@limetech/lime-elements 40.20.0 → 40.20.1

package/dist/esm/{markdown-exporter-B2vFxSLw.js → markdown-exporter-DcFcD699.js}

@@ -1,5 +1,5 @@
 import { z as zwitch, w as webNamespaces, y as s, A as h, i as convert, C as whitespace, g as structuredClone, v as visit, E as EXIT, S as SKIP, h as position, u as unified } from './index-CofmbCdQ.js';
-import { a as asciiAlpha, b as asciiAlphanumeric, h as htmlRawNames, c as htmlBlockNames, f as formatCodeAsIndented, d as formatHeadingAsSetext, e as decodeString, i as classifyCharacter, j as asciiPunctuation, k as encodeCharacterReference, p as patternInScope, l as handle$1, n as phrasing$3, t as toString, r as remarkGfm, o as remarkParse } from './markdown-parser-Cw8iW5R9.js';
+import { f as formatCodeAsIndented, h as htmlKind, c as checkRule, a as formatHeadingAsSetext, d as decodeString, p as patternInScope, b as classifyCharacter, e as encodeCharacterReference, i as handle$1, j as phrasing$2, k as containerPhrasing, t as toString, r as remarkGfm, l as remarkParse } from './markdown-parser-8aHA9IYo.js';
 import './_commonjsHelpers-B85MJLTf.js';
 
 /**
@@ -101,67 +101,6 @@ function map(left, right) {
   }
 }
 
-/**
- * @typedef {1 | 2 | 3 | 4 | 5 | 6 | 7} Kind
- *   Kind of HTML (flow) element.
- */
-
-/**
- * Infer the kind of HTML that `value` is.
- *
- * This works on valid trees as produced by `micromark`.
- * Kinds:
- *
- * 1. raw
- * 2. comment
- * 3. instruction
- * 4. declaration
- * 5. cdata
- * 6. block
- * 7. other
- *
- * See: <https://spec.commonmark.org/0.31.2/#html-blocks>.
- *
- * @param {string} value
- *   HTML.
- * @returns {Kind | undefined}
- *   Kind; `undefined` if unknown.
- */
-function htmlKind(value) {
-  if (value.charCodeAt(0) !== 60 /* `<` */) return
-
-  const next = value.charCodeAt(1);
-
-  if (next === 33 /* `!` */) {
-    const code = value.charCodeAt(2);
-    if (code === 45 /* `-` */) return 2
-    if (code === 91 /* `[` */) return 5
-    if (asciiAlpha(code)) return 4
-    return
-  }
-
-  if (next === 63 /* `?` */) return 3
-
-  const closing = next === 47; /* `/` */
-  const start = closing ? 2 : 1;
-  let end = start;
-  let code = value.charCodeAt(end);
-
-  if (!asciiAlpha(code)) return
-
-  while (code === 45 /* `-` */ || asciiAlphanumeric(code)) {
-    code = value.charCodeAt(++end);
-  }
-
-  const name = value.slice(start, end).toLowerCase();
-
-  if (!closing && code !== 47 /* `/` */ && htmlRawNames.includes(name)) {
-    return 1
-  }
-
-  return htmlBlockNames.includes(name) ? 6 : 7
-}
-
 /**
  * @import {Join} from 'mdast-util-to-markdown'
  */
@@ -189,30 +128,74 @@ function joinDefaults(left, right, parent, state) {
   // Join children of a list or an item.
   // In which case, `parent` has a `spread` field.
   if ('spread' in parent && typeof parent.spread === 'boolean') {
+    if (parent.spread) return 1
+
+    // HTML that only ends at a blank line.
+    if (left.type === 'html') {
+      const kind = htmlKind(left.value);
+      if (kind === undefined || kind === 6 || kind === 7) return 1
+    }
+
+    if (left.type === 'blockquote' && right.type === 'blockquote') return 1
+
+    // Lists with non-1 start would turn into a paragraph.
+    if (
+      (left.type === 'definition' || left.type === 'paragraph') &&
+      right.type === 'list' &&
+      ((right.ordered &&
+        typeof right.start === 'number' &&
+        right.start !== 1) ||
+        (right.children.length > 0 && right.children[0].children.length === 0))
+    ) {
+      return 1
+    }
+
+    // Thematic break with dashes would turn into a setext heading.
     if (
       left.type === 'paragraph' &&
-      // Two paragraphs.
