13

Over on Ask Ubuntu, we have been using this query to find out old posts with non-imgur images. It was working alright before this weekend, but now it's returning 0 rows. What's going wrong?

10

The query is specifically looking for images (i.e. img tags) that have a HTTP src. As stated in Roadmap to HTTPS: serving and uploading HTTPS-images only, those have been nuked (silently by the Community user, without bumping posts). They were either auto-converted to HTTPS images or plain text links; neither of which will be found by that query.

SEDE is only updated weekly, on Sundays, so the query would have returned the already fixed posts until today when it was updated.

10
  • 2
    Yep, just realised that myself. Now it seems only devs can find the images to be fixed. – muru Mar 12 '17 at 10:56
  • 2
    @muru the affected posts have all been added to the crowd rafting project. You should be able to pull the post ids from the API and filter by site, I'm going to see if I can work something when I've got a bit more time later tonight. – Cai Mar 12 '17 at 13:43
  • if you can, that would be awesome. I got stuck in the "same two posts in a loop" problem that has affected others and would rather not use that site. – muru Mar 12 '17 at 13:46
  • 2
    Yep, me too, and promptly gave up after that :) I'll let you know if I manage to come up with something – Cai Mar 12 '17 at 13:52
  • 1
    There's also a way to download full csv or json here: crowdcrafting.org/project/sehttpimagescleanup/tasks/export – Cascabel Mar 12 '17 at 14:44
  • 1
    And... it does not look like all the data is good :( It starts out sane: "info" : { "BaseHostAddress" : "meta.stackexchange.com", "PostId" : "150", but then looking for entries for cooking, I find: "info" : { "askubuntu.com" : "cooking.stackexchange.com", ... "149" : "1561" – Cascabel Mar 12 '17 at 14:52
  • @Jefromi looks like the API returns the same... not sure whats going on there :( – Cai Mar 12 '17 at 15:07
  • Yeah, I assume this is because the actual uploaded tasks are bad (maybe some kind of escaping bug), not because we're using the export/API wrong, so if people actually got to those tasks they'd be broken. I commented on the original question asking about it. For what it's worth I think the values are legit, e.g. post 1561 on cooking does have an http image link, but it makes it pretty hard to use in general, and I'm especially not sure how to use the link field. – Cascabel Mar 12 '17 at 15:12
  • 1
    @muru I've written a snippet to filter the crowdcrafting tasks by site, see here: meta.stackexchange.com/a/292279/307988 – Cai Mar 13 '17 at 13:42
  • @Jefromi see above – Cai Mar 13 '17 at 13:43
3

Images are undergoing migration from http to https as part of the grand Stack Exchange migration to https.

Removing http from the query produces some results for me, although feedback suggests it's flawed somehow.

Rene has created a better query since mine seems to be missing results.

7
  • 2 results, whereas it should be in the hundreds (about 400, keeping track with the results from last weekend) – muru Mar 12 '17 at 9:28
  • 470 results @muru – rene Mar 12 '17 at 9:36
  • @rene I've added that to the answer. – Robert Longson Mar 12 '17 at 9:39
  • 1
    @rene you're searching for HTTPS images, which don't need fixing. – muru Mar 12 '17 at 9:44
  • I need coffee then ... sorry. – rene Mar 12 '17 at 9:49
  • @muru I rechecked but there are simply no more links on http. All are replaced with https links as stated by Robert. – rene Mar 12 '17 at 10:38
  • 1
    @rene If you believe that, I have got a bridge... The last time the query worked, we got these links: etherpad.net/p/au-img-cleanup You can check those which haven't been fixed already to see that there are still posts without HTTPS images (either being dead or HTTP-only). (on that etherpad, we're removing http:// from the post links to mark them as fixed) – muru Mar 12 '17 at 10:41

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .