Grouping Issue Redocumented

This TIDAL album:

has a grouped composition when the album is un-IDed:

… but ungrouped when IDed:

Possibly, Roon’s heuristic disallows grouping for non-consecutive tracks.

I don’t know if this is a flaw, but it sure isn’t a feature. :slight_smile:

When identified, the grouping is (currently) down to an editorial decision at TiVo who deem the performance of the composition “incomplete”. When unidentified, the grouping is based on common text in the track titles. Nothing sinister or unintended here.

1 Like

Do you really believe the Roon treatment here to be more superior than grouping six contiguous tracks with identical canonical names?! Did you even think to allow the possibility that users shouldn’t have to choose between an album being IDed and a Composition being IDed? Or did you just immediately go on the defensive?

And “sinister”? You think my post inferred “sinister” behavior?

Wow, talk about the “not invented here” syndrome. As long as it is “working as intended”, it has to be right, eh? Even better with Tivo as an additional buffer, eh Joel. “Not our decision.”

I’ll save you some future time Joel. You have more than adequately revealed yourself over the last year. You rarely have anything constructive or useful to say. But if your job is to defend the status quo, they owe you a raise.

Don’t bother responding to my threads any more please. If and when you become more interested in saying something other than defending what you’ve programmed, feel free to join back in. Go tell someone else how everything is working as intended for them.

1 Like

@joel’s message reads as a clinical explanation of the current behavior to me. I don’t read it as defensive, and don’t think it merited such a harsh response.

As Joel was explaining–In your second screenshot, Rovi/Tivo decided that these tracks constitute separate performances. This was an editorial decision on their part. In the first screenshot, you are seeing a fallback attempt to group based on text of the track titles. I do not expect that an algorithm always comes up with the same result as an editor.

You and I were discussing the performance/composition distinction the other day with respect to a case that failed in the other direction–towards overzealous grouping. In this case, an editor has likely decided that because these tracks were not physically performed contiguously on the same day, in the same location, they constitute separate performances.

This is a tricky problem to solve, as there are complex historical, practical, editorial, and automated aspects to how performances are grouped. We are not going to be able to fix this problem one album at a time. Individual examples of wrong-ness are only useful in-so-far as they inform a set of rules that we could implement in the software. So far, while we have exchanged some ideas, a clear vision for improving this hasn’t emerged. I hope it does, but until then, the behavior is as it is. I’m open to overriding Rovi’s editorial decisions on this topic, but only once we have clarity on exactly what the right rules are, as this will be a pretty disruptive change, especially if we get it wrong.

I would like to continue having constructive discussions with you about problems in this area because I think you have useful ideas to contribute, but it becomes difficult when your tone is so aggressive. If your post was attacking anything other than Roon Labs, it would have been removed by the moderators, but we try to keep criticism out in the open. I’d ask that you avoid veering into personal attacks in the future.


I apologize for my tone.

By the way, Allmusic (tivi/rovi I believe) recognizes this album on its website:

Here is how they present the partial “Gadfly” composition there:


Perhaps I misinterpreted Joel’s comment. To me, it was a blow-off, albeit more polite than my response.

This was, IMO, an interesting edge case that should say more to Roon than “working as expected”.

I find this relatively common with several sub-genres. For example:

  • Ballet where there are ‘highlights’ that do not necessarily fall into a ‘cataloged’ suite, or suites have been combined or the conductor and/or the producer have just decided to play the suite in a different order.
  • Opera ‘highlights’, oratario highlights, mass/religious service highlights and excerpts of other lengthy choral/vocal work.
  • Works with accompanying spoken word where the spoken word has been removed from the performance (or alternatively is present). For example, poetry reading in a symphony like Shostokavich’s Babi Yar or Bernstein’s Carnival of the Animals.

I also don’t see how it is possible to automate rules that would catch all these and other similar cases. So what I do is I WORK/PART them to get the grouping that I want. But what I cannot do is edit the WORK title of multi-part excerpts/highlights like this to indicate that is what they are.

1 Like

Yeah, this was a TIDAL album, so WORK/PART isn’t available.

I don’t see how to automate this either. But Roon, for reasons not fully understood by me, is reluctant to give users a unilateral right to group tracks.

