * `MDUnderlineReader` is in use, it will replace the double-underscore syntax.
 */
class MDStrongReader extends MDSimplePairInlineReader {
	public function readToken(MDState $state, string $line): ?MDToken {
		if (str_starts_with($line, '*')) return new MDToken('*', MDTokenType::Asterisk);
		if (str_starts_with($line, '_')) return new MDToken('_', MDTokenType::Underscore);
		return null;
	}

	public function substituteTokens(MDState $state, int $pass, array &$tokens): bool {
		if ($this->attemptPair($state, $pass, $tokens, 'MDStrongNode', MDTokenType::Asterisk, 2)) return true;
		if ($this->attemptPair($state, $pass, $tokens, 'MDStrongNode', MDTokenType::Underscore, 2)) return true;
		return false;
	}

	public function compareSubstituteOrdering(MDReader $other, int $pass): int {
		if ($other instanceof MDEmphasisReader) {
			return -1;
		}
		return 0;
	}
}

/**
 * Reader for strikethrough syntax. Consists of two tildes on either side of
 * some text (preferred) or single tildes on either side. Note that if
 * `MDSubscriptReader` is in use, it will replace the single-tilde syntax.
 *
 * The number of recognized tildes can be configured.
 */
class MDStrikethroughReader extends MDSimplePairInlineReader {
	/** @type {boolean} */
	public bool $singleTildeEnabled = true;
	/** @type {boolean} */
	public bool $doubleTildeEnabled = true;

	public function readToken(MDState $state, string $line): ?MDToken {
		if (str_starts_with($line, '~')) return new MDToken('~', MDTokenType::Tilde);
		return null;
	}

	public function substituteTokens(MDState $state, int $pass, array &$tokens): bool {
		if ($this->singleTildeEnabled) {
			if ($this->attemptPair($state, $pass, $tokens, 'MDStrikethroughNode', MDTokenType::Tilde, 2)) return true;
		}
		if ($this->doubleTildeEnabled) {
			if ($this->attemptPair($state, $pass, $tokens, 'MDStrikethroughNode', MDTokenType::Tilde)) return true;
		}
		return false;
	}
}

/**
 * Reader for underline syntax. Consists of two underscores on either side of
 * some text. If used with `MDStrongReader` which also looks for double
 * underscores, this reader will take priority.
 */
class MDUnderlineReader extends MDSimplePairInlineReader {
	public function readToken(MDState $state, string $line): ?MDToken {
		if (str_starts_with($line, '_')) return new MDToken('_', MDTokenType::Underscore);
		return null;
	}

	public function substituteTokens(MDState $state, int $pass, array &$tokens): bool {
		return $this->attemptPair($state, $pass, $tokens, 'MDUnderlineNode', MDTokenType::Underscore, 2);
	}

	public function compareSubstituteOrdering(MDReader $other, int $pass): int {
		if ($other instanceof MDStrongReader) {
			return -1;
		}
		return 0;
	}
}

/**
 * Reader for highlight syntax. Consists of pairs of equal signs on either side
 * of some text.
 */
class MDHighlightReader extends MDSimplePairInlineReader {
	public function readToken(MDState $state, string $line): ?MDToken {
		if (str_starts_with($line, '=')) return new MDToken('=', MDTokenType::Equal);
		return null;
	}

	public function substituteTokens(MDState $state, int $pass, array &$tokens): bool {
		return $this->attemptPair($state, $pass, $tokens, 'MDHighlightNode', MDTokenType::Equal, 2);
	}
}

/**
 * Reader for inline code syntax. Consists of one or two delimiting backticks
 * around text. The contents between the backticks will be rendered verbatim,
 * ignoring any inner markdown syntax. To include a backtick inside, escape it
 * with a backslash.
 */
class MDCodeSpanReader extends MDSimplePairInlineReader {
	public function readToken(MDState $state, string $line): ?MDToken {
		if (str_starts_with($line, '`')) return new MDToken('`', MDTokenType::Backtick);
		return null;
	}

	public function substituteTokens(MDState $state, int $pass, array &$tokens): bool {
		if ($this->attemptPair($state, $pass, $tokens, 'MDCodeNode', MDTokenType::Backtick, 2, true)) return true;
		if ($this->attemptPair($state, $pass, $tokens, 'MDCodeNode', MDTokenType::Backtick, 1, true)) return true;
	}
}

/**
 * Reader for subscript syntax. Consists of single tildes on either side of
 * some text. If used with `MDStrikethroughReader`, this reader will take
 * precedence, and strikethrough can only be done with double tildes.
 */
class MDSubscriptReader extends MDSimplePairInlineReader {
	public function readToken(MDState $state, string $line): ?MDToken {
		if (str_starts_with($line, '~')) return new MDToken('~', MDTokenType::Tilde);
		return null;
	}

	public function substituteTokens(MDState $state, int $pass, array &$tokens): bool {
		return $this->attemptPair($state, $pass, $tokens, 'MDSubscriptNode', MDTokenType::Tilde);
	}

	public function compareSubstituteOrdering(MDReader $other, int $pass): int {
		if ($other instanceof MDStrikethroughReader) {
			return -1;
		}
		return 0;
	}
}

/**
 * Reader for superscript syntax. Consists of single caret characters on either
 * side of some text.
 */
class MDSuperscriptReader extends MDSimplePairInlineReader {
	public function readToken(MDState $state, string $line): ?MDToken {
		if (str_starts_with($line, '^')) return new MDToken('^', MDTokenType::Caret);
		return null;
	}

	public function substituteTokens(MDState $state, int $pass, array &$tokens): bool {
		return $this->attemptPair($state, $pass, $tokens, 'MDSuperscriptNode', MDTokenType::Caret);
	}
}