-      (left.type === right.type ||
-        right.type === 'definition' ||
-        // Paragraph followed by a setext heading.
-        (right.type === 'heading' && formatHeadingAsSetext(right, state)))
+      right.type === 'thematicBreak' &&
+      checkRule(state) === '-' &&
+      !state.options.ruleSpaces
     ) {
-      return
+      return 1
     }
 
-    // HTML that only ends at a blank line followed by something.
-    if (left.type === 'html') {
-      const kind = htmlKind(left.value);
-      if (kind === undefined || kind === 6 || kind === 7) return 1
+    // We have to “delve” into containers that form from the left margin.
+    let tail = left;
+
+    while (
+      (tail.type === 'blockquote' ||
+        tail.type === 'list' ||
+        tail.type === 'listItem') &&
+      tail.children.length > 0
+    ) {
+      tail = tail.children[tail.children.length - 1];
     }
 
-    // HTML that cannot interrupt a paragraph (kind 7) after one.
-    if (left.type === 'paragraph' && right.type === 'html') {
+    if (tail.type !== 'definition' && tail.type !== 'paragraph') return 0
+
+    // Indented code would turn into a paragraph.
+    if (right.type === 'code' && formatCodeAsIndented(right, state)) return 1
+
+    // Kind 7 HTML would turn into a paragraph.
+    if (right.type === 'html') {
       const kind = htmlKind(right.value);
       if (kind === undefined || kind === 7) return 1
     }
 
-    return parent.spread ? 1 : 0
+    // Text joins the paragraph before it, also one at the end of a container.
+    // Not a definition before it: definitions are parsed out of the start of
+    // a paragraph, so they and the text after them are read back the same.
+    if (
+      (left.type === 'paragraph' || left !== tail) &&
+      (right.type === 'definition' ||
+        (right.type === 'heading' && formatHeadingAsSetext(right, state)) ||
+        right.type === 'paragraph')
+    ) {
+      return 1
+    }
+
+    return 0
   }
 }
 
@@ -292,7 +275,22 @@ const unsafe = [
   {character: '#', inConstruct: 'headingAtx', after: '(?:[\r\n]|$)'},
   // Dollar sign and percentage are not used in markdown.
   // An ampersand could start a character reference.
-  {character: '&', after: '[#A-Za-z]', inConstruct: 'phrasing'},
+  // Labels match on their source.
+  {
+    character: '&',
+    after: '[#A-Za-z]',
+    inConstruct: [
+      'codeFencedLangGraveAccent',
+      'codeFencedLangTilde',
+      'codeFencedMetaGraveAccent',
+      'codeFencedMetaTilde',
+      'destinationLiteral',
+      'destinationRaw',
+      'phrasing',
+      'titleApostrophe',
+      'titleQuote'
+    ]
+  },
   // An apostrophe can break out of a title.
   {character: "'", inConstruct: 'titleApostrophe'},
   // A left paren could break out of a destination raw.
@@ -441,820 +439,6 @@ function compilePattern(pattern) {
   return pattern._compiled
 }
 
-/**
- * Get the first character of a value.
- *
- * Is aware of astral characters (such as emoji).
- *
- * @param {string} value
- *   Value.
- * @returns {string}
- *   First character; empty if `value` is empty.
- */
-function firstCharacter(value) {
-  const code = value.codePointAt(0);
-  return code === undefined ? '' : String.fromCodePoint(code)
-}
-
-/**
- * Get the last character of a value.
- *
- * Is aware of astral characters (such as emoji).
- *
- * @param {string} value
- *   Value.
- * @returns {string}
- *   Last character; empty if `value` is empty.
- */
-function lastCharacter(value) {
-  // Only a surrogate pair gives a code point above the BMP.
-  const code = value.codePointAt(value.length - 2);
-  return code !== undefined && code > 0xff_ff
-    ? value.slice(-2)
-    : value.slice(-1)
-}
-
-/**
- * @import {EncodeSides} from '../types.js'
- */
-
-
-/**
- * Check whether to encode (as a character reference) the characters
- * surrounding an attention run.
