2015-09-07, 11:40
I have just discovered the consequences for my music library of clicking on Album Information for a single album. I knew that there was an option to query information for all albums, and that that used scrapers to get info from online databases etc. it turns out that the scraper is also used for a single album.
Why is this an issue? Well when the scraper gets it wrong it quietly removes the correct album artist data that had been scanned into the library from tags. So for example I choose album information on the context menu for my Dvorak Symphony no. 7 by Jarvi/SNO album on Chandos, it finds a Colin Davis/LSO on a different label recording and adds them as artists but deletes Jarvi and SNO. One tiny click, no promting or warnings, and they are gone!!
The album and songs are still there and show the original artist string when playing or in lists so it took me a while to notice, but the song_artist and album_artist tables entries have been cleared. Since it is the only album by that conductor and orchestra "Jarvi" and "SNO" do not appear in the artists list and Kodi doesn't find the album or songs when I search for those names.
I can see in the code why this happens. I understand that scrapers can be useful and powerful things, and that correctly identifying recordings can be difficult. What I question is if this quiet corruption of the music library is the behaviour we really want?
I think that at very least initiating scraping should be a separate user choice from looking at information. I knew I did not want to scrape, or if I did I wanted to see and confirm what updates happen. Maybe you want an auto scrape of the entire library, fine have that as an option but unbidden deletion of correct data is not good behaviour.
Thoughts anyone?
Why is this an issue? Well when the scraper gets it wrong it quietly removes the correct album artist data that had been scanned into the library from tags. So for example I choose album information on the context menu for my Dvorak Symphony no. 7 by Jarvi/SNO album on Chandos, it finds a Colin Davis/LSO on a different label recording and adds them as artists but deletes Jarvi and SNO. One tiny click, no promting or warnings, and they are gone!!
The album and songs are still there and show the original artist string when playing or in lists so it took me a while to notice, but the song_artist and album_artist tables entries have been cleared. Since it is the only album by that conductor and orchestra "Jarvi" and "SNO" do not appear in the artists list and Kodi doesn't find the album or songs when I search for those names.
I can see in the code why this happens. I understand that scrapers can be useful and powerful things, and that correctly identifying recordings can be difficult. What I question is if this quiet corruption of the music library is the behaviour we really want?
I think that at very least initiating scraping should be a separate user choice from looking at information. I knew I did not want to scrape, or if I did I wanted to see and confirm what updates happen. Maybe you want an auto scrape of the entire library, fine have that as an option but unbidden deletion of correct data is not good behaviour.
Thoughts anyone?