Bug 1877 - Artist wth accent causes incorrect/duplicate entries in browser
: Artist wth accent causes incorrect/duplicate entries in browser
Status: RESOLVED FIXED
Product: Logitech Media Server
Classification: Unclassified
Component: Database
: 6.1.1
: Macintosh MacOS X 10.4
: P2 normal (vote)
: ---
Assigned To: Dan Sully
:
Depends on:
Blocks:
  Show dependency treegraph
 
Reported: 2005-07-25 17:38 UTC by Michael Robinson
Modified: 2009-09-08 09:13 UTC (History)
0 users

See Also:
Category: ---


Attachments
Debug logs during music folder scan (257.44 KB, text/plain)
2005-07-25 17:42 UTC, Michael Robinson
Details
slimserversql.db file artist with accent in title (95.00 KB, application/octet-stream)
2005-07-25 17:43 UTC, Michael Robinson
Details
slimserversql.db file artist without accent in title (95.00 KB, application/octet-stream)
2005-07-25 17:43 UTC, Michael Robinson
Details

Note You need to log in before you can comment on or make changes to this bug.
Description Michael Robinson 2005-07-25 17:38:21 UTC
An artist with an accent in the title causes duplicate and incorrect listings in the browser.

Symptoms are as follows.

Artist: Sin�ad O'Connor
Albums with Sin�ad O'Connor as Artist
- Collaborations		
- Sean-N�s Nua	
- Internet Alternative		
- Internet Folk	

Artist: Massive Attack
Albums with Massive Attack as Artist
- 100th Window

Sin�ad O'Connor is listed as composer for 2 tracks on 100th Window

Browsing by artist shows Sin�ad O'Connor listed twice

Selecting first Sin�ad O'Connor shows 3 albums:-
- Collaborations		
- Sean-N�s Nua
- 100th Window

First 2 albums are correct but Massive Attack is artist for 100th Window

Selecting 100th Window shows 1 album with 2 songs by 0 artists
Tracks listed are the 2 tracks with Sin�ad O'Connor as composer

Selecting 2nd Sin�ad O'Connor shows other 2 albums:-
- Internet Alternative		
- Internet Folk	

Using a tag editor, the accent the � in Sin�ad was replaced with e in all tracks.

The database was cleared and rescanned.

Browsing by artist now shows the correct listing with Sinead O'Connor appearing once with 4 albums 
listed.

Problem is reproduceable.  Re-adding the e with accent and rescanning recreated the problem.

I have attached debug logs and the slimserversl.db files for:-

Artist with accent (slimserversql.db 1) and Artist without accent (slimserversql.db 2)
Comment 1 Michael Robinson 2005-07-25 17:42:31 UTC
Created attachment 670 [details]
Debug logs during music folder scan

2 sets of debug logs when scanning music folder.  First is with artist title
with accent.  Second is with artist title without accent,
Comment 2 Michael Robinson 2005-07-25 17:43:24 UTC
Created attachment 671 [details]
slimserversql.db file artist with accent in title
Comment 3 Michael Robinson 2005-07-25 17:43:55 UTC
Created attachment 672 [details]
slimserversql.db file artist without accent in title
Comment 4 Dan Sully 2005-07-25 17:45:15 UTC
Do you have a playlist with this song in it?

Can you please upload that as well?

Thanks
Comment 5 Dan Sully 2005-07-25 17:45:37 UTC
Also - are you using iTunes?
Comment 6 Fred 2005-07-28 12:59:08 UTC
This may be related, not sure...

If a playlist references a song whose path includes an accented char, AND if the playlist is for some 
reason parsed before the directory, the server can't find the song...

2005-07-28 20:25:11.9690 New track for file:///Volumes/Sid/slimp3/Music/Compilations/Hotel%
20Costes%204/Cafe%C3%8C%C2%81%20De%20Flore%20[Charles%20Webster%27s%20Latin%20Lovers%
20Mix].mp3
2005-07-28 20:25:11.9701 readTag was 1 for file:///Volumes/Sid/slimp3/Music/Compilations/Hotel%
20Costes%204/Cafe%C3%8C%C2%81%20De%20Flore%20[Charles%20Webster%27s%20Latin%20Lovers%
20Mix].mp3
2005-07-28 20:25:11.9717 Converting file:///Volumes/Sid/slimp3/Music/Compilations/Hotel%
20Costes%204/Cafe%C3%8C%C2%81%20De%20Flore%20[Charles%20Webster%27s%20Latin%20Lovers%
20Mix].mp3 to /Volumes/Sid/slimp3/Music/Compilations/Hotel Costes 4/Cafe?�� De Flore [Charles 
Webster's Latin Lovers Mix].mp3
2005-07-28 20:25:11.9740 mp3 file type for file:///Volumes/Sid/slimp3/Music/Compilations/Hotel%
20Costes%204/Cafe%C3%8C%C2%81%20De%20Flore%20[Charles%20Webster%27s%20Latin%20Lovers%
20Mix].mp3
2005-07-28 20:25:11.9749 reading tags for: file:///Volumes/Sid/slimp3/Music/Compilations/Hotel%
20Costes%204/Cafe%C3%8C%C2%81%20De%20Flore%20[Charles%20Webster%27s%20Latin%20Lovers%
20Mix].mp3
2005-07-28 20:25:11.9775 Info: no tags found for /Volumes/Sid/slimp3/Music/Compilations/Hotel 
Costes 4/Cafe?�� De Flore [Charles Webster's Latin Lovers Mix].mp3
2005-07-28 20:25:11.9791 Info: no title found, using plain title for file:///Volumes/Sid/slimp3/
Music/Compilations/Hotel%20Costes%204/Cafe%C3%8C%C2%81%20De%20Flore%20[Charles%20Webster
%27s%20Latin%20Lovers%20Mix].mp3

