#metabrainz

iliekcomputers: why did you close this https://github.com/metabrainz/listenbrainz-serv... ?

I mean I was following it, didn't find any closing comment!

morning

mo’’in’

morning!

_lucifer: you about?

yes pristine___

I was looking at my recs. I mostly listened to indian artists. In the last week, I listed to a few songs of. Tame Impala. I see that the first page of my top artists playlist is filled with tame Impala songs, meaning these recordings have highest score, as I move to the second, third page of the top artist playlist, I see artists more evenly distributed and more indian artists. A few users have commented on the

community post, this also has troubled me for quite some time, the playlist (especially the first page) being centred around a few artists. One is reason I can think of is, LB has a few users who listened to indian songs, so maybe my similar user (based on user-user similarity) is purely based on English songs. That's is why all the song of tame Impala have highest score even though I don't listen to them so

frequently.

Just a finding, I thought it will good to share with you given you are working on a part of recs now :)

yeah, sure. thanks for sharing

the easiest way here seems to limit the number of recordings from an artist on the first page

but that is probably a work for troi

_lucifer: exactly, I have discussed about this with ruaok

We haven't reached on any conclusion though

It really looks odd when the first page is centered around only one artist, I guess most of us don't move to the next page.

_lucifer: not very critical to solve rn, but a good to have thing.

That's true, most won't go to the next oage

yeah, for now the page says tracks you might like so it not very important but when these tracks become a playlist then it'll be

I see coldplay........ coldplay aaaand coldplay

n I didnt even listen to it last week

shivam-kapila:

I see `The Chainsmokers & Coldplay` here

You can do a look up for you top artist from stats endpoint here

Most probably the artists, recordings aren't in the mapping except Coldplay, that is why you have only one artist in the playlist

Did you read the community post shivam-kapila ? I am sure it will clear your doubts regarding the nature of playlist, if not let me know, I will add in there, will be helpful for other users too :)

> the easiest way here seems to limit the number of recordings from an artist on the first page

_lucifer: I was thinking to apply `random` on the recs list and then show on site.

pristine___: thx. Will do

> _lucifer: I was thinking to apply `random` on the recs list and then show on site.

Doesnt sound right. Each time I press next and then previous the recs change

But this isnt ideal imo

shivam-kapila: do random before showing them on site, and not while you show them on site

Is what I mean

What if I reload

pristine___: could be a good idea

shivam-kapila: i think the idea is to randomize in spark instead of sorting on score so the results seen by the user will be consistent

Then we shouldnt show score then

As there wont be a reasonable pattern

yes right

pristine___: is there any way to see the current queue of requests in request consumer?

Moin!

> shivam-kapila: i think the idea is to randomize in spark instead of sorting on score so the results seen by the user will be consistent

That _lucifer :)

_lucifer: iliekcomputers would know, I have no idea on that :(

np, i have the mapping placed at the right place but i know need to upload it to hdfs somehow

Just use the command in spark_manage

I linked yesterday.

yeah, i am using that but it is not working due to some disk space issues. i'll free up some space and do that

Mr_Monkey: https://tickets.metabrainz.org/projects/LB/issu...

LB-729: Scrobble tracks streamed on ListenBrainz website

Yeah, that's been on the list for a while, hasn't it?

Shouldn't be too difficult, especially if we don't have to transform the listens in question in any way.

But I can see the increase in priority now that we have recs

In your opinion, should the listen be sent by BrainzPlayer as soon as it starts playing, or do we want to wait an arbitrary amount of time before sending it?

Something like 10 seconds?

It can be set like vlc plugin

Half of track duration

Or 4 min iirc

> In your opinion, should the listen be sent by BrainzPlayer as soon as it starts playing, or do we want to wait an arbitrary amount of time before sending it?

Mr_Monkey: arbitrary amount of time.

imo

I think Spotify considers a song listened to after 30sec, for reference

pristine___: I've moved the code in the closed pr over to https://github.com/metabrainz/listenbrainz-serv...

#1097 just needs some tests and css improvements and it should be good to go.

iliekcomputers: cool. Though a comment would have been great!

noted, will try to remember next time. it was wip, i didn't expect anyone to be monitoring it.

Thanks! Even if no one is monitoring, I guess a closing comment will be helpful for people passing by, if any. 🌹

[musicbrainz-server] reosarevok opened pull request #1719 (master…MBS-11136): MBS-11136: Support HTML in attribute descriptions https://github.com/metabrainz/musicbrainz-serve...

MBS-11136: Support HTML in attribute descriptions https://tickets.metabrainz.org/browse/MBS-11136

alastairp: Mr_Monkey https://docs.google.com/document/d/1QpggHRwFdOE...

ruaok: what's your punk tag idea?

ah, yes thanks.

bitmap, yvanzo: hmm. I was thinking about relationship gid_redirects :p

I wanted to get rid of instrument arranger and vocal arranger (there's no reason not to have those just be attributes of arranger)

Or well, at least once we merge instruments and vocals

Same with performer

But should we try and make it so that the old mbids still redirect to the main rel once that happens?

STYLE-932 is an already implemented variation of the same situation

STYLE-932: Widen ‘piano technician’ to ‘[instrument] technician’ https://tickets.metabrainz.org/browse/STYLE-932

(the previous rel still exists, although deprecated, so there's no need for a redirect *yet*)

that'd certainly be cool to have, though I'm not sure where the redirect would matter at present except for documentation

Given this should be a rare occurrence we could just do a redirect in ngnix, but

Well, if someone has, say, an old data dump, they could resolve what rel that is equivalent to now

IIUC, the question basically is how to merge relationship types?

(and yeah, see its documentation)

yvanzo: pretty much

nginx even

we could have a small struct in the code first and decide if we want to schema change later I suppose

since there will only be a few at most for the forseeable future

+1 as persistance is the main point of having MBIDs.

iliekcomputers: sign in with your meb account.

ah

got it

We just spent 90 minutes brainstorming and came up with a pile of stuff.

Fodder for future roadmaps.

Ok, I'll keep in mind for whenever we finally remove some deprecated rels :)

Also, do you feel like joining Mr_Monkey alastairp and myself for a hack weekend in Nov to implement playlist support in LB?

(playlist brainz but not tlp)