feat: extract and index inline file blocks in markdown

This commit is contained in:
hypercross 2026-07-08 15:46:19 +08:00
parent 9a5ee695c2
commit 2a9eee6410
2 changed files with 101 additions and 5 deletions

View File

@ -11,6 +11,7 @@ import {
scanCompletions,
type CompletionsPayload,
} from "../completions/index.js";
import { extractInlineBlocks } from "../inline-blocks.js";
interface ContentIndex {
[path: string]: string;
@ -105,7 +106,16 @@ export function scanDirectory(dir: string): ContentIndex {
) {
try {
const content = readFileSync(fullPath, "utf-8");
index[normalizedRelPath] = content;
if (entry.endsWith(".md")) {
const stripped = extractInlineBlocks(
content,
normalizedRelPath,
index,
);
index[normalizedRelPath] = stripped;
} else {
index[normalizedRelPath] = content;
}
} catch (e) {
console.error(`读取文件失败:${fullPath}`, e);
}
@ -295,9 +305,18 @@ export function createContentServer(
try {
const content = readFileSync(path, "utf-8");
const relPath = "/" + relative(contentDir, path).split(sep).join("/");
contentIndex[relPath] = content;
if (relPath.endsWith(".md")) {
const stripped = extractInlineBlocks(
content,
relPath,
contentIndex,
);
contentIndex[relPath] = stripped;
recomputeCompletions();
} else {
contentIndex[relPath] = content;
}
console.log(`[新增] ${relPath}`);
if (relPath.endsWith(".md")) recomputeCompletions();
} catch (e) {
console.error(`读取新增文件失败:${path}`, e);
}
@ -312,9 +331,18 @@ export function createContentServer(
try {
const content = readFileSync(path, "utf-8");
const relPath = "/" + relative(contentDir, path).split(sep).join("/");
contentIndex[relPath] = content;
if (relPath.endsWith(".md")) {
const stripped = extractInlineBlocks(
content,
relPath,
contentIndex,
);
contentIndex[relPath] = stripped;
recomputeCompletions();
} else {
contentIndex[relPath] = content;
}
console.log(`[更新] ${relPath}`);
if (relPath.endsWith(".md")) recomputeCompletions();
} catch (e) {
console.error(`读取更新文件失败:${path}`, e);
}

68
src/cli/inline-blocks.ts Normal file
View File

@ -0,0 +1,68 @@
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;
}