The first thing I would do, if I were in Roon’s position, would be to ask Rovi why its sister company (Allmusic) groups these tracks, but it tells Roon not to.

This looks like it is an excerpt of a suite (Op. 97a) arranged by Atovmian from the complete film score (Op. 97). So rovi has almost certainly got the opus number wrong and it is also missing one of the composers (arranger?)

I like this work so I looked around and as far as I can see on different disks, different excerpts of the suite op. 97a is quite common. I frequently see this sort of looseness with dramatic music (ballet, opera, sacred) so I guess a film score by such a famous composer is just a modern example of this. Personally I am ok with manually grouping such works where it is difficult to make general rules. But I would like the freedom to edit the work title so I know if it is a sub-opus or even an excerpt of a sub-opus.

I don’t have the full film score but as far as I can see there is only one (Naxos) and that is a reconstruction which would also require another composer credit:

Just to not mislead anyone reading this.

After a bit more digging in turns out Tidal’s Gadfly is an excerpt of an excerpt of the original Op. 97 orchestration (Chailley 1996). But I left the original post up because it shows how hard it can be to get the details right. I personally just don’t see that happening on the meta-data supplier side in these sorts of cases but it would be nice for those who are prepared to do the digging or have a special interest in some favorite works to have a simple editing ability to customize multi-part work titles so you know what you are listening to.

If we do this, it is good for you, but bad for the product. You will go away, stop bothering us, and go fix up your library, and the other 99% will go on with the bad experience because you are both an outlier in your willingness to edit AND in the noise level you are able to generate :stuck_out_tongue:

The above actually gives too much credit to editing. You might be willing to edit your own stuff, but even 50 of you can’t edit the whole TIDAL catalog. We have some work in the pipeline which is going to make out-of-library content more sanely navigable/accessible. A lot more people are going to come into contact with it, and thus these problems. The interesting solution to this is the one that makes a large statistical improvement across everything–not the solution that allows a few interested people to go away and make their local files perfect.

If we get to a point where no further automated improvement seems possible, we will re-consider more editing capabilities, but this is going to be treated as a last resort.

Anyways–let step back for a second. So much of the past discussion on this has focused on examples of wrongness. What I’m missing is the pattern. I could use help with a few questions. These are questions I’m trying to answer for myself, but I’d appreciate any input.

(1) What prompts you to care about performance grouping when it is wrong? in other words, what concrete user experience problems result from it being wrong/what motivates you to want to fix it or demand that we fix it.

(2) Is composition/performance grouping something that is essentially “ground truth” or is this is in the domain of “personal preference”. Or somewhere in between. If personal preference, please elaborate on the kinds of differences in preference you might expect from person to person.

This seems like a good start. I would really like to have a productive discussion where I can walk away with some actions for the team to improve this, so lets keep it focused on composition/performance-grouping and “downstream” issues that result from it for now.

Hi Brian,

I’m coming from the Opera side of things. It is important for me to have the music displayed by the following playable headers:

Act 1
…Scene 1
…Scene 2
Act 1
…Scene 1
…Scene 2

When I listen to Opera and I don’t want to listen to the whole opera, then I would usually choose an Act or a Scene grouping. If I"m being very specific I’ll just choose an aria.

The problem is that for almost every opera I have that is identified, the grouping is non-existent. As a result, I pretty much leave my operas as unidentified and let the basic track title grouping take precedence.

Please take a stab at my two discussion points above. We could easily get lost in trying to pick apart the best thing for each musical style and miss the higher level view. We’ll get into more specific stuff like examples, specific musical styles, etc later. Lets nail the high level first…

@brian I suggest that you relocate your post at the start a new thread. The issues you raised are too important to be “hidden” in an apparently unrelated topic.

Good question! I could say, “a sense of order” coupled with my inability to do anything about it. But what set me off this time was that the Performances page lists each track, so it shows me with 10 performances instead of three or so. Secondarily, I noticed that Allmusic had managed to group these same tracks, and why not Roon.

We may to iterate on this one, but I’ll try. To contrast with this case, I have another “composition” in a particular album was both 1) out of order (purposively by the publisher), and 2) non-contiguous. That composition, IMO, shouldn’t be grouped and I didn’t care beyond a mild WTF in the direction of the label.