/**
 * Reads a hypertext link. Consists of link text between square brackets
 * followed immediately by a URL in parentheses.
 */
class MDLinkReader extends MDReader {
	public function readToken(MDState $state, string $line): ?MDToken {
		$simpleEmailRegex = "^<(" . MDUtils::$baseEmailRegex . ")>";
		$simpleURLRegex = "^<(" . MDUtils::$baseURLRegex . ")>";
		if ($groups = MDToken::tokenizeLabel($line)) {
			return new MDToken($groups[0], MDTokenType::Label, $groups[1]);
		}
		if ($groups = MDToken::tokenizeEmail($line)) {
			return new MDToken($groups[0], MDTokenType::Email, $groups[1], $groups[2]);
		}
		if ($groups = MDToken::tokenizeURL($line)) {
			return new MDToken($groups[0], MDTokenType::URL, $groups[1], $groups[2]);
		}
		if (mb_eregi($simpleEmailRegex, $line, $groups)) {
			return new MDToken($groups[0], MDTokenType::SimpleEmail, $groups[1]);
		}
		if (mb_eregi($simpleURLRegex, $line, $groups)) {
			return new MDToken($groups[0], MDTokenType::SimpleLink, $groups[1]);
		}
		return null;
	}

	public function substituteTokens(MDState $state, int $pass, array &$tokens): bool {
		if ($match = MDToken::findFirstTokens($tokens, [ MDTokenType::Label, MDTokenType::META_OptionalWhitespace, MDTokenType::URL ])) {
			$text = $match->tokens[0]->content;
			$url = $match->tokens[sizeof($match->tokens) - 1]->content;
			$title = $match->tokens[sizeof($match->tokens) - 1]->extra;
			array_splice($tokens, $match->index, sizeof($match->tokens), new MDLinkNode($url, $state->inlineMarkdownToNode($text), $title));
			return true;
		}
		if ($match = MDToken::findFirstTokens($tokens, [ MDTokenType::Label, MDTokenType::META_OptionalWhitespace, MDTokenType::Email ])) {
			$text = $match->tokens[0]->content;
			$email = $match->tokens[sizeof($match->tokens) - 1]->content;
			$url = "mailto:{$email}";
			$title = $match->tokens[sizeof($match->tokens) - 1]->extra;
			array_splice($tokens, $match->index, sizeof($match->tokens), new MDLinkNode($url, $state->inlineMarkdownToNodes($text), $title));
			return true;
		}
		if ($match = MDToken::findFirstTokens($tokens, [ MDTokenType::SimpleEmail ])) {
			$token = $match->tokens[0];
			$link = "mailto:{$token->content}";
			$node = new MDLinkNode($link, new MDObfuscatedTextNode($token->content));
			array_splice($tokens, $match->index, 1, $node);
			return true;
		}
		if ($match = MDToken::findFirstTokens($tokens, [ MDTokenType::SimpleLink ])) {
			$token = $match->tokens[0];
			$link = $token->content;
			$node = new MDLinkNode($link, new MDTextNode($link));
			array_splice($tokens, $match->index, 1, $node);
			return true;
		}
		return false;
	}
}

/**
 * Reader for referential URL definitions. Consists of link text between square
 * brackets followed immediately by a reference symbol also in square brackets.
 * The URL can be defined elsewhere on a line by itself with the symbol in square
 * brackets, colon, and the URL (and optional title in quotes).
 */
class MDReferencedLinkReader extends MDLinkReader {
	public function readBlock(MDState $state): ?MDBlockNode {
		$p = $state->p;
		$line = $state->lines[$p++];
		if (mb_eregi('^\\s*\\[(.+?)]:\\s*(\\S+)\\s+"(.*?)"\\s*$', $line, $groups)) {
			$symbol = $groups[1];
			$url = $groups[2];
			$title = $groups[3];
		} else {
			if (mb_eregi('^\\s*\\[(.+?)]:\\s*(\\S+)\\s*$', $line, $groups)) {
				$symbol = $groups[1];
				$url = $groups[2];
			} else {
				return null;
			}
		}
		$state->defineURL($symbol, $url, $title);
		$state->p = $p;
		return new MDNode([]); // empty
	}

	public function substituteTokens(MDState $state, int $pass, array &$tokens): bool {
		if ($match = MDToken::findFirstTokens($tokens, [ MDTokenType::Label, MDTokenType::META_OptionalWhitespace, MDTokenType::Label ])) {
			$text = $match->tokens[0]->content;
			$ref = $match->tokens[sizeof($match->tokens) - 1]->content;
			array_splice($tokens, $match->index, sizeof($match->tokens), new MDReferencedLinkNode($ref, $state->inlineMarkdownToNodes($text)));
			return true;
		}
		return false;
	}
}

/**
 * Reader for images. Consists of an exclamation, alt text in square brackets,
 * and image URL in parentheses.
 */
class MDImageReader extends MDLinkReader {
	public function readToken(MDState $state, string $line): ?MDToken {
		$s = parent::readToken($state, $line);
		if ($s) return $s;
		if (str_starts_with($line, '!')) return new MDToken('!', MDTokenType::Bang);
		return null;
	}

	public function substituteTokens(MDState $state, int $pass, array &$tokens): bool {
		if ($match = MDToken::findFirstTokens($tokens, [ MDTokenType::Bang, MDTokenType::Label, MDTokenType::META_OptionalWhitespace, MDTokenType::URL ])) {
			$alt = $match->tokens[1]->content;
			$url = $match->tokens[sizeof($match->tokens) - 1]->content;
			$title = $match->tokens[sizeof($match->tokens) - 1]->extra;
			$node = new MDImageNode($url, $alt);
			if ($title !== null) {
				$node->attributes['title'] = $title;
			}
			array_splice($tokens, $match->index, sizeof($match->tokens), $node);
			return true;
		}
		return false;
	}

	public function compareSubstituteOrdering(MDReader $other, int $pass): int {
		if (get_class($other) === 'MDLinkReader' || get_class($other) === 'MDReferencedLinkReader') {
			return -1;
		}
		return 0;
	}
}

