mayhem: another cause of slowdown, i noticed was Spotify api throwing errors. we keep retrying those with a backoff. retries last for about 2-3 minutes. blocking the processing. i guess we could create a separate thread for retries that may help if the errors are only thrown on specific ids. but in the long run, it might not matter anyway.
odnes joined the channel
odnes has quit
odnes joined the channel
trolley has quit
riksucks lucifer: I saw the review monkey had given last week in detail and realized that things would be based on MBID and not MSID. I believe it will be easier if we work on something that generates dummy MBID for tracks locally (to ease development). What do you think?
trolley joined the channel
riksucks: dummy MBIDs for local listens?
riksucks
yes
riksucks also I accidentally sent the above as an action, so my bad. But yes, dummy MBIDs generated for tracks + a small frontend that will only run in development mode where we can change MBIDs of tracks
Would you mind doing it? Just on my phone and my pw manager is being a poo
reosarevok
Ok!
aerozol
Thanks heaps
alastairp
Pratha-Fish: hi, that's great news. make sure you don't get too far behind in your class work though!
Pratha-Fish: what are you working on now?
Pratha-Fish
alastairp: yes, I'll make sure to not fall behind :)
reosarevok
mayhem: when you have some time, see the "Mb" email from Gary
Pratha-Fish
alastairp: I've completed a lot of the README, restructured the repo, and now I am currently working on the final script!
You might want to check the journal for more info
alastairp
ok thanks, I'll read it
Pratha-Fish
alastairp: Please also checkout the README, and let me know what changes should be made
agatzk has quit
agatzk joined the channel
reosarevok
lucifer: how does LB decide which artist to link to when there's several? On https://listenbrainz.org/user/reosarevok/ I see some multi-artist recordings that are correctly linked to the MB recording, which has the right artists, but for some reason they link to only one artist and it's not even always the first
lucifer
reosarevok: i think it always picks the first artist mbid in the listen.
reosarevok
Hmm, ok
lucifer
can you share the listen which was linked to some other artist than the first?
mayhem
moooin!
reosarevok: ok, will do.
reosarevok
mayhem: also, we need to look into those stupid papers again I guess :(
(and fedex them this time)
mayhem
yep. Go ahead and mail Amity and ask to start the process again.
reosarevok
Ok, done
kellnerd joined the channel
alastairp
the MTG is running a bunch of surveys for "Challenges and Opportunities in Music Tech", if anyone is interested in filling it out then please feel free to do so: https://www.upf.edu/web/mtg/challenges-and-oppo...
specifically, metabrainz people might have some feedback on Sound and Music Archives
Also For your comment on CB#464, I think we should not check both the database rn. Instead building a report would be the best solution as it would help to make the database much better.
[listenbrainz-server] 14amCap1712 merged pull request #2151 (03master…simplify-mb-metadata-cache): Remove redundant LEFT JOINs from mb_metadata_cache query https://github.com/metabrainz/listenbrainz-serv...
ansh
alastairp: I have also updated CB#401 to add endpoints all the entities and have added caching. Please let me know if there are any changes required for it
I'll be on MeB projects tomorrow and I'll look at these
ansh
Okay, thanks :)
BrainzGit
[listenbrainz-server] 14amCap1712 merged pull request #2143 (03master…improve-api-compat-tests): Use `wait_for_query_to_have_items` in api_compat tests https://github.com/metabrainz/listenbrainz-serv...
elgranRoble joined the channel
lucifer
reosarevok: oh. i think i know why sometimes it links to not the first artist. the artist mbids stored in the mapping tables do not account for position column of artist_credit table so they may often be in a random order and LB chooses the first artist mbid in the randomly ordered list.
*artist_credit_name table
since the mbid mapping is attached to the listen at page load time, once we fix the table the fix will propagate to all listens.
reosarevok
lucifer: what's the main reason not to link to all of them?
lucifer
reosarevok: its 1 combined name, we don't have access to all parts. so 1 link. (we could add parts for listens that are mapped by LB but that still wouldn't work user submitted listens)
reosarevok
Oh, ok. I was just thinking you could link to all the MB names, but I guess that's not worth the effort at the moment
Hi everyone, I'm out of action today, still reeling from food poisoning I caught yesterday :/
lucifer
take care monkey!
monkey
Cheers
elgranRoble has quit
Shubh
monkey: get well soon!
kellnerd has quit
agatzk has quit
agatzk joined the channel
bitmap
kellnerd: hi, regarding the ids in createRelationship, as long as they're negative and unique across the page it should be fine. if you want to avoid collisions, perhaps start from Number.MIN_SAFE_INTEGER and increment from there, or I can expose the id generating function if it's useful
the update-dialog-location parameters determine which button on the page the dialog points at/is positioned by, so the parameters have to be specified in a particular way
you'll need to also specify textPhrase if you want to open the dialog for a particular 'phrase group'
if you dispatch type-value and search-after-timeout with the same search term I think it should work, though
CatQuest
monkey: aw man food poisoning is nasty! my stern wishes for your fast improvement!
lucifer
mayhem: all ids obtained from listens are finished. looked at ids in MB, there are 45k that are still absent from cache. queuing those next.
mayhem
ok, cool.
I think we should also add metrics for the total number of artists and albums in the index.
all the data we have is relative right now.
lucifer
i see, makes sense. total albums is a straightforward count(*). artist count will be more work.
mayhem
if that is too much a pain, just albums would be good enough. I just want to know the overall completeness of it.
lucifer
don't think it'll be hard, will need to check how much time it takes. don't want to block the cache to submit metrics. alternative submit metrics from a different thread
kellnerd joined the channel
kellnerd
Hi bitmap, thanks for your answers!
I've implemented my own unique negative IDs for now as you suggested, but I guess this approach is prone to break when other userscripts also need to generate their own unique IDs in the same editing session. So it would be good to have the original function exposed in the long term.
bitmap
kellnerd: hey, no problem, I'll expose this function in the next set of patches I put up
kellnerd
Also thanks for the pointer regarding update-dialog-location, I simply misunderstood the purpose of these parameters. I thought I could save the additional calls to dispatch for 'update-target-type' and 'update-link-type' if I already pass these parameters to 'update-dialog-location' :D
Then I will simply leave them away, as I don't really care whether the dialog is positioned at the main "add relationship" button or at a certain phrase group's "add another X" button.
I've just tried your suggestion to combine type-value and search-after-timeout, now it works as expected (y)
Previously I had only tested both of them individually and either only the search field was filled but no search happened (as expected) or the search field was empty and the magnifying glass was replaced by the spinning icon but no dropdown appeared, although a search term was supplied. I thought type-value would be redundant if I'm using
search-after-timeout.
bitmap
kellnerd: actually you were right, it will also pre-fill the target type and link type if everything is specified correctly (in addition to determining which button to point at)
kellnerd
Hmm, then I will give it a second try :)
v6lur joined the channel
reosarevok
Freso: I'm cycling, please don't call me too early :) CatQuest wanted to go early iirc tho