It is difficult to imagine a more specific search than a search on an Opus or Catalogue number when navigating Classical compositions. That is the purpose of opus and catalogue systems. With many famous and/or prolific composers a fuzzy search will therefore lead to quite unmanageable results. For example, if I do a search on Rachmaninov Op. 3, roon will return not 1 result but 56 results. That is, not only every valid combinatoric of Op. xxx containing a 3 but also with multi part works every valid combinatoric of Op. xxx/xx, containing a 3. This is probably an extreme case in my library as Rachmaninov is a favorite composer. But I have long noticed this. Whatâs the sense? I am not sure it was always like this but I can no longer remember when this odd behaviour was introduced.
Thatâs a filter - not a search. Different functions (but using the same damn icon - which has long been complained about).
A filter will indeed return all hits where the text contains the text string used in the filter.
However, I do now notice that the filter seems to be responding to individual text groups, so a filter of Op. 3 is including results such as Op. 26/3 and Op. 14/3.
Is this the change that you mean? It might be worthwhile raising this as a technical support request to get clarification from Support on this.
Yes, thatâs what I mean. The combinatorics of it will really pile up. Especially with core repertoire. But it doesnât matter. Itâs exactly the same. Search or filter. Itâs been there for a long time. I cannot remember when it was introduced. But I donât recall it always being there. Probably introduced as a blanket fuzzy search solution that makes sense in other contexts but not this.
I take no responsibility, beyond complaining about how long it takes to search for The The in my library
Between 7 and 8 seconds now from a cold search. It used to timeout after about 30 seconds
The fuzziness you describe here comes from the way keyword search works, even without actual fuzzy matching. The issue is that at the moment the search engine doesnât strictly match that number 3 is related to the opus number, so as long as all three terms [ârachmaninovâ, âopâ, â3â] are present, itâs a match. The ranking algorithm should have done a much better job here nevertheless. We have a few projects in the pipeline which should improve the query understanding.
Rightfully so Just wanted to say this case hasnât been forgotten, itâs just that there isnât an easy fix so it will most likely be corrected in combination with some bigger changes in the library search.
Zenith, to be honest while not perfect, things have improved greatly ( at least for most of my use cases), across multiple updates that were delivered by Roon.
I still see the odd daft result but I can often see why they were made (like typing in the Band name Bush and having Kate Bush appearing as the default response, Bush now appears second, it used to be further down the results)
It will be nice to see a return to this subject and maybe giving users some options on how they want results displayed as well. But I certainly appreciate that some good work was done
I see. The filter functionality that you showed on My Compositions is separate from the search engine and behaves also a bit differently from it. Iâll notify the team about the issue.
But at least roon is still returning variations on Bach.
Other search sections like albums and tracks seem to understand that 33 is the search term rather than 3 and generally do a better more selective job but then throw in totally unrelated results like Vol. 33 and HMV 33 which is Handel.
Just wanted to say this case hasnât been forgotten, itâs just that there isnât an easy fix so it will most likely be corrected in combination with some bigger changes in the library search.