/**
 * Reader for images with referential URL definitions. Consists of an
 * exclamation, alt text in square brackets, and link symbol in square brackets.
 * URL is defined the same as for `MDReferencedLinkReader`.
 */
class MDReferencedImageReader extends MDReferencedLinkReader {
	public function readToken(MDState $state, string $line): ?MDToken {
		$s = parent::readToken($state, $line);
		if ($s) return $s;
		if (str_starts_with($line, '!')) return new MDToken('!', MDTokenType::Bang);
		return null;
	}

	public function substituteTokens(MDState $state, int $pass, array &$tokens): bool {
		if ($match = MDToken::findFirstTokens($tokens, [ MDTokenType::Bang, MDTokenType::Label, MDTokenType::META_OptionalWhitespace, MDTokenType::Label ])) {
			$alt = $match->tokens[1]->content;
			$ref = $match->tokens[sizeof($match->tokens) - 1]->content;
			array_splice($tokens, $match->index, sizeof($match->tokens), new MDReferencedImageNode($ref, $alt));
			return true;
		}
		return false;
	}

	public function compareSubstituteOrdering(MDReader $other, int $pass): int {
		if (get_class($other) === 'MDLinkReader' || get_class($other) === 'MDReferencedLinkReader') {
			return -1;
		}
		return 0;
	}
}

/**
 * Converts line breaks within blocks into line breaks in the HTML. Not
 * included in any of the default reader sets since most flavors ignore
 * line breaks within blocks.
 */
class MDLineBreakReader extends MDReader {
	public function postProcess(MDState $state, array &$blocks) {
		MDNode::replaceNodes($state, $blocks, function(MDNode $original) {
			if (!($original instanceof MDTextNode)) return null;
			$lines = explode("\n", $original->text);
			if (sizeof($lines) == 1) return null;
			$nodes = [];
			foreach ($lines as $i => $line) {
				if ($i > 0) {
					array_push($nodes, new MDLineBreakNode());
				}
				array_push($nodes, new MDTextNode($line));
			}
			return new MDNode($nodes);
		});
	}
}

/**
 * Reads a verbatim HTML tag, and if it passes validation by `MDState.tagFilter`,
 * will be rendered in the final HTML document. Disallowed tags will be rendered
 * as plain text in the resulting document.
 */
class MDHTMLTagReader extends MDReader {
	public function readToken(MDState $state, string $line): ?MDToken {
		$tag = MDHTMLTag::fromLineStart($line, $state);
		if ($tag === null) return null;
		if (!$state->root()->tagFilter->isValidTagName($tag->tagName)) return null;
		$state->root()->tagFilter->scrubTag($tag);
		return new MDToken($tag->original, MDTokenType::HTMLTag, $tag);
	}

	public function substituteTokens(MDState $state, int $pass, array &$tokens): bool {
		if ($match = MDToken::findFirstTokens($tokens, [ MDTokenType::HTMLTag ])) {
			$tag = $match->tokens[0]->tag;
			array_splice($tokens, $match->index, sizeof($match->tokens), new MDHTMLTagNode($tag));
			return true;
		}
		return false;
	}
}

/**
 * Reads tag modifiers. Consists of curly braces with one or more CSS classes,
 * IDs, or custom attributes separated by spaces to apply to the preceding
 * node. Validation is performed on modifiers and only acceptable values are
 * applied.
 */
class MDModifierReader extends MDReader {
	public function readToken(MDState $state, string $line): ?MDToken {
		$modifier = MDTagModifier::fromStart($line);
		if ($modifier) return new MDToken($modifier->original, MDTokenType::Modifier, $modifier);
		return null;
	}

	public function substituteTokens(MDState $state, int $pass, array &$tokens): bool {
		// Modifiers are applied elsewhere, and if they're not it's fine if they're
		// rendered as the original syntax.
		return false;
	}
}


// -- Nodes -----------------------------------------------------------------


/**
 * Base class for nodes in the assembled document tree.
 */
class MDNode {
	/**
	 * Array of CSS classes to add to the node when rendered as HTML.
	 * @var string[]
	 */
	public array $cssClasses = [];

	public ?string $cssId = null;

	/**
	 * Mapping of CSS attributes to values.
	 * @var array
	 */
	public array $cssStyles = [];

	/**
	 * Mapping of arbitrary attributes and values to add to this node's top-level
	 * tag when rendered as HTML. For `class`, `id`, and `style` attributes, use
	 * `cssClasses`, `cssId`, and `cssStyles` instead.
	 * @var array
	 */
	public array $attributes = [];

	/**
	 * All child nodes in this node.
	 * @var MDNode[]
	 */
	public array $children = [];

	/**
	 * @param MDNode[]|MDNode $children
	 */
	public function __construct(array|MDNode $children=[]) {
		if (is_array($children)) {
			foreach ($children as $elem) {
				if (!($elem instanceof MDNode)) {
					$thisClassName = get_class($this);
					$elemClassName = get_class($elem);
					throw new Error("{$thisClassName} expects children of type MDNode[] or MDNode, got array with {$elemClassName} element");
				}
			}
			$this->children = $children;
		} elseif ($children instanceof MDNode) {
			$this->children = [ $children ];
		} else {
			$thisClassName = get_class($this);
			$elemClassName = gettype($children) == 'object' ? get_class($children) : gettype($children);
			throw new Error("{$thisClassName} expects children of type MDNode[] or MDNode, got {$elemClassName}");
		}
	}

	/**
	 * Adds a CSS class. If already present it will not be duplicated.
	 */
	public function addClass(string $cssClass): bool {
		if (array_search($cssClass, $this->cssClasses) !== false) return false;
		array_push($this->cssClasses, $cssClass);
		return true;
	}