I didn’t particularly care about this album until I saw the Performance page issue (above). Then I said, “no biggie, I’ll just force a grouping and go about my business.” First rub: this is a TIDAL album, so my trusted WORK/PART tricks weren’t applicable.

Ten minutes later, after performing all the troubleshooting and workarounds I could muster, I concluded that Roon was only going to let me group that composition if I un-IDed the album. But if I un-IDed, I would lose: my album cover (unless I could find a substitute : more time); the other comp’s IDs; and the related metadata for them. Not a good trade.

It was about then that I went on tilt, much to my regret. :unamused:

In sum, I hated “being stuck” without an acceptable solution. And this “hate” (poor word choice) transcended the actual event, because I had encountered this many times before. I could care less about Roon’s lack of a “correct” heuristic. I just wanted to fix an “obvious” problem, and couldn’t.

It’s about now the therapist says, “well, our time is up.” But follow-ons welcome.

My take on it …

One thing - admittedly foremost a collector’s problem - which comes with grouping issues: wrong performance counts. So sorting by performance count becomes unreliable. I may not have 4 performances of a work but two or even just one. Or not 20 but 5.

Another related thing: I may not be able to play a performance as a whole just by selecting it on the Composition page but might have to clamp it together after switching into album view.

Since it is a typical classical music problem: rightly joined performances might also help to somehow better get around the “mixed genre problem”. Genres could get assigned to compositions and thereby help with “multigenre” albums (boxsets especially). I understand genres are not for track level - compositions / performances might be a different thing.

Also, at times it just doesn’t feel right. A few days ago I wanted to listen to some Haydn String Quartets (I have a multidisc set with all quartets) and choose to play one CD (out of 19). The Roon grouping was set to opus / movement - which made me wonder after an hour or so why the queue wasn’t empty yet. But if the work is grouped that way, it merges multiple quartets into one work and spans more than one CD. Roon - always keeping compositions together as grouped - filled the queue with much more tracks than I’d intended. Grouped that way, the actual quartet structure is gone. It feels wrong to me.

For compositions some of it should be considered a ground truth - if there exists a score then this is (somehow) definitive. If one looks up the hierarchy - starting at the smallest piece the composer separated in the score, like the movement - things become lesser distinct the higher one goes. Therefore I’d prefer a bottom-up approach especially when only two levels of distinction are available - which I believe is a major problem for classical music. An opus assignment might not originate from the composer but some publisher:

… it may have been the publisher who decided there had to be Three Trios op. 22 … but Trio No. 2 definitely has four movements so this information I’d like to see.

There’s sure enough room for discussion or quite some academics would be out of their jobs … but most of it should be reasonably settled for a metadata service. Some of it may be something like “additional” truth - as certain editions and collections.

For performances I’d say a single performance for live recordings is easy to determine (same place same day same personnel). For studio recordings things become more difficult - one recording (a performance) could take more than a day. A week at times. But if there are months in between or the location changes I would go for different performances even if it’s pressed on the same album.

(There are special cases too - where the mixing becomes part of the artists intent; edge cases)

What I do miss most regarding performance grouping is: count distributed parts not as individual but as a single performance. (see here)

Surely you’ve noticed that there’s more than a little OCD among Roon’s customers. We want consistency. You are right that much of Roon’s value comes from its automated library organization. But mistakes and inconsistencies are inevitable.

  1. The others have discussed why it matters. Besides feeling right, having the right grouping makes it easier to find and play the performance we want.

  2. If a composer intended a composition to be played at once, its movements should certainly be grouped. Even if some just want the hear the last movement of Beethoven’s Ninth, I’m comfortable with imposing a grouping on the entire symphony. But collections of dances or concertos are much more an issue of preference. Vivaldi’s Four Seasons is a collection of four concertos. Should it be grouped as four, or one? Roon is inconsistent on this. And I’ll bet its customers are divided in their preferences.

Think of it this way: mistakes are inevitable, and sufficiently common that users should be able to make their own corrections. If you allow that, wouldn’t most editing for preferences follow the same process?


1 Like

Point 1.

No one feels a need to justify structuring other forms of written communication. A book with no phrases, no sentences, no paragraphs, no chapters, no thematic sections? I don’t think so. Half the lengthy posts on this forum would simply be unreadable without a grammatical structure.

