2013-03-18, 09:41
There seems to be a problem when retrieving cast thumbnails via JSON, caused by cast thumb entries not being inserted into the MyVideos75.art table.
I'm testing with OpenELEC (built from git a couple of days ago), Confluence Modified skin, using MySQL for Music and Video.
I've loaded NFO's for TV Shows that all contain valid URLs for actor thumbnails, however when the NFOs were scraped I had "Download actor thumbnails when adding to library" set to disabled, so that when retrieving cast properties via JSON, there were no cast thumbnails returned - I'm happy with that.
Now that I've enabled "Download actor thumbnails when adding to library", actor thumbnails have started to appear in the GUI, downloading and caching on demand, and I can see the urls being retrieved by CurlFile, however the "cast" property results returned by JSON are still missing the thumbnail urls despite the GUI having successfully retrieved and cached a thumbnail.
Looking in MyVideos75.art I'm not seeing any rows corresponding to the dynamically cached cast artwork that is now visible in the GUI, so I'm guessing that the CurlFile process is simply failing to update the MyVideos75.art table correctly.
This debug log extract shows two actor images being loaded, neither of which exist in MyVideos75.art but are being successfully displayed by the GUI.
Is this a known bug, or by design?
Hopefully it's the former as it currently makes it is impossible to query the media library using JSON for cast thumbnails with any degree of reliability.
It seems that cast artwork will be present in the MyVideos75.art table if a show is scraped while "Download actor thumbnails" is enabled - I have 10 such rows in the art table corresponding to two shows scraped in this way. However cast thumbs that are cached/loaded dynamically are entirely missing so I'm guessing the CurlFile process is accessing the database directly rather than using JSON to obtain the url for cast artwork (as JSON denies there are any thumbnails present), and the CurlFile process is then not inserting a row in the art table once it has cached a cast artwork item, ensuring that JSON is oblivious to this artwork.
Movies (scraped with the Universal Movie Scraper) suffer from the same problem, so the issue is not specific to TV Shows (scraped with TMDB).
EDIT: Thinking about this some more, I'm not sure if the rows missing from the art table are the crux of this problem. The JSON queries should be returning cast thumbnail urls once the "Download actor thumbnails" switch is flipped to enabled... the process of downloading these urls should then result in new rows appearing in the art table. Maybe the problem is with JSON and not CurlFile?
I'm testing with OpenELEC (built from git a couple of days ago), Confluence Modified skin, using MySQL for Music and Video.
I've loaded NFO's for TV Shows that all contain valid URLs for actor thumbnails, however when the NFOs were scraped I had "Download actor thumbnails when adding to library" set to disabled, so that when retrieving cast properties via JSON, there were no cast thumbnails returned - I'm happy with that.
Now that I've enabled "Download actor thumbnails when adding to library", actor thumbnails have started to appear in the GUI, downloading and caching on demand, and I can see the urls being retrieved by CurlFile, however the "cast" property results returned by JSON are still missing the thumbnail urls despite the GUI having successfully retrieved and cached a thumbnail.
Looking in MyVideos75.art I'm not seeing any rows corresponding to the dynamically cached cast artwork that is now visible in the GUI, so I'm guessing that the CurlFile process is simply failing to update the MyVideos75.art table correctly.
This debug log extract shows two actor images being loaded, neither of which exist in MyVideos75.art but are being successfully displayed by the GUI.
Is this a known bug, or by design?
Hopefully it's the former as it currently makes it is impossible to query the media library using JSON for cast thumbnails with any degree of reliability.
It seems that cast artwork will be present in the MyVideos75.art table if a show is scraped while "Download actor thumbnails" is enabled - I have 10 such rows in the art table corresponding to two shows scraped in this way. However cast thumbs that are cached/loaded dynamically are entirely missing so I'm guessing the CurlFile process is accessing the database directly rather than using JSON to obtain the url for cast artwork (as JSON denies there are any thumbnails present), and the CurlFile process is then not inserting a row in the art table once it has cached a cast artwork item, ensuring that JSON is oblivious to this artwork.
Movies (scraped with the Universal Movie Scraper) suffer from the same problem, so the issue is not specific to TV Shows (scraped with TMDB).
EDIT: Thinking about this some more, I'm not sure if the rows missing from the art table are the crux of this problem. The JSON queries should be returning cast thumbnail urls once the "Download actor thumbnails" switch is flipped to enabled... the process of downloading these urls should then result in new rows appearing in the art table. Maybe the problem is with JSON and not CurlFile?