[danbooru][error] "KeyError: 3456687" when downloading posts with extractor.danbooru.metadata=true

[HexHummingbird]

Hello.
I get an error when I'm trying to download favorite images of a user from danbooru with metadata option set to true.

gallery-dl.exe --verbose -s -o extractor.danbooru.metadata=true 'https://danbooru.donmai.us/posts?tags=ordfav%3Akiyah123'
[gallery-dl][debug] Version 1.25.3 - Executable
[gallery-dl][debug] Python 3.8.10 - Windows-10-10.0.19045-SP0
[gallery-dl][debug] requests 2.28.2 - urllib3 1.26.15
[gallery-dl][debug] Configuration Files []
[gallery-dl][debug] Starting SimulationJob for 'https://danbooru.donmai.us/posts?tags=ordfav%3Akiyah123'
[danbooru][debug] Using DanbooruTagExtractor for 'https://danbooru.donmai.us/posts?tags=ordfav%3Akiyah123'
[urllib3.connectionpool][debug] Starting new HTTPS connection (1): danbooru.donmai.us:443
[urllib3.connectionpool][debug] https://danbooru.donmai.us:443 "GET /posts.json?tags=ordfav%3Akiyah123&limit=200 HTTP/1.1" 200 None
[danbooru][debug] Sleeping 1.00 seconds (request)
[urllib3.connectionpool][debug] https://danbooru.donmai.us:443 "GET /posts.json?tags=ordfav%3Akiyah123&limit=200&page=b6223003&only=artist_commentary%2Cchildren%2Cnotes%2Cparent%2Cuploader%2Cid HTTP/1.1" 200 None
[danbooru][error] An unexpected error occurred: KeyError - 3456687. Please run gallery-dl again with the --verbose flag, copy its output and 
report this issue on https://github.com/mikf/gallery-dl/issues .
[danbooru][debug] 
Traceback (most recent call last):
  File "gallery_dl\job.pyc", line 96, in run
  File "gallery_dl\extractor\danbooru.pyc", line 61, in items
  File "gallery_dl\extractor\danbooru.pyc", line 114, in _pagination
KeyError: 3456687

Above command works correctly without metadata option:

gallery-dl.exe --verbose -s 'https://danbooru.donmai.us/posts?tags=ordfav%3Akiyah123'
[gallery-dl][debug] Version 1.25.3 - Executable
[gallery-dl][debug] Python 3.8.10 - Windows-10-10.0.19045-SP0
[gallery-dl][debug] requests 2.28.2 - urllib3 1.26.15
[gallery-dl][debug] Configuration Files []
[gallery-dl][debug] Starting SimulationJob for 'https://danbooru.donmai.us/posts?tags=ordfav%3Akiyah123'
[danbooru][debug] Using DanbooruTagExtractor for 'https://danbooru.donmai.us/posts?tags=ordfav%3Akiyah123'
[urllib3.connectionpool][debug] Starting new HTTPS connection (1): danbooru.donmai.us:443
[urllib3.connectionpool][debug] https://danbooru.donmai.us:443 "GET /posts.json?tags=ordfav%3Akiyah123&limit=200 HTTP/1.1" 200 None
# danbooru_6223002_dcae6a56a31897e056cb9e9dec329cc0.jpg
...

[mikf]

The "improvements" from 75666cf do not work all that well with non-standard sorting orders like ordfav:, it seems.

Try gallery-dl version 1.25.1 where this change wasn't included yet, or do not order the posts by time favorited.

[mikf]

Turns out that this problem is not confined to only metadata; the current pagination logic does not return the correct/full results for any post order other than the default order:id_desc.

[mikf]

Fixed in fd0e1ff and 494acab.

Using -o metadata=true will no longer result in a crash when using ordfav: or any other order: tag, and it will actually download posts in the correct order instead of always implicitly using order:id_desc.