- *
- * Which characters are around an attention run influence whether it works or
- * not.
- *
- * See <https://github.com/orgs/syntax-tree/discussions/60> for more info.
- * See this markdown in a particular renderer to see what works:
- *
- * ```markdown
- * |                         | A (letter inside) | B (punctuation inside) | C (whitespace inside) | D (nothing inside) |
- * | ----------------------- | ----------------- | ---------------------- | --------------------- | ------------------ |
- * | 1 (letter outside)      | x*y*z             | x*.*z                  | x* *z                 | x**z               |
- * | 2 (punctuation outside) | .*y*.             | .*.*.                  | .* *.                 | .**.               |
- * | 3 (whitespace outside)  | x *y* z           | x *.* z                | x * * z               | x ** z             |
- * | 4 (nothing outside)     | *x*               | *.*                    | * *                   | **                 |
- * ```
- *
- * @param {number} outside
- *   Code point on the outer side of the run.
- * @param {number} inside
- *   Code point on the inner side of the run.
- * @param {string} marker
- *   Marker of the run.
- *   Underscores are handled more strictly (they form less often) than
- *   asterisks.
- * @returns {EncodeSides}
- *   Whether to encode characters.
- */
-// Important: punctuation must never be encoded.
-// Punctuation is solely used by markdown constructs.
-// And by encoding itself.
-// Encoding them will break constructs or double encode things.
-function encodeInfo(outside, inside, marker) {
-  const outsideKind = classifyCharacter(outside);
-  const insideKind = classifyCharacter(inside);
-
-  // Letter outside:
-  if (outsideKind === undefined) {
-    return insideKind === undefined
-      ? // Letter inside:
-        // we have to encode *both* letters for `_` as it is looser.
-        // it already forms for `*` (and GFMs `~`).
-        marker === '_'
-        ? {inside: true, outside: true}
-        : {inside: false, outside: false}
-      : insideKind === 1
-        ? // Whitespace inside: encode both (letter, whitespace).
-          {inside: true, outside: true}
-        : // Punctuation inside: encode outer (letter)
-          {inside: false, outside: true}
-  }
-
-  // Whitespace outside:
-  if (outsideKind === 1) {
-    return insideKind === undefined
-      ? // Letter inside: already forms.
-        {inside: false, outside: false}
-      : insideKind === 1
-        ? // Whitespace inside: encode both (whitespace).
-          {inside: true, outside: true}
-        : // Punctuation inside: already forms.
-          {inside: false, outside: false}
-  }
-
-  // Punctuation outside:
-  return insideKind === undefined
-    ? // Letter inside: already forms.
-      {inside: false, outside: false}
-    : insideKind === 1
-      ? // Whitespace inside: encode inner (whitespace).
-        {inside: true, outside: false}
-      : // Punctuation inside: already forms.
-        {inside: false, outside: false}
-}
-
-/**
- * @import {Attention, Handle, Info, State} from 'mdast-util-to-markdown'
- * @import {PhrasingParents} from '../types.js'
- */
-
-
-/**
- * Markers for builtin attention (emphasis, strong) to whether they are stricter:
- * whether they cannot open and close inside words.
- *
- * Builtin runs pair following CommonMark rules for emphasis and strong:
- * they are split,
- * using one or two markers at a time,
- * with the rule of 3.
- * Other markers pair like GFM strikethrough:
- * whole runs pair with a run of the same size.
- *
- * @type {Map<string, boolean>}
- */
-const builtins = new Map([
-  ['*', false],
-  ['_', true]
-]);
-
-/**
- * @typedef AttentionResult
- *   Attention, of which the sequence is chosen later.
- * @property {Array<Item>} children
- *   Serialized children.
- * @property {Array<string>} sequences
- *   Sequences that can be used, in order of preference.
- */
-
-/**
- * @typedef Improvement
- *   Improvement of attention sequences.
- * @property {Map<AttentionResult, string>} chosen
- *   Chosen sequences for each attention.
- * @property {Array<Token>} tokens
- *   Tokens involved in the improvement.
- * @property {Mistake | undefined} mistake
- *   Mistake that led to this improvement, if any.
- */
-
-/**
- * @typedef {AttentionResult | string} Item
- *   Serialized phrasing.