I also think “it looks wrong” can be quantified. It is the lack of a “sense of ownership” you get without it. Its at the root of comments about “premium product”. With roon this spills over into a lack of focus on album and artist art. Same issue as a lack of focus on composition grouping. The general air of a product rough around the edges really destroys a sense of ownership and loyalty. It’s the reason a car manufacturer will spend so much R&D on getting the “clunk” right when you close the door.

This is a common problem with digital product. For example Google Books often gets the formatting, indenting and other structuring wrong in an e-Book. Recently I had to buy a technical e-Book quickly for professional reasons, so I downloaded it. Formatting of footnotes and references was about 90% right, but 10% wrong is plenty wrong. Despite the advantages of being able to search, highlight, annotate etc. I found the book difficult and tiring to read on-line and ended up getting a physical copy as well. Easier to read and an undefinable sense of ownership.

Point 2.

My general preference is always to group even when I have an incomplete composition. This is extremely common with ballet, opera, sacred and lengthy vocal/choral work generally and also with live and recital format disks where excerpts from more lengthy works is the norm. The only reason I break that rule is because of other (for me) inconsistencies in roon functionality.

So for example I like to merge compositions so I have easy access to all versions of a work. But I would also like to do that at a part level where the part is more famous than the work, or it is just a favorite part. For example, the Debussy Prelude Book 1/8 “Girl with the flaxen hair”. In this case where I have an incomplete recital of 2 or 3 of the Preludes including “Flaxen girl” I will often break “Flaxen girl” out of the incomplete group. But this is only so I can merge it with other stand-alone versions I might have. My preference would always be to maintain the grouping but be able to merge compositions at a part as well as a work level.

The other reason I break up a group is because of the way that roon treats groups in radio and playlists. Vivaldi’s four seasons is a good example. I have a lot of Baroque music and my preference in radio mode would be to listen to a single Four Seasons movement, not even a complete concerto. So as a compromise I always break that one up into 4 where roon has concatenated the 4 concertos, so radio doesn’t boil down to listening to Four Seasons all day. I do a similar thing where roon has concatenated Books of preludes, etudes or songs. But it is only a work-around to fix these other issues. My preference would be that roon supported a composition hierarchy greater than 2 and that it was context aware in its treatment of grouped music. That is, I want to play the entire group through when browsing in album or composition mode but I want to shuffle single parts when in radio or playlist mode. Others will have different preferences.

1 Like

Playback must be considered when discussing grouping.

It is difficult (for me) to discuss grouping on playback without imagining “how”.

Here goes: I want one extra toggle switch, located near the other playback options, that says “Grouping On” and that is the default setting. A user, at any time, for any type of playback, can toggle the switch. Immediately, Roon shifts from playing compositions, to playing tracks. Elegant, simple, neat. (Or, if not on screen, put it in preferences)

Easier said than done I know.

I too would like innovation on Classical music playback. Roon does a great job now, but please keep improving the user experience.

I strongly advocate for options for manually being able to assign genre by work and randomized play by track. An album might couple a 20thC Orch piece and a Romantic Concerto. If you don’t believe all Classical music is the same, these are different genres. To recognize as the same or one genre negatively impacts effective sorting for many Classical music enthusiasts.

And why not have the option to listen to randomized Classical music tracks? As with some others on the forum we’d like to be able to do this from time to time.

1 Like

Yes. I have never understood why grouped playback of Classical works is fixed in stone. It depends entirely on the listening scenario. Pretty much all “mood”, “best of”, live and recital format Classical discs shuffle excerpts. Some full length ballets have as many as 4/5 different suite “excerpts” written by the original composer and/or an arranger/orchestrator for recital use. Most “Books” of preludes, etudes, songs etc. were collected together by the publishers, not the composers who wrote them at wildly different stages of their career and never imagined or intended that they would always be played all together, all at once. Large numbers of popular opera discs are compilations of arias going back to the early days of recording with Carusso. It’s a long time since most people had the time to sit through an entire Bach High Mass, Handel Oratario or Grand Opera. Almost all physical world Classical radio stations shuffle single movements. It is an extremely common Classical music listening scenario which makes most sense in a roon context with radio, discovery or playlist mode. Listening through an entire work end to end makes perfect sense in the context of the album or composition playing mode.

1 Like