	/**
	 * Removes a CSS class.
	 *
	 * @param {string} cssClass
	 * @returns {boolean} whether the class was present and removed
	 */
	public function removeClass(string $cssClass): bool {
		$beforeLength = sizeof($this->cssClasses);
		$this->cssClasses = array_diff($this->cssClasses, [ $cssClass ]);
		return sizeof($this->cssClasses) != $beforeLength;
	}

	/**
	 * Renders this node and any children as an HTML string. If the node has no
	 * content an empty string should be returned.
	 */
	public function toHTML(MDState $state): string {
		return MDNode::arrayToHTML($this->children, $state);
	}

	/**
	 * Renders this node and any children as a plain text string. The conversion
	 * should only render ordinary text, not attempt markdown-like formatting
	 * (e.g. list items should not be prefixed with asterisks, only have their
	 * content text returned). If the node has no renderable content an empty
	 * string should be returned.
	 */
	public function toPlaintext(MDState $state): string {
		return MDNode::arrayToPlaintext($this->children, $state);
	}

	/**
	 * Protected helper method that renders an HTML fragment of the attributes
	 * to apply to the root HTML tag representation of this node.
	 *
	 * Example result with a couple `cssClasses`, a `cssId`, and a custom
	 * `attributes` key-value pair:
	 *
	 * ```
	 *  class="foo bar" id="baz" lang="en"
	 * ```
	 *
	 * The value includes a leading space if it's non-empty so that it can be
	 * concatenated directly after the tag name and before the closing `>`.
	 */
	protected function htmlAttributes(): string {
		$html = '';
		if (sizeof($this->cssClasses) > 0) {
			$classlist = implode(' ', $this->cssClasses);
			$html .= " class=\"{$classList}\"";
		}
		if ($this->cssId !== null && mb_strlen($this->cssId) > 0) {
			$html .= " id=\"{$this->cssId}\"";
		}
		$styles = [];
		foreach ($this->cssStyles as $key => $value) {
			array_push($styles, "{$key}: {$value};");
		}
		if (sizeof($styles) > 0) {
			$escaped = htmlspecialchars(implode(' ', $styles));
			$html .= " style=\"{$escaped}\"";
		}
		foreach ($this->attributes as $key => $value) {
			if ($key === 'class' || $key === 'id' || $key === 'style') continue;
			$cleanKey = MDUtils::scrubAttributeName($key);
			if (mb_strlen($cleanKey) == 0) continue;
			$cleanValue = htmlspecialchars($value);
			$html .= " {$cleanKey}=\"{$cleanValue}\"";
		}
		return $html;
	}

	/**
	 * Protected helper that renders and concatenates the HTML of all children
	 * of this node. Mostly for use by subclasses in their `toHTML`
	 * implementations.
	 */
	protected function childHTML(MDState $state): string {
		return MDNode::arrayToHTML($this->children, $state);
	}

	/**
	 * Protected helper that renders and concatenates the plaintext of all
	 * children of this node.
	 */
	protected function childPlaintext(MDState $state): string {
		return MDNode::arrayToPlaintext($this->children, $state);
	}

	/**
	 * Protected helper for rendering nodes represented by simple paired HTML
	 * tags. Custom CSS classes and attributes will be included in the result,
	 * and child content will be rendered between the tags.
	 */
	protected function simplePairedTagHTML(MDState $state, string $tagName): string {
		$openTagSuffix = $this->children[0] instanceof MDBlockNode ? "\n" : "";
		$closeTagPrefix = $this->children[sizeof($this->children) - 1] instanceof MDBlockNode ? "\n" : '';
		$closeTagSuffix = $this instanceof MDBlockNode ? "\n" : '';
		$attr = $this->htmlAttributes();
		$childHTML = $this->childHTML($state);
		return "<{$tagName}{$attr}>{$openTagSuffix}{$childHTML}{$closeTagPrefix}</{$tagName}>{$closeTagSuffix}";
	}

	/**
	 * Calls the given callback function with every child node, recursively.
	 * Nodes are visited depth-first.
	 */
	public function visitChildren(callable $fn) {
		foreach ($this->children as $child) {
			$fn($child);
			$child->visitChildren($fn);
		}
	}

	/**
	 * Helper for rendering and concatenating HTML from an array of `MDNode`s.
	 *
	 * @param MDNode[] $nodes
	 * @param MDState $state
	 * @return string HTML string
	 */
	public static function arrayToHTML(array $nodes, MDState $state): string {
		return implode('', array_map(function($node) use ($state) {
			return $node->toHTML($state) . ($node instanceof MDBlockNode ? "\n" : '');
		}, $nodes));
	}

	/**
	 * Helper for rendering and concatenating plaintext from an array of `MDNode`s.
	 *
	 * @param MDNode[] $nodes
	 * @param MDState $state
	 * @return string plaintext
	 */
	public static function arrayToPlaintext(array $nodes, MDState $state): string {
		return implode('', array_map(fn($node) => $node->toPlaintext($state), $nodes));
	}

	/**
	 * Recursively searches and replaces nodes in a tree. The given `replacer`
	 * is passed every node in the tree. If `replacer` returns a new `MDNode`
	 * the original will be replaced with it. If the function returns `null` no
	 * change will be made to that node. Traversal is depth-first.
	 *
	 * @param {MDState} state
	 * @param {MDNode[]} nodes
	 * @param {function} replacer - takes a node as an argument, returns either
	 *   a new node or `null` to leave it unchanged
	 */
	public static function replaceNodes(MDState $state, array &$nodes, callable $replacer) {
		for ($i = 0; $i < sizeof($nodes); $i++) {
			$originalNode = $nodes[$i];
			$replacement = $replacer($originalNode);
			if ($replacement instanceof MDNode) {
				array_splice($nodes, $i, 1, [$replacement]);
			} else {
				self::replaceNodes($state, $originalNode->children, $replacer);
			}
		}
	}
}

/**
 * Marker subclass that indicates a node represents block syntax.
 */
class MDBlockNode extends MDNode {}