- */
-
-/**
- * @typedef Mistake
- *   Sequence that does not form its attention.
- * @property {Array<AttentionResult>} attention
- *   Attention to try other sequences for.
- * @property {number} index
- *   Index of the token.
- */
-
-/**
- * @typedef Run
- *   Sequences next to each other with the same marker.
- * @property {boolean} close
- *   Whether it can close.
- * @property {number} end
- *   Index after the last marker not used (to open).
- * @property {Array<Token>} markers
- *   Sequence of each marker.
- * @property {boolean} open
- *   Whether it can open.
- * @property {boolean} split
- *   Whether it can be split (built-in attention).
- * @property {number} start
- *   Index of the first marker not used (to close).
- * @property {Array<Token>} tokens
- *   Sequences.
- */
-
-/**
- * @typedef Token
- *   Rendered phrasing.
- * @property {AttentionResult | undefined} attention
- *   Attention, if this is one of its sequences.
- * @property {string} value
- *   Value.
- */
-
-/**
- * Get candidates for a mistake:
- * attention to try other sequences for,
- * later first.
- *
- * @param {Array<Token>} tokens
- *   Tokens.
- * @param {Array<Token>} involved
- *   Sequences involved in the mistake.
- * @param {Map<Token, Run>} runOf
- *   Run of each sequence.
- * @returns {Array<AttentionResult>}
- *   Candidates.
- */
-function candidates(tokens, involved, runOf) {
-  /** @type {Set<AttentionResult | undefined>} */
-  const related = new Set();
-
-  // The attention involved, and attention in runs with its sequences.
-  for (const token of tokens) {
-    const siblings = runOf.get(token);
-    if (
-      siblings &&
-      token.attention &&
-      involved.some((d) => d.attention === token.attention)
-    ) {
-      for (const sibling of siblings.tokens) {
-        related.add(sibling.attention);
-      }
-    }
-  }
-
-  /** @type {Array<AttentionResult>} */
-  const ordered = [];
-
-  for (const token of tokens) {
-    if (token.attention && related.delete(token.attention)) {
-      ordered.push(token.attention);
-    }
-  }
-
-  // Later first: attention that opens later is inner or next.
-  ordered.reverse();
-  return ordered
-}
-
-/**
- * Check whether sequences form their attention, like `micromark`.
- *
- * Sequences of `*` and `_` pair like emphasis and strong in CommonMark,
- * others like strikethrough in GFM.
- *
- * @param {Array<Token>} tokens
- *   Tokens.
- * @param {string} before
- *   Characters before.
- * @param {string} after
- *   Characters after.
- * @returns {Mistake | undefined}
- *   First mistake, if any.
- */
-function check(tokens, before, after) {
-  /** @type {Array<Run>} */
-  const runs = [];
-  /** @type {Map<Token, Array<Token>>} */
-  const wrong = new Map();
-  /** @type {Map<Token, Run>} */
-  const runOf = new Map();
-  let index = 0;
-
-  while (index < tokens.length) {
-    if (!tokens[index].attention) {
-      index++;
-      continue
-    }
-
-    const marker = tokens[index].value.charAt(0);
-    const strict = builtins.get(marker);
-    let end = index + 1;
-
-    while (
-      end < tokens.length &&
-      tokens[end].attention &&
-      tokens[end].value.charAt(0) === marker
-    ) {
-      end++;
-    }
-
-    const head = index ? tokens[index - 1].value : before;
-    const previous = classifyCharacter(head.charCodeAt(head.length - 1));
-    const next = classifyCharacter(
-      (end < tokens.length ? tokens[end].value : after).charCodeAt(0)
-    );
-    const open = !next || (next === 2 && Boolean(previous));
-    const close = !previous || (previous === 2 && Boolean(next));
-    /** @type {Run} */
-    const run = {
-      tokens: tokens.slice(index, end),
-      markers: [],
-      start: 0,
-      end: 0,
-      // Underscores are strict, they cannot open or close inside words.
-      open: strict ? open && (Boolean(previous) || !close) : open,
-      close: strict ? close && (Boolean(next) || !open) : close,
-      // All builtins can split runs.
-      split: strict !== undefined
-    };
-
-    for (const token of run.tokens) {
-      // Every sequence is wrong until it pairs with its other sequence.
