@kolkrabbi/kol-workshop 0.29.0 → 0.30.0

package/src/engine/parse-markdown.js

@@ -1,374 +0,0 @@
-/**
- * Handrolled markdown parser for documentation pages.
- * Block-level + inline tokenizer — no remark/markdown-it/marked/MDX.
- * Lifted verbatim from the monorepo, minus the React render layer
- * (`renderInlineTokens` lives with the docs viewer in Phase 2).
- *
- * `parseDocsMarkdown(md)` → { sections, toc, introBlocks, inlineTags }
- */
-
-/** Extract #hashtags from markdown content. */
-export const extractHashtags = (content) => {
-  // Strip markdown link targets (…) to avoid matching anchor fragments as tags
-  const cleaned = content.replace(/\([^)]*\)/g, '')
-  const regex = /#([a-z0-9-]+)/gi
-  const matches = cleaned.match(regex)
-  return matches ? [...new Set(matches.map((t) => t.slice(1).toLowerCase()))] : []
-}
-
-/**
- * Process inline markdown within text content.
- * Supports: bold, italic, inline code, links, images, color swatches, hashtags.
- * Returns a token array; the render layer turns tokens into elements.
- */
-/* EXPORTED 2026-08-01. The props table needs the token producer so its cells
- * can render `backticked` code as a chip instead of printing the backticks.
- * It was module-private, which is why the table had no way to do it and the
- * gap read as "the table is bad" rather than "the parser was unreachable". */
-export const processInlineMarkdown = (text) => {
-  if (!text) return []
-
-  const tokens = []
-  let remaining = text
-
-  while (remaining.length > 0) {
-    // Image: ![alt](src)
-    const imageMatch = remaining.match(/^!\[([^\]]*)\]\(([^)]+)\)/)
-    if (imageMatch) {
-      tokens.push({ type: 'image', alt: imageMatch[1], src: imageMatch[2] })
-      remaining = remaining.slice(imageMatch[0].length)
-      continue
-    }
-
-    // Wikilink: [[target]] / [[target|display]] / [[target#anchor|display]] (Obsidian).
-    // Emitted as a .md link so it routes through the same resolveDocLink cross-link path.
-    const wikiMatch = remaining.match(/^\[\[([^\]|#]+)(?:#([^\]|]+))?(?:\|([^\]]+))?\]\]/)
-    if (wikiMatch) {
-      const target = wikiMatch[1].trim().replace(/\.md$/, '')
-      const anchor = wikiMatch[2]
-        ? wikiMatch[2].trim().toLowerCase().replace(/\s+/g, '-').replace(/[^a-z0-9-]/g, '')
-        : ''
-      const display = wikiMatch[3] ? wikiMatch[3].trim() : (wikiMatch[1].trim().split('/').pop())
-      tokens.push({ type: 'link', text: display, url: `${target}.md${anchor ? `#${anchor}` : ''}` })
-      remaining = remaining.slice(wikiMatch[0].length)
-      continue
-    }
-
-    // Link: [text](url)
-    const linkMatch = remaining.match(/^\[([^\]]+)\]\(([^)]+)\)/)
-    if (linkMatch) {
-      tokens.push({ type: 'link', text: linkMatch[1], url: linkMatch[2] })
-      remaining = remaining.slice(linkMatch[0].length)
-      continue
-    }
-
-    // Color swatch: {#rrggbb}
-    const swatchMatch = remaining.match(/^\{(#[0-9a-fA-F]{6})\}/)
-    if (swatchMatch) {
-      tokens.push({ type: 'colorswatch', color: swatchMatch[1] })
-      remaining = remaining.slice(swatchMatch[0].length)
-      continue
-    }
-
-    // Inline code: `code`
-    const codeMatch = remaining.match(/^`([^`]+)`/)
-    if (codeMatch) {
-      tokens.push({ type: 'code', content: codeMatch[1] })
-      remaining = remaining.slice(codeMatch[0].length)
-      continue
-    }
-
-    // Bold: **text** (must check before italic)
-    const boldMatch = remaining.match(/^\*\*([^*]+)\*\*/)
-    if (boldMatch) {
-      tokens.push({ type: 'bold', content: boldMatch[1] })
-      remaining = remaining.slice(boldMatch[0].length)
-      continue
-    }
-
-    // Italic: *text* (must not match list markers)
-    const italicMatch = remaining.match(/^\*([^*\s][^*]*)\*/)
-    if (italicMatch) {
-      tokens.push({ type: 'italic', content: italicMatch[1] })
-      remaining = remaining.slice(italicMatch[0].length)
-      continue
-    }
-
-    // Text followed by a hashtag (but not inside markdown links)
-    const textBeforeHashtag = remaining.match(/^([^#[]+)(#([a-z0-9-]+))/i)
-    if (textBeforeHashtag && !remaining.startsWith('[')) {
-      if (textBeforeHashtag[1]) {
-        tokens.push({ type: 'text', content: textBeforeHashtag[1] })
-      }
-      tokens.push({ type: 'hashtag', tag: textBeforeHashtag[3].toLowerCase() })
-      remaining = remaining.slice(textBeforeHashtag[1].length + textBeforeHashtag[2].length)
-      continue
-    }
-
-    // Standalone hashtag at start of remaining text
-    const hashtagMatch = remaining.match(/^#([a-z0-9-]+)/i)
-    if (hashtagMatch) {
-      tokens.push({ type: 'hashtag', tag: hashtagMatch[1].toLowerCase() })
-      remaining = remaining.slice(hashtagMatch[0].length)
-      continue
-    }
-
-    // Regular text - take until next special character or end
-    const textMatch = remaining.match(/^([^*`[\]!]+)/)
-    if (textMatch) {
-      tokens.push({ type: 'text', content: textMatch[1] })
-      remaining = remaining.slice(textMatch[0].length)
-      continue
-    }
-
-    // If no match, take one character as text
-    tokens.push({ type: 'text', content: remaining[0] })
-    remaining = remaining.slice(1)
-  }
-
-  return tokens
-}
-
-/**
- * Parse markdown text into structured blocks.
