app/browse/lib.py

   1 import requests
   2 from datetime import datetime, timezone
   3
   4 def fetch_searchresults(q=None, page=1, sp=None):
   5   for _ in range(2):
   6     today = datetime.now(timezone.utc).strftime("%Y%m%d")
   7     r = requests.get(f"https://www.youtube.com/results", {
   8         'q': q, # Note: if we use '?search_query=' (as yt.com does), we can't use '&page=', but have to use a continuation url that requires an api key
   9         'pbj': 1, # makes youtube return a json-response
  10         'hl': 'en', #'en_US',
  11         'page': page,
  12         'sp': sp,
  13     }, headers={
  14         'x-youtube-client-name': '1',
  15         'x-youtube-client-version': f'2.{today}.01.01', # the version is parsed as a date, and if it's invalid (e.g. month>12 or even feb>=30), youtube throws an encrypted stacktrace :D (but any random date >= 20160323 as of 20200802 works (even year 3000)
  16     })
  17     if not r.ok:
  18         return None
  19
  20     # Sometimes, youtube throws an exception after the response already begun.
  21     # So the status code is 200, begins with JSON and switches to HTML half way
  22     # through. WTF?! (This should be "fixed" by retrying, though)
  23     if r.text.endswith("</html>"):
  24         continue # will return None once we break out of the loop
  25
  26     return r.json()
  27
  28 def fetch_ajax(params):
  29     """
  30     fetch data using a continuation protobuf
  31     """
  32     # TODO: handle auto_generated!
  33     today = datetime.now(timezone.utc).strftime("%Y%m%d")
  34
  35     r = requests.get(f"https://www.youtube.com/browse_ajax", {
  36         'continuation': params,
  37         'gl': 'US',
  38         'hl': 'en',
  39     }, headers={
  40         'x-youtube-client-name': '1',
  41         'x-youtube-client-version': f'2.{today}.01.01', # see fetch_searchresults()
  42     })
  43
  44     if not r.ok:
  45         return None
  46
  47     return r.json()