g1t/apps/web/app/routes/repo/archive.ts

67 lines3,707 bytesCodeBlame
1/**
2 * Download ZIP: a branch, tag or commit of a repository as one zip, its
3 * files under `<repo>-<ref>/` as `git archive` would name them. For anyone
4 * who can see the repository. Made on request and not kept, so it is capped:
5 * a repository past the caps is cloned instead.
6 */
7import type { Route } from "./+types/archive";
8import { repos } from "../../lib/services.server";
9import { getViewer } from "../../lib/session.server";
10import { zip } from "../../lib/zip";
11
12/** The most files, and bytes in all, an archive is made of. */
13const MAX_FILES = 10_000;
14// The bytes, their base64 and the zip are all held at once, in a Worker of 128 MB.
15const MAX_BYTES = 24 * 1024 * 1024;
16/** Blobs read per call to the repository, and calls at once. */
17const PER_READ = 100;
18const READS_AT_ONCE = 8;
19
20function refused(status: number, message: string): Response {
21 return new Response(`${message}\n`, { status, headers: { "content-type": "text/plain; charset=utf-8", "cache-control": "no-store" } });
22}
23
24export async function loader({ params, context }: Route.LoaderArgs) {
25 const viewer = getViewer(context);
26 const asked = params["*"] ?? "";
27 if (!asked.endsWith(".zip") || asked.length <= 4) return refused(404, "Ask for <branch, tag or commit>.zip.");
28 const ref = asked.slice(0, -4);
29 const repo = await repos.get({ namespace: params.owner, name: params.repo }, viewer).catch(() => null);
30 if (!repo?.ok) return refused(404, "There is no such repository, or you cannot see it.");
31 const listed = await repos.listFiles(repo.value.id, ref, MAX_FILES + 1).catch(() => null);
32 if (!listed?.commit) return refused(404, `There is no branch, tag or commit named ${ref}.`);
33 const clone = `git clone https://g1t.sh/${repo.value.namespace}/${repo.value.name}.git`;
34 if (listed.truncated || listed.files.length > MAX_FILES) {
35 return refused(413, `It has more than ${MAX_FILES.toLocaleString("en-US")} files, too many for a download. Clone it instead: ${clone}`);
36 }
37 const files = listed.files.filter((file): file is { path: string; hash: string } => file.hash != null);
38 const unique = [...new Set(files.map((file) => file.hash))];
39 const bytes = new Map<string, Uint8Array>();
40 let total = 0;
41 let tooLarge = false;
42 const batches: string[][] = [];
43 for (let at = 0; at < unique.length; at += PER_READ) batches.push(unique.slice(at, at + PER_READ));
44 // A few reads at a time, each taking the next batch, until all are read or it is too large.
45 const reader = async () => {
46 for (let batch = batches.shift(); batch && !tooLarge; batch = batches.shift()) {
47 for (const blob of await repos.rawBlobs(repo.value.id, batch, MAX_BYTES)) {
48 total += blob.size;
49 if (total > MAX_BYTES || (blob.data == null && blob.size > 0)) tooLarge = true;
50 else bytes.set(blob.hash, Uint8Array.from(atob(blob.data ?? ""), (c) => c.charCodeAt(0)));
51 }
52 }
53 };
54 await Promise.all(Array.from({ length: READS_AT_ONCE }, reader));
55 if (tooLarge) return refused(413, `It is over ${MAX_BYTES / 1024 / 1024} MB, too large for a download. Clone it instead: ${clone}`);
56 // A branch name can hold slashes; the folder and file name cannot.
57 const label = `${repo.value.name}-${ref.replaceAll("/", "-")}`;
58 const archive = await zip(files.map((file) => ({ path: `${label}/${file.path}`, data: bytes.get(file.hash) ?? new Uint8Array() })));
59 return new Response(archive, {
60 headers: {
61 "content-type": "application/zip",
62 "content-disposition": `attachment; filename="${label}.zip"`,
63 // A commit's files never change; a branch's do.
64 "cache-control": /^[0-9a-f]{40}$/.test(ref) ? "private, max-age=31536000, immutable" : "private, no-cache",
65 },
66 });
67}