instagram scraping for likes broke

[snarfed]

first reported by @thorkon yesterday. (thank you!) started sometime in the last few weeks. looks like instagram likes now need an extra HTTP fetch to scrape.

eg if you fetch https://www.instagram.com/p/BmbU2qSFXrR/ , the embedded JSON media object has:

  "edge_media_preview_like": {
    "count": 9,
    "edges": [],
  }

that edges field used to have the individual likes, but now it's empty. if you click on the 9 likes link on the instagram UI, it GETs this URL:

https://www.instagram.com/graphql/query/?query_hash=e0f59e4a1c8d78d0161873bc2ee7ec44&variables={"shortcode":"BmmWVV9lHjI","include_reel":false,"first":24}

which returns:

{
  "status": "ok",
  "data": {
    "shortcode_media": {
      "id": "1848262920796141768",
      "shortcode": "BmmWVV9lHjI",
      "edge_liked_by": {
        "count": 14,
        "page_info": {"..."},
        "edges": [
          {
            "node": {
              "id": "1072653878",
              "username": "kaydeedubya",
              "full_name": "",
              "profile_pic_url": "https://instagram.fsnc1-1.fna.fbcdn.net/vp/2cd0c658b9123a8f67d05301aa875598/5C15E91D/t51.2885-19/s150x150/13712803_750357865105707_625900552_a.jpg",
              "is_private": false,
              "is_verified": false,
              "followed_by_viewer": false,
              "requested_by_viewer": false
            }
          },
          {
            "node": {
              "id": "185218713",
              "username": "smawson",
              "full_name": "Sven",
              "profile_pic_url": "https://instagram.fbkk1-1.fna.fbcdn.net/vp/eaf4663b993b22ab3c90681222cba10e/5C0B007A/t51.2885-19/11906329_960233084022564_1448528159_a.jpg",
              "is_private": true,
              "is_verified": false,
              "followed_by_viewer": false,
              "requested_by_viewer": false
            }
          },
          "..."
        ]
      }
    }
  }
}

i can work with that.

also note that the daily instagram_live_test.py run didn't catch this because it wasn't checking for likes. sigh. i'll fix that.

[snarfed]

shit. this graphql query url requires a login cookie. it 403s if you don't include one.

this may spell the death of backfeeding instagram likes. shit.

[snarfed]

example from the one other project i've found that implemented this: ping/instagram_private_api@78294a8 (thank you @ping!)

[snarfed]

current tentative plan: do all scraping with a login cookie from a throwaway account. IG does already rate limit our scraping, but i believe that's automatic, not us specifically. i expect bridgy is still small enough (5k users, <1k instagram) that IG employees haven't noticed us in particular yet.

[snarfed]

it's alive. god help me, it's alive.

now we wait to see how long it takes to get blocked. 😕