View Full Version : IMDB Scapper problem
ultrabrutal
2007-04-22, 02:41
A fresh install of latest T3CH build from 0420 still does not work with a movie called "Jalla! Jalla!.avi". It fetches info from another movie even though the movie is on the list if you manually fetch movie info. Howcome it doesn't use the name of the item which matches 100%? Should be easy to reproduce with a dummy file named just this. The problem has been there since forever ;)
jmarshall
2007-04-22, 03:06
Is it at the top of the list?
The movie lookup stuff doesn't use fuzzy string matching on the returned results (yeah, it probably should)
ultrabrutal
2007-04-22, 09:53
No it's not on the top, it's a bit down the list. I think I have the same problem with some other files but this is an example.
Fuzzy string matching, or sound-a-like as I call it (MSSQL), really is a must if you have misspelled your movie, but this is not the case here. They are 100% identical.
Should be fairly simple to iterate through all the movies returned and if a perfect match, like in this case, is found just pick it. Alot easier to do than the fuzzy stuff unless there is functionality in the xdk for it.
Tnx
jmarshall
2007-04-22, 13:24
We do that already in the allmusic.com search, so the functionality is there in some regard already. Whether or not it will offer a better result than imdb's matching (which is the order we provide currently) is non-obvious really though.
Will have a chat with some other devs about it.
Remember, you can always add a .nfo with the correct imdb url in to make sure it finds the correct info.
ultrabrutal
2007-04-23, 11:18
jgawera, thanks I know, but I rather have XBMC work better than having to create text files to workaround ;)
to fix this we have to rewrite the imdb scraper to dont fetch cronologically but rather skip the 'popular searches' matches. iirc that's the main issue
I don't think any implementation is going to be perfect. I use nfo files with the correct url. That leaves me with the peace of mind to know that the result is always going to be right, and if I have to clean my database I can leave the wrong when the scan is being done and I know it will get everything.
It's only a 2 second job to create the files.
perfect; no. but its certainly possible to improve on todays situation, thats all i'm saying.
i dont have the inspiration to do it, but there's the receipe