- * Returns { sections, toc, introBlocks, inlineTags }.
- */
-export const parseDocsMarkdown = (markdown) => {
-  if (!markdown) {
-    return { sections: [], toc: [], introBlocks: [], inlineTags: [] }
-  }
-
-  const lines = markdown.split('\n')
-  const sections = []
-  const toc = []
-  const introBlocks = []
-
-  let current = null
-  let inCode = false
-  let inFrontmatter = false
-  let paragraphBuffer = []
-
-  const getTargetBlocks = () => (current ? current.blocks : introBlocks)
-
-  const flushParagraph = () => {
-    if (paragraphBuffer.length > 0) {
-      const text = paragraphBuffer.join(' ').trim()
-      if (text) {
-        getTargetBlocks().push({
-          type: 'paragraph',
-          content: text,
-          tokens: processInlineMarkdown(text),
-        })
-      }
-      paragraphBuffer = []
-    }
-  }
-
-  const startNewSection = (level, heading) => {
-    flushParagraph()
-
-    // H2 starts new section
-    if (level === 2) {
-      if (current) {
-        sections.push(current)
-      }
-      const id = heading.toLowerCase().replace(/\s+/g, '-').replace(/[^a-z0-9-]/g, '')
-      const tocLabel = heading.replace(/^[\d]+[a-z]?[.)]\s*/i, '')
-      toc.push({ id, label: tocLabel, level: 2 })
-      current = { heading, id, level: 2, blocks: [] }
-      return
-    }
-
-    // H3/H4 added as blocks within current section (not in TOC)
-    if (level === 3 || level === 4) {
-      const id = heading.toLowerCase().replace(/\s+/g, '-').replace(/[^a-z0-9-]/g, '')
-      if (current) {
-        current.blocks.push({ type: level === 3 ? 'heading3' : 'heading4', id, content: heading })
-      } else {
-        // H3/H4 before any H2 goes to intro
-        introBlocks.push({ type: level === 3 ? 'heading3' : 'heading4', id, content: heading })
-      }
-    }
-  }
-
-  lines.forEach((line, lineIndex) => {
-    const trimmed = line.trim()
-
-    // Code blocks - check first so content inside code blocks is preserved
-    if (line.startsWith('```')) {
-      if (inCode) {
-        inCode = false
-      } else {
-        flushParagraph()
-        const lang = line.replace(/^```/, '').trim()
-        getTargetBlocks().push({ type: 'code', lang, lines: [] })
-        inCode = true
-      }
-      return
-    }
-
-    // If inside a code block, add line as-is (preserves ---, #, etc.)