Later on, the link is deleted...

2005-07-28 20:25:20.5301 Merging entry for file:///Volumes/Sid/slimp3/Music/Compilations/Hotel%
20Costes%206/Un%20Monde%20Sans%20Frime.mp3
2005-07-28 20:25:20.5516 CacheItem: Checking status of file:///Volumes/Sid/slimp3/Music/
Compilations/Hotel%20Costes%204/Cafe%C3%8C%C2%81%20De%20Flore%20[Charles%20Webster%27s%
20Latin%20Lovers%20Mix].mp3 (TTL: 1122575117).
2005-07-28 20:25:20.5557 deleting file:///Volumes/Sid/slimp3/Music/Compilations/Hotel%20Costes
%204/Cafe%C3%8C%C2%81%20De%20Flore%20[Charles%20Webster%27s%20Latin%20Lovers%
20Mix].mp3 from cache as it no longer exists

This last delete because "if (-e $filepath) {" fails in _hasChanged... The song is named "Caf� de Flore 
[bla bla].mp3".

In this example I end up with an extra track for the ghost song. A rescan playlists fixes it for some 
reason, but a rescan or wipecache don't.

HTH
Comment 7 Dan Sully 2005-07-29 14:01:45 UTC
Is this still the case using a 6.2 nightly?
Comment 8 Michael Robinson 2005-07-29 15:32:32 UTC
Dan - I tried the 6.2 nightly and the problems with duplicate artists and incorrect tracks under Browse 
Artists still exist however I have been doing more investigation into the problem and believe there are 
two separate issues:-

1. Slimserver is treating the same artist name with different encodings as different artists
2. Tracks where the composer is the same as the artist are being added to the Browse Artists listings 
even if the Server Settings / Behaviour setting is "Don't include composer, band and orchestra in artists"

I mentioned in the bug report that the artist name Sin�ad O'Connor appeared twice when browsing by 
artist.  It occured to me that the tracks under one instance of the name had been ripped using iTunes 
and the tracks under the other instance had been downloaded from the internet.

I use iTunes to edit tags and there is a menu option in iTunes under Advanced called Convert ID3 Tags.

The Convert ID3 Tags menu is grayed out when I selected tracks ripped by iTunes but it could be used 
for the tracks I had downloaded from the internet

Options within this menu include Translate text characters: ASCII to ISO Latin-1, ISO Latin-1 to ASCII 
and Reverse Unicode.

Selecting various conversion options changed the � in Sin�ad to various other characters.  I used the 
option to convert from ISO Latin-1 to ASCII and then in iTunes I copied and pasted the text from the 
artist field of a track I had ripped using iTunes into the artist field of a track I had downloaded from the 
internet.

I did this for all the tracks under the second instance of Sin�ad O'Connor then cleared and rescanned 
the library.

Sin�ad O'Connor now appeared only once under Browse Artists.

It looks very much that even though the artist name was the same in iTunes, the tag data was encoded 
in different ways as the tracks were from different sources. Slimserver then treated them as different 
artists under Browse Artists.

The second issue mentioned in the bug report was that tracks from a Massive Attack album with Sin�ad 
O'Connor as the composer were listed under Sin�ad O'Connor when browsing by artist.

I had "Don't include composer, band and orchestra in artists" set under Server Settings / Behaviour.

I didn't have any other tracks in my music where the composer in one artist's track was the same as the 
artist name in another so I tried the following:-

Composer set to "A" in album "P" of artist "X"
Composer set to "B" in album "Q" of artist "Y"
Composer set to "C" in album "R" of artist "Z"

Where "A", "B" and "C" were artist names already in the library.

After a clear and rescan, browsing artists "A", "B" and "C" listed albums "P", "Q" and "R" of artists "X", "Y" 
and "Z" respectively.

This is a different issue from the duplicate artists - I just noticed it because I had tracks with Sin�ad 
O'Connor as the artist and composer.

