import { posix } from "path";

/**
 * Regex to match a fenced code block with a `file="..."` attribute, e.g.:
 *
 *   ```csv file="card-data.csv"
 *   name, cost, effect
 *   Fireball, 1, Blast em
 *   ```
 *
 *   ```yaml file="monsters.yaml"
 *   goblin:
 *     hp: 7
 *   ```
 *
 * The language identifier is optional — only `file="..."` matters.
 *
 * Group 1: the info string after the opening fence (e.g. `csv file="card-data.csv"`)
 * Group 2: the code block body
 */
const INLINE_FILE_BLOCK_RE = /^```\w*\s+file="([^"]+)"(?:\s+\w+="[^"]*")*\s*\n([\s\S]*?)```\s*$/gm;

/**
 * Process fenced code blocks that contain a `file="..."` attribute.
 *
 * For each block like:
 *   ```lang file="card-data.csv"
 *   content
 *   ```
 *
 * 1. Strips the block from the returned text (so it doesn't render)
 * 2. Injects a synthetic entry into `index` at the resolved file path
 *
 * @param content - Raw file content
 * @param fileRelativePath - Normalized path of the containing file (e.g. "/rules/combat.md")
 * @param index - The content index to inject synthetic entries into
 * @returns Content with inline blocks removed
 */
export function extractInlineBlocks(
  content: string,
  fileRelativePath: string,
  index: Record<string, string>,
): string {
  const matches: { filename: string; body: string }[] = [];
  let match: RegExpExecArray | null;

  INLINE_FILE_BLOCK_RE.lastIndex = 0;

  while ((match = INLINE_FILE_BLOCK_RE.exec(content)) !== null) {
    matches.push({ filename: match[1], body: match[2] });
  }

  if (matches.length === 0) return content;

  const stripped = content.replace(INLINE_FILE_BLOCK_RE, "");

  for (const { filename, body } of matches) {
    const fileDir = posix.dirname(fileRelativePath);
    const resolvedPath = posix.join(fileDir, filename);

    index[resolvedPath] = body;
    console.log(
      `[inline] Extracted ${resolvedPath} from ${fileRelativePath}`,
    );
  }

  return stripped;
}