/**
 * Paragraph block.
 */
class MDParagraphNode extends MDBlockNode {
	public function toHTML(MDState $state): string {
		return $this->simplePairedTagHTML($state, 'p');
	}
}

/**
 * A heading block with a level from 1 to 6.
 */
class MDHeadingNode extends MDBlockNode {
	public int $level;

	public function __construct(int $level, array $children) {
		parent::__construct($children);
		if (!is_int($level) || ($level < 1 || $level > 6)) {
			$thisClassName = get_class($this);
			throw new Error("{$thisClassName} requires heading level 1 to 6");
		}
		$this->level = $level;
	}

	public function toHTML(MDState $state): string {
		return $this->simplePairedTagHTML($state, "h{$this->level}");
	}
}

/**
 * A sub-text block with smaller, less prominent text.
 */
class MDSubtextNode extends MDBlockNode {
	public function toHTML(MDState $state): string {
		$this->addClass('subtext');
		return $this->simplePairedTagHTML($state, 'div');
	}
}

/**
 * Node for a horizontal dividing line.
 */
class MDHorizontalRuleNode extends MDBlockNode {
	public function toHTML(MDState $state): string {
		return "<hr" . $this->htmlAttributes() . ">";
	}
}

/**
 * A block quote, usually rendered indented from other text.
 */
class MDBlockquoteNode extends MDBlockNode {
	public function toHTML(MDState $state): string {
		return $this->simplePairedTagHTML($state, 'blockquote');
	}
}

/**
 * A bulleted list. Contains `MDListItemNode` children.
 */
class MDUnorderedListNode extends MDBlockNode {
	/** @var MDListItemNode[] $children */

	public function toHTML(MDState $state): string {
		return $this->simplePairedTagHTML($state, 'ul');
	}
}

/**
 * A numbered list. Contains `MDListItemNode` children.
 */
class MDOrderedListNode extends MDBlockNode {
	/** @var MDListItemNode[] $children */

	public int $startOrdinal;

	/**
	 * @param MDListItemNode[] $children
	 * @param ?int $startOrdinal
	 */
	public function __construct(array $children, ?int $startOrdinal=null) {
		parent::__construct($children);
		$this->startOrdinal = $startOrdinal;
	}

	public function toHTML(MDState $state): string {
		if ($this->startOrdinal !== null && $this->startOrdinal != 1) {
			$this->attributes['start'] = strval($this->startOrdinal);
		}
		return $this->simplePairedTagHTML($state, 'ol');
	}
}

/**
 * An item in a bulleted or numbered list.
 */
class MDListItemNode extends MDBlockNode {
	public int $ordinal;

	/**
	 * @param MDNode|MDNode[] $children
	 * @param ?int $ordinal
	 */
	public function __construct(array $children, ?int $ordinal=null) {
		parent::__construct($children);
		$this->ordinal = $ordinal;
	}

	public function toHTML(MDState $state): string {
		return $this->simplePairedTagHTML($state, 'li');
	}
}

/**
 * A block of preformatted computer code. Inner markdown is ignored.
 */
class MDCodeBlockNode extends MDBlockNode {
	public string $text;

	/**
	 * The programming language of the content.
	 */
	public ?string $language;

	public function __construct(string $text, ?string $language=null) {
		super([]);
		$this->text = $text;
		$this->language = $language;
	}

	public function toHTML(MDState $state): string {
		$languageModifier = ($this->language !== null) ? " class=\"language-{$this->language}\"" : '';
		return "<pre" . $this->htmlAttributes() . "><code{$languageModifier}>" .
			htmlentities($this->text) . "</code></pre>\n";
	}
}

/**
 * A table node with a single header row and any number of body rows.
 *
 * If modifying the rows, use the `headerRow` and `bodyRows` accessors,
 * otherwise `children` may get out of sync.
 */
class MDTableNode extends MDBlockNode {
	/** @var MDTableRowNode[] $children */

	public function headerRow(): ?MDTableRowNode { return $this->children[0] ?? null; }

	public function bodyRows(): array { return array_slice($this->children, 1); }

	/**
	 * How to align each column. Columns beyond the length of the array or with
	 * corresponding `null` elements will have no alignment set. Values should
	 * be valid CSS `text-align` values.
	 *
	 * @var string[]
	 */
	public array $columnAlignments = [];

	/**
	 * @param MDTableRowNode $headerRow
	 * @param MDTableRowNode[] $bodyRows
	 */
	public function __construct(MDTableRow $headerRow, array $bodyRows) {
		parent::__construct(array_merge([ $headerRow ], $bodyRows));
	}

	public function applyAlignments() {
		foreach ($this->children as $child) {
			$this->applyAlignmentsToRow($child);
		}
	}

	private function applyAlignmentsToRow(MDTableRowNode $row) {
		foreach ($row->children as $columnIndex => $cell) {
			$alignment = $this->columnAlignments[$columnIndex] ?? null;
			$this->applyAlignmentToCell($cell, $alignment);
		}
	}

	public function applyAlignmentToCell(MDTableCellNode $cell, ?string $alignment) {
		if ($alignment) {
			$cell->cssStyles['text-align'] = $alignment;
		} else {
			unset($cell->cssStyles['text-align']);
		}
	}

	public function toHTML(MDState $state): string {
		$this->applyAlignments();
		$html = '';
		$html .= "<table" . $this->htmlAttributes() . ">\n";
		$html .= '<thead>\n';
		$html .= $this->headerRow->toHTML($state) . "\n";
		$html .= "</thead>\n";
		$html .= "<tbody>\n";
		$html .= MDNode::toHTML($this->bodyRows, $state) . "\n";
		$html .= "</tbody>\n";
		$html .= "</table>\n";
		return $html;
	}
}