I don't think this is the correct behaviour - if the option "Don't include composer, band and orchestra in 
artists" is set then surely it shouldn't include ANY composer, but it appears to list tracks where the 
composer is the same as the artist.

I no longer have the problem with the duplicate listing of artists as the encoding of my tag data has 
been cleaned up but I'm guessing other people may have the same issue caused by different encodings 
of accented characters from different sources.


Comment 9 Dan Sully 2005-08-01 19:08:31 UTC
Michael - I've fixed the first part of this from happening again in subversion
change 3844.

Could you open a separate bug for the second portion?

Thanks.
Comment 10 Blackketter Dean 2005-08-11 12:30:18 UTC
Comment in bug 1877 indicates that this problem is back. 
Comment 11 Dan Sully 2005-08-11 12:33:21 UTC
The bug is actually with MusicMagic, and is reported as bug 1938.

Closing this bug.

Will have a fix shortly for bug 1938.
Comment 12 Michael Robinson 2005-08-22 16:44:02 UTC
This problem has reoccurred in the 2005-08-22 nightly.

An artist title with an accent (Sin�ad O'Connor) is listed twice when browsing by artist.

One listing contains .m4a albums (some AAC, come ALAC) and the other listing contains mp3 albums.

I took a look in the slimserversql.db file and there are two different Sin�ad O'Connor artist contributors.

id	name	                namesort	                        namesearch
15	Sin?�ad O'Connor	SINA�AD O CONNOR	SINA�AD O CONNOR
404	Sin?�ad O'Connor	SINEAD O CONNOR	        SINEAD O CONNOR

(The tool I use to export slimserversql.db tables doesn't appear to handle accents properly)

The m4a tracks have contributor id 15

The mp3 tracks listed separately have contributor id 404
Comment 13 Dan Sully 2005-08-22 17:28:21 UTC
Mike - did this change since my last fix, or has it always been like this?

Can you attach one of each of the mp3 & m4a files?

thanks.
Comment 14 Michael Robinson 2005-08-23 12:14:32 UTC
Dan - I thought this had been fixed in a previous nightly... I have copied 2 files to ftp.electricrain.com 
Trance Nation 2002 s12 Sinead O'Connor - Troy ( Push Remix ).mp3 and Mandinka.m4a
Comment 15 Dan Sully 2005-08-23 17:47:52 UTC
Ok - I believe I have this fixed in subversion change 4037.
Comment 16 Michael Robinson 2005-08-24 14:07:07 UTC
Dan, I tried the 2005-08-24 nightly and the duplicate names issue appears to be fixed.

However there appear to be other issues when browsing tracks in this nightly. No Album, No Artists and 
No Genre appear when browsing by Album, Artist and Genre however tracks are correctly tagged.  Albums 
with the compilation tag set also appear under Various Artists even when all tracks on these albums have 
the same artist.

This has appeared between 2005-08-23 and 2005-08-24 nightlies.
Comment 17 Dan Sully 2005-08-24 14:19:55 UTC
Michael - that's correct behavior for the Compilation albums. If you don't want that - remove the tag.

Can you upload your current db file so I can track down the No * entries?

Thanks.
Comment 18 Michael Robinson 2005-08-24 17:19:35 UTC
I just noticed the new Compilation settings under Server Settings / Behavior

Personally, I think it is a bit of a stretch to assume that compilation = Various Artists as iTunes/CDDB 
will tag most single artist "Best of..." type albums as compilations.

I'm also a bit confused about the default settings of:
- List compilation albums under each artist
- Don't include compilation artists in artist list

I don't understand the logic of listing compilation albums under the artist but not including the artist in 
the artist list.

In my collection, there are several examples where "Best of..." single artist albums that are tagged as 
compilations by iTunes are now listed under Various Artists rather than the actual artist.

Another consequence is that MMM mixes based on these albums will fail.  I believe this is because 
Slimserver requests a mix using Various Artists / Album Title while MMM stores the album as Artist / 
Album Title.  I have provided more information in bugid 1938.

Personally I would prefer the default behaviour to ignore the compilation setting as I believe this was 
the previous behavior.  Similar to the points I made previously about the Composer field, iTunes or 
other tagging programs may set the Compilation field but Slimserver didn't do anything with it (until 
now...) so I concentrated on sorting out a tagging convention for the fields that Slimserver did use (such 
as genre, album and artist).  Doing clever things depending on the Compilation field should be 
optional.

I have copied mr_slimserversql.db.zip to ftp.electricrain.com.

The No Artist / No Genre / No Album tracks all seem to be downloads from www.stompy.com and 
www.beatport.com.  These are either WAV downloads that I tagged and converted to Apple Lossless 
using iTunes or non-DRM AAC downloads.

They all appeared correctly prior to the 08-24 nightly.