-      wrong.set(token, [...run.tokens]);
-      runOf.set(token, run);
-      let size = token.value.length;
-      while (size--) run.markers.push(token);
-    }
-
-    run.end = run.markers.length;
-    runs.push(run);
-    index = end;
-  }
-
-  pair(runs, wrong);
-
-  index = -1;
-
-  while (++index < tokens.length) {
-    const involved = wrong.get(tokens[index]);
-    if (involved) return {attention: candidates(tokens, involved, runOf), index}
-  }
-}
-
-/**
- * Serialize the children of a parent that contains phrasing children.
- *
- * @param {PhrasingParents} parent
- *   Parent of phrasing nodes.
- * @param {State} state
- *   Info passed around about the current state.
- * @param {Info} info
- *   Info on where we are in the document we are generating.
- * @returns {string}
- *   Serialized result.
- */
-function containerPhrasing(parent, state, info) {
-  return serialize(phrasing$2(parent, state, info), info.before, info.after)
-}
-
-/**
- * Encode the first or last character of tokens as a character reference,
- * if it is text.
- *
- * @param {Array<Token>} tokens
- *   Tokens.
- * @param {boolean} start
- *   Whether to encode the first or last character.
- * @returns {undefined}
- *   Nothing.
- */
-function encode(tokens, start) {
-  const token = tokens[start ? 0 : tokens.length - 1];
-
-  if (!token || token.attention) return
-
-  const character = start
-    ? firstCharacter(token.value)
-    : lastCharacter(token.value);
-  const reference = encodeCharacterReference(
-    /** @type {number} */ (character.codePointAt(0))
-  );
-
-  token.value = start
-    ? reference + token.value.slice(character.length)
-    : token.value.slice(0, token.value.length - character.length) + reference;
-}
-
-/**
- * Try other sequences for the candidates of a mistake and use the first that
- * moves the mistake further.
- *
- * @param {Array<Item>} items
- *   Items.
- * @param {Map<AttentionResult, string>} chosen
- *   Sequences to use instead of preferred sequences.
- * @param {Mistake} mistake
- *   Mistake.
- * @param {string} before
- *   Characters before.
- * @param {string} after
- *   Characters after.
- * @returns {Improvement | undefined}
- *   Result if better.
- */
-function improve(items, chosen, mistake, before, after) {
-  for (const attention of mistake.attention) {
-    const current = chosen.get(attention) || attention.sequences[0];
-
-    for (const sequence of attention.sequences) {
-      if (sequence === current) continue
-      const trial = new Map(chosen);
-      trial.set(attention, sequence);
-      const tokens = render(items, trial, before, after);
-      const next = check(tokens, before, after);
-
-      if (!next || next.index > mistake.index) {
-        return {chosen: trial, mistake: next, tokens}
-      }
-    }
-  }
-}
-
-/**
- * Pair runs like `micromark` and mark sequences that do form their attention
- * as fine.
- *
- * Each pair uses markers from the end of an opener and the start of a closer.
- * It forms the intended attention if those markers are exactly the whole
- * opening and closing sequence of one attention.
- * Otherwise attention is mixed up:
- * either sequences of different attention pair,
- * part of a sequence is used (such as `**` of strong as `*`),
- * or several sequences are used together (such as `*` and `*` as `**`).
- * Then the sequences involved are recorded with each other.
- *
- * @param {Array<Run>} runs
- *   Runs.
- * @param {Map<Token, Array<Token>>} wrong
- *   Sequences that are wrong, with the sequences involved.
- * @returns {undefined}
- *   Nothing, `wrong` is changed.
- */
-function pair(runs, wrong) {
-  let index = 0;
-
-  while (index < runs.length) {
-    const closer = runs[index];
-    const marker = closer.tokens[0].value.charAt(0);
-    const closerSize = closer.end - closer.start;
-    let open = closer.close && closerSize ? index : 0;
-
-    while (open--) {
-      const opener = runs[open];
-      const openerSize = opener.end - opener.start;
-
-      if (
-        opener.tokens[0].value.charAt(0) === marker &&
-        opener.open &&
-        openerSize &&
-        (closer.split
-          ? // The rule of 3, on the sizes that are left.