-    if (inCode) {
-      const blocks = getTargetBlocks()
-      const lastBlock = blocks[blocks.length - 1]
-      if (lastBlock && lastBlock.type === 'code') {
-        lastBlock.lines.push(line)
-      }
-      return
-    }
-
-    // Handle frontmatter (skip it) - only when not in code block
-    if (trimmed === '---') {
-      if (lineIndex === 0) {
-        inFrontmatter = true
-        return
-      }
-      if (inFrontmatter) {
-        inFrontmatter = false
-        return
-      }
-      // Otherwise it's a divider
-      flushParagraph()
-      getTargetBlocks().push({ type: 'divider' })
-      return
-    }
-
-    if (inFrontmatter) {
-      return
-    }
-
-    // H1 - Title (add to intro, not sections)
-    if (line.startsWith('# ')) {
-      flushParagraph()
-      introBlocks.push({ type: 'heading1', content: line.replace(/^#\s+/, '') })
-      return
-    }
-
-    // H2 - Main sections
-    if (line.startsWith('## ')) {
-      startNewSection(2, line.replace(/^##\s+/, ''))
-      return
-    }
-
-    // H3 - Sub-sections
-    if (line.startsWith('### ')) {
-      startNewSection(3, line.replace(/^###\s+/, ''))
-      return
-    }
-
-    // H4 - Sub-sub-sections
-    if (line.startsWith('#### ')) {
-      startNewSection(4, line.replace(/^####\s+/, ''))
-      return
-    }
-
-    // Empty line
-    if (trimmed === '') {
-      flushParagraph()
-      return
-    }
-
-    // Unordered list
-    const unorderedMatch = line.match(/^[-*]\s+(.*)/)
-    if (unorderedMatch) {
-      flushParagraph()
-      const blocks = getTargetBlocks()
-      const lastBlock = blocks[blocks.length - 1]
-      const itemContent = unorderedMatch[1]
-      if (lastBlock && lastBlock.type === 'list' && !lastBlock.ordered) {
-        lastBlock.items.push({ content: itemContent, tokens: processInlineMarkdown(itemContent) })
-      } else {
-        blocks.push({
-          type: 'list',
-          ordered: false,
-          items: [{ content: itemContent, tokens: processInlineMarkdown(itemContent) }],
-        })
-      }
-      return
-    }
-
-    // Ordered list
-    const orderedMatch = line.match(/^\d+\.\s+(.*)/)
-    if (orderedMatch) {
-      flushParagraph()
-      const blocks = getTargetBlocks()
-      const lastBlock = blocks[blocks.length - 1]
-      const itemContent = orderedMatch[1]
-      if (lastBlock && lastBlock.type === 'list' && lastBlock.ordered) {
-        lastBlock.items.push({ content: itemContent, tokens: processInlineMarkdown(itemContent) })
-      } else {
-        blocks.push({
-          type: 'list',
-          ordered: true,
-          items: [{ content: itemContent, tokens: processInlineMarkdown(itemContent) }],
-        })
-      }
-      return
-    }
-
-    // Blockquote
-    if (line.startsWith('>')) {
-      flushParagraph()
-      const content = line.replace(/^>\s?/, '')
-      getTargetBlocks().push({ type: 'blockquote', content, tokens: processInlineMarkdown(content) })
-      return
-    }
-
-    // Table row (starts with |)
-    if (trimmed.startsWith('|') && trimmed.endsWith('|')) {
-      flushParagraph()
-      const blocks = getTargetBlocks()
-      const lastBlock = blocks[blocks.length - 1]
-
-      /* Split on UNESCAPED pipes only, then unescape: Obsidian tables carry
-       * `[[target\|display]]` wikilinks, and a naive split('|') exploded the
-       * link across two cells and left literal `\` in the text — the reason
-       * hub INDEX pages (tables of wikilinks) rendered as junk (wave-4). */
-      const cells = trimmed
-        .slice(1, -1)
-        .split(/(?<!\\)\|/)
-        .map((cell) => cell.trim().replace(/\\\|/g, '|'))
-
-      // Separator row (|---|---|)
-      const isSeparator = cells.every((cell) => /^:?-+:?$/.test(cell))
-
-      if (isSeparator) {
-        if (lastBlock && lastBlock.type === 'table') {
-          lastBlock.hasSeparator = true
-        }
-        return
-      }
-
-      if (lastBlock && lastBlock.type === 'table') {
-        if (!lastBlock.hasSeparator) {
-          lastBlock.headers = cells
-        } else {
-          lastBlock.rows.push(cells.map((cell) => ({ content: cell, tokens: processInlineMarkdown(cell) })))
-        }
-      } else {
-        blocks.push({ type: 'table', headers: cells, rows: [], hasSeparator: false })
-      }
-      return
-    }
-
-    // Regular text line - add to paragraph buffer
-    paragraphBuffer.push(line)
-  })
-
-  flushParagraph()
-
-  if (current) {
-    sections.push(current)
-  }
-
-  // Extract hashtags from all content
-  const allContent = [
-    ...introBlocks.map((b) => b.content || ''),
-    ...sections.flatMap((s) => [s.heading, ...s.blocks.map((b) => b.content || '')]),
-  ].join(' ')
-  const inlineTags = extractHashtags(allContent)
-
-  return { sections, toc, introBlocks, inlineTags }
-}
-
-export default parseDocsMarkdown

package/src/engine/search.js

@@ -1,36 +0,0 @@
-/**
- * matchSearchItems — the pure substring-match predicate extracted from
- * ShellSearchOverlay. Case-insensitive `includes` across each item's
- * label / tags / headings / keywords, returning the matched items annotated
- * with `matchedHeading` and `matchedKeyword` for the overlay's subtext rows.
