Fix xclid again. Fixes #312

twscrape/xclid.py

@@ -1,10 +1,11 @@
+import asyncio
 import base64
 import hashlib
 import math
 import random
 import re
 import time
-from typing import Iterator
+from urllib.parse import urljoin
 
 import bs4
 
@@ -45,16 +46,28 @@ def script_url(k: str, v: str):
     return f"https://abs.twimg.com/responsive-web/client-web/{k}.{v}.js"
 
 
-def get_scripts_list(text: str) -> Iterator[str]:
+# Current X web build (Vite): script bundles are linked directly in the page
+# HTML under https://abs.twimg.com/x-web/.../*.js (modulepreload links + entry).
+ASSET_URL_RE = re.compile(r"https://[\w.-]+/x-web/[\w./-]+\.js")
+
+
+def get_scripts_list(text: str) -> list[str]:
     """
     Extract chunk script URLs from the X homepage HTML.
 
-    As of 2026-05, X embeds two separate maps directly in the page HTML:
+    Current build (x-web / Vite): scripts are linked directly in the page HTML,
+    so we just collect those URLs. If none are found we fall back to the legacy
+    webpack build, which embeds two maps in the page and requires URL
+    reconstruction:
       - Hash map  {chunk_id: "7hexchars"}            values are exactly 7 lowercase hex digits
       - Name map  {chunk_id: "human_readable_name"}  values contain non-hex characters
-
-    URL format: https://abs.twimg.com/responsive-web/client-web/{name}.{hash}a.js
+      URL format: https://abs.twimg.com/responsive-web/client-web/{name}.{hash}a.js
     """
+    urls = list(dict.fromkeys(ASSET_URL_RE.findall(text)))
+    if urls:
+        return urls
+
+    # Legacy webpack build fallback.
     # Hash map: values are exactly 7 lowercase hex digits (distinguishes them from name-map values)
     hash_map = {m.group(1): m.group(2) for m in re.finditer(r'(\d+):"([0-9a-f]{7})"', text)}
 
@@ -68,9 +81,10 @@ def get_scripts_list(text: str) -> Iterator[str]:
         if not re.fullmatch(r"[0-9a-f]{7}", val):
             name_map[m.group(1)] = val
 
-    for chunk_id, hash_val in hash_map.items():
-        name = name_map.get(chunk_id, chunk_id)
-        yield script_url(name, hash_val + "a")
+    return [
+        script_url(name_map.get(chunk_id, chunk_id), hash_val + "a")
+        for chunk_id, hash_val in hash_map.items()
+    ]
 
 
 # MARK: XClientTxId parsing
@@ -216,13 +230,50 @@ def parse_vk_bytes(soup: bs4.BeautifulSoup) -> list[int]:
     return list(base64.b64decode(bytes(el, "utf-8")))
 
 
+# File holding the animation indices: legacy build linked `ondemand.s.*.js`,
+# current x-web build dynamically imports `sign.o-*.js` from a bundle chunk.
+# \b guards against substring hits like `design.o-*.js`.
+INDICES_FILE_RE = re.compile(r"(?:\.{0,2}/)?[\w./-]*?\b(?:ondemand\.s|sign\.o)[\w.-]*\.js")
+
+
+async def _find_indices_url(scripts: list[str], clt: HttpClient) -> str:
+    # The indices file (sign.o-*.js) is not linked in the page directly — it is
+    # dynamically imported from one of the bundle chunks. Scan chunks
+    # concurrently and resolve the first reference we find, then stop.
+    sem = asyncio.Semaphore(16)
+
+    async def fetch(url: str) -> tuple[str, str]:
+        async with sem:
+            try:
+                return url, (await clt.get(url)).text
+            except Exception:
+                return url, ""
+
+    tasks = [asyncio.create_task(fetch(u)) for u in scripts]
+    try:
+        for fut in asyncio.as_completed(tasks):
+            url, body = await fut
+            m = INDICES_FILE_RE.search(body)
+            if m:
+                return urljoin(url, m.group(0))
+    finally:
+        for t in tasks:
+            t.cancel()
+
+    raise Exception("Couldn't get XClientTxId indices script")
+
+
 async def parse_anim_idx(text: str, clt: HttpClient) -> list[int]:
     scripts = list(get_scripts_list(text))
-    scripts = [x for x in scripts if "/ondemand.s." in x]
     if not scripts:
         raise Exception("Couldn't get XClientTxId scripts")
 
-    text = await get_tw_page_text(scripts[0], clt)
+    # Legacy build links the indices file directly; the current x-web build
+    # hides it behind a dynamic import inside a bundle chunk.
+    direct = [x for x in scripts if INDICES_FILE_RE.search(x)]
+    url = direct[0] if direct else await _find_indices_url(scripts, clt)
+
+    text = await get_tw_page_text(url, clt)
 
     items = [int(x.group(2)) for x in INDICES_REGEX.finditer(text)]
     if not items:
@@ -313,6 +364,4 @@ async def main():
 
 
 if __name__ == "__main__":
-    import asyncio
-
     asyncio.run(main())