Messages

31

Bugs / Re: Scan results not remembered

« on: April 02, 2011, 18:53:30 »

Program is inconsistent in remembering previous data after new start: Comparison results are discarded, analysis data are kept. The latter may be sometimes favorable but it may clutter the analysis window in other cases. The only way I found to get rid of them is to start an analysis run and immediately stop it thereafter.
I'd like to have a reset button for analysis data similar to that for the cache.
The same applies to ignore data (files marked to be ignored). A reset button for that is a much cleaner way than to delete an ignore file somewhere in Microsoft's program data jungle.

32

Bugs / Re: "Cannot delete file. Cannor read from source file or disc". Accept missing files

« on: April 02, 2011, 18:31:47 »

Cache indication of 0 after comparison start seems to be a bug: After restart (before comparison run) former cache size is shown again. Comparison speed is also a hint that cache is used: The fast speed at the beginning means that signatures from cache are used, slower speed appears when signature is not in cache and has to be calculated from audio source.
I agree that missing files should be misregarded. A similar annoying problem: If you mark a couple of files and try to execute an action on these (e.g. delete), execution is refused if one file does no longer exist. You have to reduce the marking range or even act on single files until you have the noexisting file no longer in your selection. If you have another nonexisting file in the rest of your selection - same procedure again.

33

Wishlist / Re: Consider different releases

« on: March 28, 2011, 17:58:38 »

I've been using Jaikoz, Picard and Classical MB Tagger for years but they give me only results for 10% of newly introduced material (off mainstream). With similarity I get a reasonable hint whether I have that song already in my collection and especially if it's of better quality.
Without clean definitions you will only get arbitrary statements, no matter how the wind blows 

34

Wishlist / Re: Consider different releases

« on: March 28, 2011, 10:58:42 »

The meaning of "identical" depends on user. Some want to get rid of any duplicates no matter on what CD they were published. Others want ot keep all copies of a song in varying samplers. Some of them may originate from the same recording session and from the same digital sampling but with different compression, others may differ in mixing and digital conversion (re-edit) from the same analog tapes and additionally others are different recording sessions (e.g. single and album versions). At the extreme live sessions with greatly varying content and length may have the same filename and tags (artist - title).
As a consequence: Give as much information as possible to the user (tags, musical similarity, quality) and let him decide. It should be his decision whether to rely on automatic generated data what to throw away and what to keep.

35

Bugs / Re: Similarity stops scanning after a few songs

« on: February 24, 2011, 10:00:46 »

Had the same problem (not after a few but after several thousand files). Reappeared after restart.
I disabled all codecs except ACM (I only examined MP3s). It didn't stop every time at the same file. Seems to depend on processing history, too. Trial with debug version (build 1033) is on the way.
My bug fix: I split the task in groups (e.g. A-N + O-Z) using the /portable switch in order to to keep two separate caches. They can then be run sequently or even parallel on a multicore system.
Hint:
If I want to know which file is being processed, I use resource monitor (included in windows 7). For former windows versions the process monitor in the sysinternals suite does the same job. Task manager does not deliver enough information.

36

Bugs / Spectrum analysis overoptimistic

« on: February 22, 2011, 18:44:39 »

The max. frequency value in the spectrum analysis window is in some cases too optimistic. It seems to be oriented on max. frequency level (red), not on average frequency level (blue). When the average level drops down, while the max. level stays high, this an indication that the noise level drastically increased at higher frequencies caused by alias effects by violating the nyquist limit or by high frequency mismatch during compression/decompression.
In this case the rating value is misleading (higher value - worse quality).

37

Bugs / 1.5.3 odd behaviour

« on: February 05, 2011, 18:35:18 »

Sorry to say that the current version of similarity (1.5.3 build 1001) got less stable:

For large numbers of songs (>50K) the duplicate scan freezes from time to time, no CPU and file activity any more. When restarted, program stops at the same progress indication (xx.x%) again. When starting environment is changed (different directories or cache cleared) program sometimes arrives at 100%, sometimes stops elsewhere.

After successful search usage of GUI resources often fails: No reaction to ctrl-S or right click to spectrum analysis, no reaction to Enter or Play and other right click menu items. After some playing around (resorting, resizing etc.) ctrl-S suddenly works but after a few successful actions fails again.

Additional strange behaviour: Cursor shape does not change from border icon (N-S-arrow) to normal in client area, toggles rapidly back and forth between normal and N-S-cursor on action fields. Menu bar icons disappear and reappear whereas balloon tips and click actions function normally.

38

General / Re: How do you find files with the same name?

« on: January 30, 2011, 10:48:37 »

I still consider file name similarity not to be in the primary scope of similarity.
But there might be a workaround for you:
If you are willing to build artist & song name tags with rules from filename (tools like MP3tag can do that using sophisticated rules with a mass processing ability), you can use the tag similarity slider for building duplicate groups. Tag similarity is done on a fuzzy approach. This could work as a substitute for a file name slider, as long as you haven't found a fuzzy file name comparison tool or it is not implemented in similarity.

39

General / Re: How do you find files with the same name?

« on: January 29, 2011, 19:27:38 »

OK, I understand that your request is not just looking for same name files. I suppose you want to keep a more original version even if it doesn't have the same name.
If there were an option to move all files in a duplicate group into a directory by some user input or even by an automatic \\root\artist scheme, it would do what you want and even more. That would make sense for me in the similarity context. By adjusting the %-threshold you can fine tune what you get in a duplicate group. You can play with that after having performed an extensive search (raising the threshold), you don't need to restart the search.
This group member moving/copying is an idea I would strongly support.

If you are only interested in working on identically named files, a simple explorer search would do. In Windows 7 a following mark and copy/move even renames them to <file> (1), <file> (2) etc. Other explorer clones may do that even in earlier windows versions.

40

General / Re: How do you find files with the same name?

« on: January 28, 2011, 20:29:21 »

I'm not sure whether that makes sense. There are dozens of other tools doing exactly that.
It is almost contrary to the primary aim of similarity. I would rate that as a peripheral option with low priority.

41

Bugs / Re: Spectrum analysis for groups crashes

« on: January 28, 2011, 20:21:27 »

Happens each time (except for single file). Error reports have been submitted, don't know whether they are received/processed.

42

Bugs / Spectrum analysis for groups crashes

« on: January 26, 2011, 20:09:34 »

When spectrum analysis is called by right click for two or more files, it crashes constantly  .
Fortunately only function call crashes, results windows remains operable, single call of spectrum via Ctrl-S works.

43

Bugs / Re: File editing

« on: January 26, 2011, 19:50:52 »

Two selections "copy tags" - "copy all" would be ok.
But at the moment there is a nasty bug: If you copy filename from another file in the group, afterwards the rest by copy all (or vice versa) and then press "apply" the changes from copy all are discarded.
So you are forced to call the edit window twice, once for filename and once for copy all.

44

Bugs / File editing

« on: January 16, 2011, 22:22:16 »

Copy all in the edit window does not copy filename. Doesn't matter too much: Click on file name tag completes task. Otherwise good idea: Copying filenames from another grouped  file is what is almost always wanted.

45

Bugs / Re: false 100% similarity

« on: January 10, 2011, 20:18:18 »

The clue point of similarity is that it works just by comparing two tracks acoustically, not by comparison with a database entry. Musicbrainz IDs are no longer updated, they were replaced by Picard IDs (PUID). Both cover only part of released material. If you are not in mainstream you may get matches for only a a minor fraction of your candidates (I get typically less than 10%). Additionally, the problem of different instances of the same recording (with differing quality) isn't adressed at all.