-            !((opener.close || closer.open) && closerSize % 3) ||
-            (openerSize + closerSize) % 3
-          : // Others pair whole runs of the same size.
-            openerSize === opener.markers.length && openerSize === closerSize)
-      ) {
-        const size = closer.split
-          ? openerSize > 1 && closerSize > 1
-            ? 2
-            : 1
-          : closerSize;
-        const opening = opener.markers.slice(opener.end - size, opener.end);
-        const closing = closer.markers.slice(closer.start, closer.start + size);
-        const openToken = opening[0];
-        const closeToken = closing[0];
-
-        if (
-          // Whole sequences:
-          // the markers of a sequence are next to each other,
-          // so if the first and last marker are of one sequence of that size,
-          // all markers are exactly that sequence.
-          openToken === opening[size - 1] &&
-          openToken.value.length === size &&
-          closeToken === closing[size - 1] &&
-          closeToken.value.length === size &&
-          // Of the same attention, so its opening and closing sequence.
-          openToken.attention === closeToken.attention
-        ) {
-          wrong.delete(openToken);
-          wrong.delete(closeToken);
-        } else {
-          const involved = [...new Set([...opening, ...closing])];
-          // Add to what is involved already: a sequence can be in several pairs.
-          for (const token of involved) {
-            const list = wrong.get(token);
-            if (list) list.push(...involved);
-          }
-        }
-
-        opener.end -= size;
-        closer.start += size;
-        break
-      }
-    }
-
-    // Stay on this run if it can close more.
-    if (open === -1 || closer.start === closer.end) index++;
-  }
-}
-
-/**
- * Get the sequences that can be used for attention, in order of preference.
- *
- * @param {string} type
- *   Node type.
- * @param {Array<string>} markers
- *   Markers.
- * @param {Array<number>} sizes
- *   Sizes.
- * @returns {Array<string>}
- *   Sequences.
- */
-function attentionSequences(type, markers, sizes) {
-  if (markers.length === 0) {
-    throw new Error(
-      'Cannot serialize `' +
-        type +
-        '` as attention without markers, expected one or more markers'
-    )
-  }
-
-  if (sizes.length === 0) {
-    throw new Error(
-      'Cannot serialize `' +
-        type +
-        '` as attention without sizes, expected one or more sizes'
-    )
-  }
-
-  for (const size of sizes) {
-    if (!Number.isSafeInteger(size) || size < 1) {
-      throw new Error(
-        'Cannot serialize `' +
-          type +
-          '` as attention with `' +
-          size +
-          '` as size, expected positive integer'
-      )
-    }
-  }
-
-  /** @type {Array<string>} */
-  const sequences = [];
-
-  for (const marker of markers) {
-    if (marker.length !== 1) {
-      throw new Error(
-        'Cannot serialize `' +
-          type +
-          '` as attention with `' +
-          marker +
-          '` as marker, expected a single ascii character'
-      )
-    }
-
-    if (!asciiPunctuation(marker.charCodeAt(0))) {
-      throw new Error(
-        'Cannot serialize `' +
-          type +
-          '` as attention with `' +
-          marker +
-          '` as marker, expected ascii punctuation'
-      )
-    }
-
-    for (const size of sizes) {
-      sequences.push(marker.repeat(size));
-    }
-  }
-
-  return sequences
-}
-
-/**
- * Serialize the children of a parent that contains phrasing children,
- * keeping attention separate,
- * so that its markers can be chosen later.
- *
- * @param {PhrasingParents} parent
- *   Parent of phrasing nodes.
- * @param {State} state
- *   Info passed around about the current state.
- * @param {Info} info
- *   Info on where we are in the document we are generating.
- * @returns {Array<Item>}
- *   Items.
- */
-function phrasing$2(parent, state, info) {
-  const indexStack = state.indexStack;
-  const children = parent.children || [];
-  /** @type {Array<Item>} */
-  const results = [];
-  let index = -1;
-  let before = info.before;
-
-  indexStack.push(-1);
-  let tracker = state.createTracker(info);
-
-  while (++index < children.length) {
-    const child = children[index];
-    /** @type {string} */
-    let after;
-
-    indexStack[indexStack.length - 1] = index;
-
-    if (index + 1 < children.length) {
-      /** @type {Handle} */
-      // @ts-expect-error: hush, it’s actually a `zwitch`.
