Slowdown with beet web - regression #2182
Comments
|
@cobra2 said that with this commit included it takes 80-100 seconds, and with the change reversed it's about 10 seconds. |
sampsyo
added
the
needinfo
|
Thanks for the report. @maxammann, you added this path field—how necessary is it really? We could consider just exposing the real file path, which should be much faster, although I'm not 100% sure why we need the path here at all. |
|
@sampsyo : probably this /~https://github.com/maxammann/music-cyclon "Android app to synchronize music or files over network by using the beets web server"" EDIT: definitely that. it says "Waiting for pull request: #2050 " . That's the change that introduced this. |
|
Got it; thanks. I'm just not 100% sure why it needs us to generate paths… couldn't that happen client-side? Anyway, we could make it a config option to decide whether to expose real file paths. (Off by default for privacy's sake.) |
|
Should we consider this a blocker bug? |
|
Maybe we should just roll back the change and call it good? If @maxammann is interested in adding it back, that's of course cool but not worth blocking release. |
|
So what about #2177 ? You mention that PR as the solution |
|
Yeah, I'm not exactly sure what's going on there. It's not entirely clear to me that they need the path in the web API---in any case, it shouldn't be hard to add back if necessary. |
|
This is also in the changelog, so if it does get backed out, it'd have to be removed from there too. |
ghost
added
the
|
Currently traveling for this month but I'll look into it when I come home again. For now just revert that commit if it makes trouble. |
|
I think its still a valuable feature. Maybe we should add a flag to toggle "expensive" output? |
|
Yeah, absolutely—I do think it should be an option. |
|
Back at home, going to fix this and push a pull request in a few days. Is it urgent right now, because it's a blocker? Or does the new release take a few more weeks? |
|
@maxammann : which route are you gonna take for a fix? |
|
@jrobeson sorry, had some family related problems the past week, I'll introduce a new flag to disable the expensive output. |
|
a flag to disable? shouldn't it be a flag to enable? In the future, it shouldn't nearly be so expensive, so hopefully we can can drop the flag all together. |
|
Only @sampsyo can say how urgent this is. I'm not sure when he's ready to push a new release. |
|
@jrobeson yeah, sorry, I mean a flag to enable :P |
…he 'path' variable" This reverts commit 5e8ac9e, because of a slowdown. Resolves beetbox#2182.
|
Sorry for the delay, a lot of stuff going on. Just reverted that commit. :D |
There is a massive slowdown in queries with python2 and python3 when using the beet web interface.
When large queries are run, i.e. 'format:flac' on a flac only library, the web interface posts queries 10x slower than before the regression. To clarify that does seem to be a relative time based on the length of time originally taken by the query.
This is caused by a regression in the commit.
5e8ac9e
beet stats
Tracks: 43913
Total time: 51.4 weeks
Approximate total size: 976.18 GiB
Artists: 7345
Albums: 12004
Album artists: 1800
The text was updated successfully, but these errors were encountered: