diff --git a/docs/spec/fail-open-to-original.md b/docs/spec/fail-open-to-original.md
index 86e309a..bf462ee 100644
--- a/docs/spec/fail-open-to-original.md
+++ b/docs/spec/fail-open-to-original.md
@@ -10,7 +10,9 @@ adr: null
---
分享連結端點 `GET /<原網域>/<路徑>` 在抓取或解析失敗時，一律 302 回清理後的原網址，不回錯誤頁，也不回缺少媒體的 og 頁。
-抓取和解析包在同一個失敗邊界裡。邊界內的例外、上游非 200 回應、找不到必要欄位，都走同一個 302。已知的失敗訊號包括：Instagram 被導到 `/accounts/login` 或 `unsupportedbrowser`、embed 頁只顯示 `WatchOnInstagram`、Facebook `post.php` 找不到 `t39.30808-6` 貼文圖或作者、Facebook `video.php` 找不到 `hd_src`/`sd_src`、`original_width`/`original_height` 或作者。Facebook 的失敗頁一律回 200，有些完全沒有提示文字，所以只能靠缺欄位判定，不能靠比對字樣。
+抓取和解析包在同一個失敗邊界裡。邊界內的例外、上游非 200 回應、找不到必要欄位，都走同一個 302。已知的失敗訊號包括：Instagram 被導到 `/accounts/login` 或 `unsupportedbrowser`、圖片貼文的 embed 頁只顯示 `WatchOnInstagram`、多圖貼文被標為版權封鎖（`copyright_blocked` 或 `WatchOnInstagram`）、被標為版權封鎖又沒有封面圖的影片、Facebook `post.php` 找不到 `t39.30808-6` 貼文圖或作者、Facebook `video.php` 找不到 `hd_src`/`sd_src`、`original_width`/`original_height` 或作者。Facebook 的失敗頁一律回 200，有些完全沒有提示文字，所以只能靠缺欄位判定，不能靠比對字樣。
+
+被標為版權封鎖的 Instagram 影片，embed 頁沒有 `video_url`，但仍有封面圖，這時改用封面圖產生 og 頁，不算缺少媒體。
這條只管分享連結端點。`GET /?url=` 轉換端點和 `/media/` 端點的錯誤處理不在範圍內。
diff --git a/src/instagram.ts b/src/instagram.ts
index 48234fe..1028805 100644
--- a/src/instagram.ts
+++ b/src/instagram.ts
@@ -277,21 +277,27 @@ function richItems(type: string, context: unknown): Media[] | null {
return items;
}
+function imagePost(username: string, html: string): Post | null {
+ const mediaUrl = mediaUrlOf(html);
+ if (mediaUrl === null) return null;
+ return { username, caption: captionOf(html), media: [{ kind: "image", url: mediaUrl }] };
+}
+
export function parseEmbed(html: string): Post | null {
try {
- if (html.includes("WatchOnInstagram")) return null;
+ const watchOnly = html.includes("WatchOnInstagram");
const type = mediaType(html);
const username = usernameOf(html);
if (!username) return null;
- if (type === "GraphImage") {
- const mediaUrl = mediaUrlOf(html);
- if (mediaUrl === null) return null;
- return { username, caption: captionOf(html), media: [{ kind: "image", url: mediaUrl }] };
- }
+ if (type === "GraphImage") return watchOnly ? null : imagePost(username, html);
if (type !== "GraphVideo" && type !== "GraphSidecar") return null;
const context = readContext(html);
if (context === null) return null;
- if (asRecord(asRecord(context)?.context)?.copyright_blocked === true) return null;
+ // A copyright-blocked video has no video_url in the embed, but the page still
+ // shows its cover frame, which beats no preview at all.
+ if (watchOnly || asRecord(asRecord(context)?.context)?.copyright_blocked === true) {
+ return type === "GraphVideo" ? imagePost(username, html) : null;
+ }
const media = richItems(type, context);
if (media === null) return null;
return { username, caption: captionOf(html), media };
diff --git a/test/fail-open-to-original.test.ts b/test/fail-open-to-original.test.ts
index 040d5c0..94fe9ca 100644
--- a/test/fail-open-to-original.test.ts
+++ b/test/fail-open-to-original.test.ts
@@ -106,10 +106,25 @@ test("a crawler is redirected when the cache binding throws", async () => {
}
});
-test("a crawler is redirected when the embed video is blocked", async () => {
+test("a crawler receives an og page with the cover frame for a blocked reel", async () => {
const html = await fixture("embed-Dd0M_ifNfXO.html");
- const fake = await expectOpen(() => Promise.resolve(new Response(html, { status: 200 })), 1);
- expect(fake.entries.size).toBe(0);
+ const app = createWorker({ cache: () => fakeCache().cache });
+ const spy = spyOn(globalThis, "fetch").mockImplementation(
+ (() => Promise.resolve(new Response(html, { status: 200 }))) as unknown as typeof fetch,
+ );
+ try {
+ const res = await app.fetch(
+ new Request("https://bl.example/www.instagram.com/reel/Dd0M_ifNfXO/", { headers: { "User-Agent": DISCORD } }),
+ {} as Env,
+ ctx,
+ );
+ expect(res.status).toBe(200);
+ const body = await res.text();
+ expect(body).toContain('');
+ expect(body).not.toContain("og:video");
+ } finally {
+ spy.mockRestore();
+ }
});
test("a crawler on an instagram item link is sent to the post when fetch fails", async () => {
@@ -149,8 +164,8 @@ test("a crawler asking for a missing carousel item is sent to the post", async (
);
});
-test("a crawler on a blocked reel is sent to the reel", async () => {
- const html = await fixture("embed-Dd0M_ifNfXO.html");
+test("a crawler on a blocked reel without a cover frame is sent to the reel", async () => {
+ const html = (await fixture("embed-Dd0M_ifNfXO.html")).replace('class="EmbeddedMediaImage"', 'class="Other"');
const fake = await expectOpen(
() => Promise.resolve(new Response(html, { status: 200 })),
1,
diff --git a/test/instagram.test.ts b/test/instagram.test.ts
index b9f7730..3d1cc9a 100644
--- a/test/instagram.test.ts
+++ b/test/instagram.test.ts
@@ -155,24 +155,49 @@ function richPage(type: string, node: unknown): string {
return `u
`;
}
-test("a copyright-blocked video embed yields no post", async () => {
- expect(parseEmbed(await fixture("embed-Dd0M_ifNfXO.html"))).toBeNull();
+const COVER_FRAME = [
+ {
+ kind: "image" as const,
+ url: expect.stringMatching(
+ /^https:\/\/scontent\.cdninstagram\.com\/v\/t51\.82787-15\/825269646_18629137711044933_165147988821785013_n\.jpg\?/,
+ ),
+ },
+];
+
+test("a copyright-blocked video embed yields its cover frame as an image", async () => {
+ const post = parseEmbed(await fixture("embed-Dd0M_ifNfXO.html"));
+ expect(post?.username).toBe("atlanticrecords");
+ expect(post?.media).toEqual(COVER_FRAME);
+ expect(post?.media[0]?.url).not.toContain("&");
+});
+
+test("a copyright-blocked video embed without a cover frame yields no post", async () => {
+ const html = (await fixture("embed-Dd0M_ifNfXO.html")).replace('class="EmbeddedMediaImage"', 'class="Other"');
+ expect(parseEmbed(html)).toBeNull();
});
-test("a watch-on-instagram marker on a playable video yields no post", async () => {
+test("a copyright-blocked carousel yields no post", async () => {
+ const html = (await fixture("embed-Dd0M_ifNfXO.html")).replace(
+ 'data-media-type="GraphVideo"',
+ 'data-media-type="GraphSidecar"',
+ );
+ expect(parseEmbed(html)).toBeNull();
+});
+
+test("a watch-on-instagram marker on a playable video yields its cover frame", async () => {
const html = (await fixture("embed-DJvkjAlvNc8.html")).replace(
'vatsalya_therapy',
'vatsalya_therapyWatch on Instagram',
);
- expect(parseEmbed(html)).toBeNull();
+ expect(parseEmbed(html)?.media).toEqual([{ kind: "image", url: expect.any(String) }]);
});
-test("a video whose context is copyright blocked yields no post", async () => {
+test("a video whose context is copyright blocked yields its cover frame", async () => {
const html = (await fixture("embed-DJvkjAlvNc8.html")).replaceAll(
'\\"copyright_blocked\\":false',
'\\"copyright_blocked\\":true',
);
- expect(parseEmbed(html)).toBeNull();
+ expect(parseEmbed(html)?.media).toEqual([{ kind: "image", url: expect.any(String) }]);
});
test("a video without a video url yields no post", async () => {