-      let handle = state.handle.handlers[children[index + 1].type];
-      // @ts-expect-error: hush, it’s actually a `zwitch`.
-      if (handle && handle.peek) handle = handle.peek;
-      after = handle
-        ? firstCharacter(
-            handle(children[index + 1], parent, state, {
-              before: '',
-              after: '',
-              ...tracker.current()
-            })
-          )
-        : '';
-    } else {
-      after = info.after;
-    }
-
-    // In some cases, html (text) can be found in phrasing right after an eol.
-    // When we’d serialize that, it could be seen as html (flow): all its kinds
-    // except 7 can interrupt a paragraph.
-    // As we can’t escape or so to prevent it from happening,
-    // we take a somewhat reasonable approach:
-    // replace that eol with a space.
-    // See: <https://github.com/syntax-tree/mdast-util-to-markdown/issues/15>
-    const previous = results[results.length - 1];
-
-    if (
-      typeof previous === 'string' &&
-      (before === '\r' || before === '\n') &&
-      child.type === 'html' &&
-      htmlKind(child.value) !== 7
-    ) {
-      results[results.length - 1] = previous.replace(/(\r?\n|\r)$/, ' ');
-      before = ' ';
-
-      // To do: does this work to reset tracker?
-      tracker = state.createTracker(info);
-      tracker.move(serialize(results, info.before, info.after));
-    }
-
-    /** @type {(Handle & {attention?: Attention | undefined}) | undefined} */
-    // @ts-expect-error: hush, it’s actually a `zwitch`.
-    const handle = state.handle.handlers[child.type];
-
-    if (handle && handle.attention) {
-      const {construct, markers, sizes} = handle.attention(child, state);
-      const sequences = attentionSequences(child.type, markers, sizes);
-      const sequence = sequences[0];
-      const exit = state.enter(construct);
-      tracker.move(sequence);
-      const inside = phrasing$2(/** @type {PhrasingParents} */ (child), state, {
-        ...tracker.current(),
-        before: sequence,
-        after: sequence
-      });
-      tracker.move(serialize(inside, sequence, sequence));
-      tracker.move(sequence);
-      exit();
-      results.push({children: inside, sequences});
-      before = sequence.charAt(sequence.length - 1);
-    } else {
-      const value = state.handle(child, parent, state, {
-        ...tracker.current(),
-        after,
-        before
-      });
-
-      if (!value) continue
-      tracker.move(value);
-      results.push(value);
-      before = lastCharacter(value);
-    }
-  }
-
-  indexStack.pop();
-
-  return results
-}
-
-/**
- * Render items with sequences;
- * encoding characters around attention where needed.
- *
- * @param {Array<Item>} items
- *   Serialized phrasing.
- * @param {Map<AttentionResult, string>} chosen
- *   Sequences to use instead of preferred sequences.
- * @param {string} before
- *   Characters before.
- * @param {string} after
- *   Characters after.
- * @returns {Array<Token>}
- *   Tokens.
- */
-function render(items, chosen, before, after) {
-  /** @type {Array<Token>} */
-  const tokens = [];
-  /** @type {string | undefined} */
-  let encodeAfter;
-  let previous = lastCharacter(before);
-  let index = -1;
-
-  while (++index < items.length) {
-    const item = items[index];
-
-    if (typeof item === 'string') {
-      /** @type {Token} */
-      const token = {value: item, attention: undefined};
-
-      // If attention before had to encode the character after it, encode it.