/**
 * Node for one row (header or body) in a table.
 */
class MDTableRowNode extends MDBlockNode {
	/** @var MDTableCellNode[] $children */

	public function toHTML(MDState $state): string {
		return $this->simplePairedTagHTML($state, 'tr');
	}
}

/**
 * Node for one cell in a table row.
 */
class MDTableCellNode extends MDBlockNode {
	public function toHTML(MDState $state): string {
		return $this->simplePairedTagHTML($state, 'td');
	}
}

/**
 * Node for a header cell in a header table row.
 */
class MDTableHeaderCellNode extends MDBlockNode {
	public function toHTML(MDState $state): string {
		return $this->simplePairedTagHTML($state, 'th');
	}
}

/**
 * Definition list with `MDDefinitionListTermNode` and
 * `MDDefinitionListDefinitionNode` children.
 */
class MDDefinitionListNode extends MDBlockNode {
	public function toHTML(MDState $state): string {
		return $this->simplePairedTagHTML($state, 'dl');
	}
}

/**
 * A word or term in a definition list.
 */
class MDDefinitionListTermNode extends MDBlockNode {
	public function toHTML(MDState $state): string {
		return $this->simplePairedTagHTML($state, 'dt');
	}
}

/**
 * The definition of a word or term in a definition list. Should follow a
 * definition term, or another definition to serve as an alternate.
 */
class MDDefinitionListDefinitionNode extends MDBlockNode {
	public function toHTML(MDState $state): string {
		return $this->simplePairedTagHTML($state, 'dd');
	}
}

/**
 * Block at the bottom of a document listing all the footnotes with their
 * content.
 */
class MDFootnoteListNode extends MDBlockNode {
	private function footnoteId(MDState $state, string $symbol): int {
		$lookup = $state->root()->userInfo['footnoteIds'];
		if (!$lookup) return null;
		return $lookup[$symbol] ?? null;
	}

	public function toHTML(MDState $state): string {
		$footnotes = $state->userInfo['footnotes'];
		$symbolOrder = array_keys($footnotes);
		if (sizeof($footnotes) == 0) return '';
		$footnoteUniques = $state->root()->footnoteInstances;
		$html = '';
		$html .= '<div class="footnotes">';
		$html .= '<ol>';
		foreach ($symbolOrder as $symbol) {
			$content = $footnotes[$symbol];
			if (!$content) continue;
			$footnoteId = $this->footnoteId($state, $symbol);
			$contentHTML = MDNode::toHTML($content, $state);
			$html .= "<li value=\"{$footnoteId}\" id=\"{$state->root()->elementIdPrefix}footnote_{$footnoteId}\">{$contentHTML}";
			$uniques = $footnoteUniques[$symbol];
			if ($uniques) {
				foreach ($uniques as $unique) {
					$html .= " <a href=\"#{$state->root()->elementIdPrefix}footnoteref_{$unique}\" class=\"footnote-backref\">↩︎</a>";
				}
			}
			$html .= "</li>\n";
		}
		$html .= '</ol>';
		$html .= '</div>';
		return html;
	}

	public function toPlaintext(MDState $state): string {
		$footnotes = $state->userInfo['footnotes'];
		$symbolOrder = array_keys($footnotes);
		if (sizeof($footnotes) == 0) return '';
		$text = '';
		foreach ($symbolOrder as $symbol) {
			$content = $footnotes[$symbol];
			if (!$content) continue;
			$text .= "{$symbol}. " . $this->childPlaintext(state) . "\n";
		}
		return trim($text);
	}
}

/**
 * Marker subclass that indicates a node represents inline syntax.
 */
class MDInlineNode extends MDNode {}

/**
 * Contains plain text. Special HTML characters are escaped when rendered.
 */
class MDTextNode extends MDInlineNode {
	public string $text;

	public function __construct(string $text) {
		parent::__construct([]);
		if (!is_string($text) || mb_strlen($text) == 0) throw new Error("Meh!");
		$this->text = $text;
	}

	public function toHTML(MDState $state): string {
		return htmlentities($this->text);
	}

	public function toPlaintext(MDState $state): string {
		return $this->text;
	}
}

/**
 * Contains plain text which is rendered with HTML entities when rendered to
 * be marginally more difficult for web scapers to decipher. Used for
 * semi-sensitive info like email addresses.
 */
class MDObfuscatedTextNode extends MDTextNode {
	public function toHTML(MDState $state): string {
		return MDUtils::escapeObfuscated($this->text);
	}
}

/**
 * Emphasized (italicized) content.
 */
class MDEmphasisNode extends MDInlineNode {
	public function toHTML(MDState $state): string {
		return $this->simplePairedTagHTML($state, 'em');
	}
}

/**
 * Strong (bold) content.
 */
class MDStrongNode extends MDInlineNode {
	public function toHTML(MDState $state): string {
		return $this->simplePairedTagHTML($state, 'strong');
	}
}

/**
 * Content rendered with a line through it.
 */
class MDStrikethroughNode extends MDInlineNode {
	public function toHTML(MDState $state): string {
		return $this->simplePairedTagHTML($state, 's');
	}
}

/**
 * Underlined content.
 */
class MDUnderlineNode extends MDInlineNode {
	public function toHTML(MDState $state): string {
		return $this->simplePairedTagHTML($state, 'u');
	}
}

/**
 * Highlighted content. Usually rendered with a bright colored background.
 */
class MDHighlightNode extends MDInlineNode {
	public function toHTML(MDState $state): string {
		return $this->simplePairedTagHTML($state, 'mark');
	}
}

/**
 * Superscripted content.
 */
class MDSuperscriptNode extends MDInlineNode {
	public function toHTML(MDState $state): string {
		return $this->simplePairedTagHTML($state, 'sup');
	}
}

/**
 * Subscripted content.
 */
class MDSubscriptNode extends MDInlineNode {
	public function toHTML(MDState $state): string {
		return $this->simplePairedTagHTML($state, 'sub');
	}
}

