feat: Implement caching and rate limiting for API requests

server/schedules.js

@@ -1,13 +1,18 @@
 import { readFile, writeFile, mkdir } from "node:fs/promises";
 import { fileURLToPath } from "node:url";
 import { parseWord, PARSER_VERSION } from "./parser.js";
-import { request } from "./source.js";
+import { request, MIN_REFRESH } from "./source.js";
 
 const data = fileURLToPath(new URL("../.data/parsed/", import.meta.url));
+// When each document was last downloaded, and the downloads under way: a forced refresh
+// downloads a document at most once per MIN_REFRESH, and simultaneous ones share a download.
+const fetchedAt = new Map(),
+  loading = new Map();
 
 /** A parsed catalog document, from the disk cache while the file and parser are unchanged. */
 export async function scheduleFor(file, { refresh = false } = {}) {
   const cache = data + file.id + ".json";
+  if (refresh && Date.now() - (fetchedAt.get(file.id) || 0) < MIN_REFRESH) refresh = false;
   if (!refresh)
     try {
       const cached = JSON.parse(await readFile(cache, "utf8"));
@@ -17,7 +22,17 @@ export async function scheduleFor(file, { refresh = false } = {}) {
       )
         return cached;
     } catch {}
+  if (!loading.has(file.id))
+    loading.set(
+      file.id,
+      download(file, cache).finally(() => loading.delete(file.id)),
+    );
+  return loading.get(file.id);
+}
+
+async function download(file, cache) {
   const buffer = await request(file.path);
+  fetchedAt.set(file.id, Date.now());
   const parsed = await parseWord(buffer, {
     name: file.name,
     url: file.url,