Function parseMemoryFileContent

source code/utils/claudemd.ts:345–402

* Parses raw memory file content into a MemoryFileInfo. Pure function — no I/O. * * When includeBasePath is given, @include paths are resolved in the same lex * pass and returned alongside the parsed file (so processMemoryFile doesn't * need to lex the same content a second time).

(
  rawContent: string,
  filePath: string,
  type: MemoryType,
  includeBasePath?: string,
)

 * need to lex the same content a second time).
 */
function parseMemoryFileContent(
  rawContent: string,
  filePath: string,
  type: MemoryType,
  includeBasePath?: string,
): { info: MemoryFileInfo | null; includePaths: string[] } {
  // Skip non-text files to prevent loading binary data (images, PDFs, etc.) into memory
  const ext = extname(filePath).toLowerCase()
  if (ext && !TEXT_FILE_EXTENSIONS.has(ext)) {
    logForDebugging(`Skipping non-text file in @include: ${filePath}`)
    return { info: null, includePaths: [] }
  }

  const { content: withoutFrontmatter, paths } =
    parseFrontmatterPaths(rawContent)

  // Lex once so strip and @include-extract share the same tokens. gfm:false
  // is required by extract (so ~/path doesn't tokenize as strikethrough) and
  // doesn't affect strip (html blocks are a CommonMark rule).
  const hasComment = withoutFrontmatter.includes('<!--')
  const tokens =
    hasComment || includeBasePath !== undefined
      ? new Lexer({ gfm: false }).lex(withoutFrontmatter)
      : undefined

  // Only rebuild via tokens when a comment actually needs stripping —
  // marked normalises \r\n during lex, so round-tripping a CRLF file
  // through token.raw would spuriously flip contentDiffersFromDisk.
  const strippedContent =
    hasComment && tokens
      ? stripHtmlCommentsFromTokens(tokens).content
      : withoutFrontmatter

  const includePaths =
    tokens && includeBasePath !== undefined
      ? extractIncludePathsFromTokens(tokens, includeBasePath)
      : []

  // Truncate MEMORY.md entrypoints to the line AND byte caps
  let finalContent = strippedContent
  if (type === 'AutoMem' || type === 'TeamMem') {
    finalContent = truncateEntrypointContent(strippedContent).content
  }

  // Covers frontmatter strip, HTML comment strip, and MEMORY.md truncation
  const contentDiffersFromDisk = finalContent !== rawContent
  return {
    info: {
      path: filePath,
      type,
      content: finalContent,
      globs: paths,
      contentDiffersFromDisk,
      rawContent: contentDiffersFromDisk ? rawContent : undefined,
    },
    includePaths,
  }
}