- *
- * Match semantics (unchanged from the overlay):
- *   - an item survives if the query hits ANY of label / tags / headings / keywords
- *   - `matchedHeading` = the first heading that matched, else null
- *   - `matchedKeyword` = the first keyword that matched, but ONLY surfaced when
- *      the label and headings did NOT match (a keyword-only reason-for-inclusion)
- *
- * Pure JS, zero dependencies.
- *
- * @param {Array<Object>} items  items shaped { label, tags?, headings?, keywords? }
- * @param {string}        query  raw query string
- * @returns {Array<Object>} filtered items, each spread with matchedHeading/matchedKeyword
- */
-export function matchSearchItems(items, query) {
-  if (!query || query.length === 0) return []
-  const q = query.toLowerCase()
-  return (items || [])
-    .map((item) => {
-      const labelMatch = item.label.toLowerCase().includes(q)
-      const tagMatch = (item.tags || []).some((tag) => tag.toLowerCase().includes(q))
-      const headingMatch = (item.headings || []).find((h) => h.toLowerCase().includes(q))
-      const keywordMatch = (item.keywords || []).find((k) => k.toLowerCase().includes(q))
-      if (!labelMatch && !tagMatch && !headingMatch && !keywordMatch) return null
-      return {
-        ...item,
-        matchedHeading: headingMatch || null,
-        matchedKeyword: !labelMatch && !headingMatch && keywordMatch ? keywordMatch : null,
-      }
-    })
-    .filter(Boolean)
-}

package/src/engine/tags.js

@@ -1,75 +0,0 @@
-/**
- * Pure tag aggregation over a docs inventory — zero dependencies, no React, no d3.
- *
- * Extracted from TagModeOverlay (`buildTagCounts`) and TagGraph
- * (`buildTagCooccurrence`) so the counting/co-occurrence math is testable in
- * isolation and the components stay presentational. Each takes the docs
- * `inventory` (array of `{ id, metadata: { tags: string[] } }`) and returns
- * plain data.
- *
- * Case handling is deliberately NOT unified — it preserves the two call sites'
- * original behaviour:
- *   - buildTagCounts       keeps tags exact-case (the overlay's tag list)
- *   - buildTagCooccurrence lowercases tags (the graph's node ids)
- */
-
-/**
- * Count docs per tag, tags kept exact-case. Returns `[{ tag, count }]` sorted
- * by count descending.
- */
-export const buildTagCounts = (inventory = []) => {
-  const counts = {}
-  inventory.forEach((d) => {
-    if (Array.isArray(d.metadata?.tags)) {
-      d.metadata.tags.forEach((t) => {
-        counts[t] = (counts[t] || 0) + 1
-      })
-    }
-  })
-  return Object.entries(counts)
-    .map(([tag, count]) => ({ tag, count }))
-    .sort((a, b) => b.count - a.count)
-}
-
-/**
- * Build the tag co-occurrence graph. Tag ids are lowercased. Nodes are all
- * tags with their doc counts (sorted desc); edges connect any two tags that
- * share at least one doc, weighted by the number of shared docs.
- *
- * Returns `{ nodes: [{ id, count, type: 'tag' }], edges: [{ source, target, weight }] }`.
- */
-export const buildTagCooccurrence = (inventory = []) => {
-  const tagCounts = {}
-  const tagToDocs = {} // tag -> Set of doc ids
-
-  inventory.forEach((doc) => {
-    const tags = doc.metadata?.tags || []
-    tags.forEach((tag) => {
-      const tagLower = tag.toLowerCase()
-      tagCounts[tagLower] = (tagCounts[tagLower] || 0) + 1
-      if (!tagToDocs[tagLower]) {
-        tagToDocs[tagLower] = new Set()
-      }
-      tagToDocs[tagLower].add(doc.id)
-    })
-  })
-
-  const nodes = Object.entries(tagCounts)
-    .map(([tag, count]) => ({ id: tag, count, type: 'tag' }))
-    .sort((a, b) => b.count - a.count)
-
-  const edges = []
-  const tagList = Object.keys(tagToDocs)
-  for (let i = 0; i < tagList.length; i++) {
-    for (let j = i + 1; j < tagList.length; j++) {
-      const tagA = tagList[i]
-      const tagB = tagList[j]
-      const commonDocs = [...tagToDocs[tagA]].filter((id) => tagToDocs[tagB].has(id))
-      if (commonDocs.length > 0) {
-        edges.push({ source: tagA, target: tagB, weight: commonDocs.length })
-      }
-    }
-  }
-
-  return { nodes, edges }
-}