-      if (encodeAfter && encodeAfter === firstCharacter(item)) {
-        encode([token], true);
-      }
-
-      encodeAfter = undefined;
-      tokens.push(token);
-      if (token.value) previous = lastCharacter(token.value);
-      continue
-    }
-
-    const sequence = chosen.get(item) || item.sequences[0];
-    const marker = sequence.charAt(0);
-    const next = items[index + 1];
-    const outsideBefore = previous;
-    const outsideAfter = firstCharacter(
-      next === undefined
-        ? after
-        : typeof next === 'string'
-          ? next
-          : chosen.get(next) || next.sequences[0]
-    );
-    const inside = render(item.children, chosen, sequence, sequence);
-    const head = inside.length > 0 ? inside[0].value : '';
-    const open = encodeInfo(
-      outsideBefore.charCodeAt(outsideBefore.length - 1),
-      head.charCodeAt(0),
-      marker
-    );
-
-    if (open.inside) encode(inside, true);
-
-    const tail =
-      inside.length > 0 ? lastCharacter(inside[inside.length - 1].value) : '';
-    const close = encodeInfo(
-      outsideAfter.charCodeAt(0),
-      tail.charCodeAt(tail.length - 1),
-      marker
-    );
-
-    if (close.inside) encode(inside, false);
-
-    if (open.outside && outsideBefore !== '\n' && outsideBefore !== '\r') {
-      encode(tokens, false);
-    }
-
-    if (close.outside) encodeAfter = outsideAfter;
-
-    tokens.push({value: sequence, attention: item}, ...inside, {
-      value: sequence,
-      attention: item
-    });
-    previous = sequence.charAt(sequence.length - 1);
-  }
-
-  return tokens
-}
-
-/**
- * Serialize phrasing with attention.
- *
- * Whether attention forms depends on everything around it.
- * This checks that it does and tries other sequences if not.
- *
- * @param {Array<Item>} items
- *   Phrasing items.
- * @param {string} before
- *   Characters before.
- * @param {string} after
- *   Characters after.
- * @returns {string}
- *   Serialized markdown.
- */
-function serialize(items, before, after) {
-  /** @type {Map<AttentionResult, string>} */
-  let chosen = new Map();
-  const initial = render(items, chosen, before, after);
-  let tokens = initial;
-  let mistake = check(tokens, before, after);
-
-  while (mistake) {
-    const next = improve(items, chosen, mistake, before, after);
-
-    if (!next) {
-      tokens = initial;
-      break
-    }
-
-    chosen = next.chosen;
-    tokens = next.tokens;
-    mistake = next.mistake;
-  }
-
-  // Join the tokens.
-  let result = '';
-  for (const token of tokens) result += token.value;
-  return result
-}
-
 /**
  * @import {State} from 'mdast-util-to-markdown'
  * @import {FlowChildren, FlowParents, TrackFields} from '../types.js'
@@ -1535,7 +719,8 @@ function safe(state, input, config) {
         infos[position].after &&
         !infos[position + 1].before &&
         !infos[position + 1].after) ||
-      (positions[index - 1] === position - 1 &&
+      (position - 1 >= offset &&
+        positions[index - 1] === position - 1 &&
         infos[position].before &&
         !infos[position - 1].before &&
         !infos[position - 1].after)
@@ -1551,7 +736,7 @@ function safe(state, input, config) {
       result.push(percentEncode ? slice : escapeBackslashes(slice, '\\'));
     }
 
-    start = position;
+    start = position + 1;
 
     // Percent-encoding such as `%20`.
     if (percentEncode) {
@@ -1559,17 +744,15 @@ function safe(state, input, config) {
         '%' +
           value.charCodeAt(position).toString(16).toUpperCase().padStart(2, '0')
       );
-      start++;
     } else if (
       /[!-/:-@[-`{-~]/.test(value.charAt(position)) &&
       (!config.encode || !config.encode.includes(value.charAt(position)))
     ) {
       // Character escape.
-      result.push('\\');
+      result.push('\\' + value.charAt(position));
     } else {
       // Character reference.
       result.push(encodeCharacterReference(value.charCodeAt(position)));
-      start++;
     }
   }
 
@@ -1763,7 +946,7 @@ function toMarkdown(tree, options) {
   });
 
   let result = state.handle(
-    phrasing$3(tree) ? {type: 'root', children: [tree]} : tree,
+    phrasing$2(tree) ? {type: 'root', children: [tree]} : tree,
     undefined,
     state,
     {before: '\n', after: '\n', now: {line: 1, column: 1}, lineShift: 0}
@@ -4821,7 +4004,7 @@ function phrasing(node) {
   const tagName = node.data && node.data.hName;
   return tagName
     ? phrasing$1({type: 'element', tagName, properties: {}, children: []})
-    : phrasing$3(node)
+    : phrasing$2(node)
 }
 
 /**