/**
 * Inline plaintext indicating computer code.
 */
class MDCodeNode extends MDInlineNode {
	public string $text;

	public function __construct(string $text) {
		parent::__construct([]);
		$this->text = $text;
	}

	public function toHTML(MDState $state): string {
		return "<code" . $this->htmlAttributes() . ">" . MDUtils::escapeHTML($this->text) . "</code>";
	}
}

/**
 * A footnote symbol in a document. Denoted as a superscripted number that can
 * be clicked to go to its content at the bottom of the document.
 */
class MDFootnoteNode extends MDInlineNode {
	/**
	 * Symbol the author used to match up the footnote to its content definition.
	 */
	public string $symbol;

	/**
	 * The superscript symbol rendered in HTML. May be the same or different
	 * than `symbol`.
	 */
	public ?string $displaySymbol = null;

	/**
	 * Unique ID for the footnote definition.
	 */
	public ?int $footnoteId = null;

	/**
	 * Unique number for backlinking to a footnote occurrence. Populated by
	 * `MDFootnoteReader.postProcess`.
	 */
	public ?int $occurrenceId = null;

	public function __construct(string $symbol, ?string $title=null) {
		parent::__construct([]);
		$this->symbol = $symbol;
		if ($title) $this->attributes['title'] = $title;
	}

	public function toHTML(MDState $state): string {
		if ($this->differentiator !== null) {
			return "<sup class=\"footnote\" id=\"{$state->root()->elementIdPrefix}footnoteref_{$this->occurrenceId}\"" . $this->htmlAttributes() . ">" .
				"<a href=\"#{$state->root()->elementIdPrefix}footnote_{$this->footnoteId}\">" . htmlentities($this->displaySymbol ?? $this->symbol) . "</a></sup>";
		}
		return "<!--FNREF:{{$this->symbol}}-->";
	}
}

/**
 * A clickable hypertext link.
 */
class MDLinkNode extends MDInlineNode {
	public string $href;

	/**
	 * @param string $href
	 * @param MDNode[]|MDNode $children
	 */
	public function __construct(string $href, array|MDNode $children, ?string $title=null) {
		parent::__construct($children);
		$this->href = $href;
		if ($title !== null) $this->attributes['title'] = $title;
	}

	public function toHTML(MDState $state): string {
		if (str_starts_with($this->href, 'mailto:')) {
			$escapedLink = MDUtils::escapeObfuscated($this->href);
		} else {
			$escapedLink = htmlentities($this->href);
		}
		return "<a href=\"{$escapedLink}\"" . $this->htmlAttributes() . ">" . $this->childHTML($state) . "</a>";
	}
}

/**
 * A clickable hypertext link where the URL is defined elsewhere by reference.
 */
class MDReferencedLinkNode extends MDLinkNode {
	public string $reference;

	public function __construct(string $reference, array|MDNode $children) {
		parent::__construct('', $children);
		$this->reference = $reference;
	}

	public function toHTML(MDState $state): string {
		if ($this->href === '') {
			$url = $state->urlForReference($this->reference);
			if ($url) $this->href = $url;
			$title = $state->urlTitleForReference($this->reference);
			if ($title) $this->attributes['title'] = $title;
		}
		return $super->toHTML($state);
	}
}

/**
 * An inline image.
 */
class MDImageNode extends MDInlineNode {
	public string $src;

	public ?string $alt;

	public function __construct(string $src, ?string $alt) {
		super([]);
		$this->src = $src;
		$this->alt = $alt;
	}

	public function toHTML(MDState $state): string {
		$html = "<img src=\"" . htmlentities($this->src) . "\"";
		if ($this->alt) $html .= " alt=\"" . htmlentities($this->alt) . "\"";
		$html .= $this->htmlAttributes() . ">";
		return $html;
	}
}

/**
 * An inline image where the URL is defined elsewhere by reference.
 */
class MDReferencedImageNode extends MDImageNode {
	public string $reference;

	public function __construct(string $reference, ?string $alt=null) {
		parent::__construct('', $alt, []);
		$this->reference = $reference;
	}

	public function toHTML(MDState $state): string {
		if ($this->src === '') {
			$url = $state->urlForReference($this->reference);
			if ($url !== null) $this->src = $url;
			$title = $state->urlTitleForReference($this->reference);
			if ($title !== null) $this->attributes['title'] = $title;
		}
		return super.toHTML(state);
	}
}

/**
 * An abbreviation that can be hovered over to see its full expansion.
 */
class MDAbbreviationNode extends MDInlineNode {
	/** @type {string} */
	public string $abbreviation;

	/**
	 * @param {string} abbreviation
	 * @param {string} definition
	 */
	public function __construct(string $abbreviation, string $definition) {
		super([]);
		$this->abbreviation = $abbreviation;
		$this->attributes['title'] = $definition;
	}

	public function toHTML(MDState $state): string {
		return "<abbr" . $this->htmlAttributes() . ">" . htmlentities($this->abbreviation) . "</abbr>";
	}
}

/**
 * A line break that is preserved when rendered to HTML.
 */
class MDLineBreakNode extends MDInlineNode {
	public function toHTML(MDState $state): string {
		return '<br>';
	}

	public function toPlaintext(MDState $state): string {
		return "\n";
	}
}

/**
 * A verbatim HTML tag. May be altered to strip out disallowed attributes or
 * CSS values.
 */
class MDHTMLTagNode extends MDInlineNode {
	public MDHTMLTag $tag;

	public function __construct(MDHTMLTag $tag) {
		parent::__construct([]);
		$this->tag = $tag;
	}

	public function toHTML(MDState $state): string {
		return "{$this->tag}";
	}
}


// -- Main class ------------------------------------------------------------


/**
 * Markdown parser.
 */
class Markdown {
	/**
	 * Set of standard readers to handle common syntax.
	 * @type {MDReader[]}
	 */
	public static function standardReaders(): array {
		if (self::$sharedStandardReaders === null) {
			self::$sharedStandardReaders = [
				new MDUnderlinedHeadingReader(),
				new MDHashHeadingReader(),
				new MDBlockQuoteReader(),
				new MDHorizontalRuleReader(),
				new MDUnorderedListReader(),
				new MDOrderedListReader(),
				new MDFencedCodeBlockReader(),
				new MDIndentedCodeBlockReader(),
				new MDParagraphReader(),
		
				new MDStrongReader(),
				new MDEmphasisReader(),
				new MDCodeSpanReader(),
				new MDImageReader(),
				new MDLinkReader(),
				new MDHTMLTagReader(),
			];
		}
		return self::$sharedStandardReaders;
	}
	private static ?array $sharedStandardReaders = null;

	/**
	 * All supported readers except `MDLineBreakReader`.
	 * @type {MDReader[]}
	 */
	public static function allReaders(): array {
		if (self::$sharedAllReaders === null) {
			$sharedAllReaders = array_merge(self::standardReaders(), [
				new MDSubtextReader(),
				new MDTableReader(),
				new MDDefinitionListReader(),
				new MDFootnoteReader(),
				new MDAbbreviationReader(),
		
				new MDUnderlineReader(),
				new MDSubscriptReader(),
				new MDStrikethroughReader(),
				new MDHighlightReader(),
				new MDSuperscriptReader(),
				new MDReferencedImageReader(),
				new MDReferencedLinkReader(),
				new MDModifierReader(),
			]);
		}
		return $sharedAllReaders;
	}
	private static ?array $sharedAllReaders = null;

	/**
	 * Shared instance of a parser with standard syntax.
	 */
	public static function standardParser(): Markdown {
		if (self::$sharedStandardMarkdown === null) {
			self::$sharedStandardMarkdown = new Markdown(self::standardReaders());
		}
		return self::$sharedStandardMarkdown;
	}
	private static ?Markdown $sharedStandardMarkdown = null;

	/**
	 * Shared instance of a parser with all supported syntax.
	 */
	public static function completeParser(): Markdown {
		if (self::$sharedCompleteParser === null) {
			self::$sharedCompleteParser = new Markdown(self::allReaders());
		}
		return self::$sharedCompleteParser;
	}
	public static ?Markdown $sharedCompleteParser = null;

	/**
	 * Filter for what non-markdown HTML is permitted. HTML generated as a
	 * result of markdown is unaffected.
	 */
	public MDHTMLFilter $tagFilter;

	/** @var MDReader[] */
	private array $readers;

	/** @var MDReader[] */
	private array $readersByBlockPriority;
	/** @var MDReader[] */
	private array $readersByTokenPriority;
	private array $readersBySubstitutePriority;

	/**
	 * Creates a Markdown parser with the given syntax readers.
	 *
	 * @param MDReader[] $readers
	 */
	public function __construct(?array $readers=null) {
		$this->readers = $readers ?? self::allReaders();
		$this->readersByBlockPriority = MDReader::sortReaderForBlocks($this->readers);
		$this->readersByTokenPriority = MDReader::sortReadersForTokenizing($this->readers);
		$this->readersBySubstitutePriority = MDReader::sortReadersForSubstitution($this->readers);
		$this->tagFilter = new MDHTMLFilter();
	}

	/**
	 * Converts a markdown string to an HTML string.
	 *
	 * @param string $markdown
	 * @param string $elementIdPrefix  Optional prefix for generated element
	 *   `id`s and links to them. For differentiating multiple markdown docs in
	 *   the same HTML page.
	 * @return string HTML
	 */
	public function toHTML(string $markdown, string $elementIdPrefix='') {
		$lines = mb_split('(?:\\n|\\r|\\r\\n)', $markdown);
		try {
			return $this->parse($lines, $elementIdPrefix);
		} catch (Error $e) {
			$this->investigateException($lines, $elementIdPrefix);
			throw $e;
		}
	}

	/**
	 * @param string[] $lines
	 * @param string $elementIdPrefix
	 */
	private function parse(array $lines, string $elementIdPrefix) {
		$state = new MDState($lines);
		$state->readersByBlockPriority = $this->readersByBlockPriority;
		$state->readersByTokenPriority = $this->readersByTokenPriority;
		$state->readersBySubstitutePriority = $this->readersBySubstitutePriority;
		$state->tagFilter = $this->tagFilter;
		$state->elementIdPrefix = $elementIdPrefix;
		foreach ($this->readers as $reader) {
			$reader->preProcess($state);
		}
		$nodes = $state->readBlocks();
		foreach ($this->readers as $reader) {
			$reader->postProcess($state, $nodes);
		}
		return MDNode::arrayToHTML($nodes, $state);
	}

	/**
	 * Keeps removing first and last lines of markdown to locate the source of
	 * an exception and prints the minimal snippet.
	 *
	 * @param string[] $lines
	 * @param string $elementIdPrefix
	 */
	private function investigateException(array $lines, string $elementIdPrefix) {
		$startIndex = 0;
		$endIndex = sizeof($lines);
		// Keep stripping away first line until an exception stops being thrown
		for ($i = 0; $i < sizeof($lines); $i++) {
			try {
				$this->parse(array_slice($lines, $i, $endIndex), $elementIdPrefix);
				break;
			} catch (Error $e0) {
				$startIndex = $i;
			}
		}
		// Keep stripping away last line until an exception stops being thrown
		for ($i = sizeof($lines); $i > $startIndex; $i--) {
			try {
				$this->parse(array_slice($lines, $startIndex, $i), $elementIdPrefix);
				break;
			} catch (Error $e0) {
				$endIndex = $i;
			}
		}
		$problematicMarkdown = implode("\n", array_slice($lines, $startIndex, $endIndex));
		print("This portion of markdown caused an unexpected exception: {$problematicMarkdown}");
	}
}
?>