#metabrainz

/

      • alastairp
        it's always dns
      • 2022-12-14 34854, 2022

      • mayhem
        yeah, the "it takes way longer to connect, but it works" is always DNS
      • 2022-12-14 34821, 2022

      • mayhem
        lucifer: its not a troi problem as I thought: https://datasets.listenbrainz.org/top-discoveries…
      • 2022-12-14 34834, 2022

      • mayhem
        the track appears in the raw data. time to go look at the query again.
      • 2022-12-14 34847, 2022

      • lucifer
        sql query link?
      • 2022-12-14 34815, 2022

      • mayhem
      • 2022-12-14 34854, 2022

      • lucifer
        hmm, are you running this query on wolf or gaga?
      • 2022-12-14 34801, 2022

      • mayhem
        yes.
      • 2022-12-14 34806, 2022

      • mayhem
        wolf with tunnel to gaga.
      • 2022-12-14 34810, 2022

      • lucifer
        i see.
      • 2022-12-14 34810, 2022

      • mayhem
        might be a dataset hoster query issue.
      • 2022-12-14 34828, 2022

      • lucifer
        oh i know what it is.
      • 2022-12-14 34836, 2022

      • lucifer
        you are using user_name instead of user_id.
      • 2022-12-14 34841, 2022

      • lucifer
        i changed user_name this year
      • 2022-12-14 34841, 2022

      • lucifer
        replace user_name with user_id everywhere and it'll work fine i think.
      • 2022-12-14 34851, 2022

      • mayhem
        ok, let me try.
      • 2022-12-14 34821, 2022

      • lucifer
        LB will invoke troi patches directly this time so shouldn't be an issue to get user ids to it.
      • 2022-12-14 34851, 2022

      • mayhem
        for running yim that will invoke troi?
      • 2022-12-14 34824, 2022

      • lucifer
        yup. we have the necessary infra in LB and it seems less clunky than importing text files.
      • 2022-12-14 34800, 2022

      • mayhem
        great!
      • 2022-12-14 34835, 2022

      • mayhem
        do you know your user_id, lucifer ?
      • 2022-12-14 34848, 2022

      • lucifer
        5746
      • 2022-12-14 34818, 2022

      • lucifer
        the user id has created enough bugs in LB on its alone so i remember it distinctly :)
      • 2022-12-14 34849, 2022

      • mayhem
        mine is 3, so not hard to remember.
      • 2022-12-14 34858, 2022

      • mayhem
        only VA and deleted artist beat me to it.
      • 2022-12-14 34809, 2022

      • lucifer
        in LB, its 1.
      • 2022-12-14 34829, 2022

      • mayhem
        oh. ha. let me restart this.
      • 2022-12-14 34859, 2022

      • lucifer
      • 2022-12-14 34801, 2022

      • lucifer
        interesting!
      • 2022-12-14 34819, 2022

      • vibhoo_24 joined the channel
      • 2022-12-14 34834, 2022

      • vibhoo_24 has quit
      • 2022-12-14 34806, 2022

      • mayhem
        I like the smugmug guy -- glad he is doing this.
      • 2022-12-14 34822, 2022

      • lucifer
        zas: i figured some more things. connection fails on ipv6. psql tries ipv6 first, after a few retries gives up then connects on ipv4 and works. if i use ipv6 for direct connection to pink, it fails to pass firewall/.
      • 2022-12-14 34837, 2022

      • lucifer
      • 2022-12-14 34812, 2022

      • lucifer
        monkey: hi! someone is looking to build an open source book recommendation engine and is looking for ISBN data. how much relevant data would BB have in your opinion?
      • 2022-12-14 34834, 2022

      • mayhem
      • 2022-12-14 34837, 2022

      • mayhem
        better?
      • 2022-12-14 34844, 2022

      • lucifer
        yup looks great.
      • 2022-12-14 34859, 2022

      • lucifer
        (i was already browsing it before you shared the link :D)
      • 2022-12-14 34814, 2022

      • lucifer
        the issue is that my multiple top tracks are unmapped so missing from this
      • 2022-12-14 34828, 2022

      • monkey
        lucifer: Not significant. I see 6772 ISBNs in the database
      • 2022-12-14 34840, 2022

      • lucifer
        hmm but this one is not using MB data so those should be there actually
      • 2022-12-14 34845, 2022

      • mayhem
        sweet. let me fix the troi side of things and then update the playlist.
      • 2022-12-14 34848, 2022

      • zas
        lucifer: I'll add Ipv6 rules
      • 2022-12-14 34827, 2022

      • mayhem
        lucifer: this ignores unmapped tracks.
      • 2022-12-14 34837, 2022

      • lucifer
        ah ok.
      • 2022-12-14 34800, 2022

      • lucifer
        well, priority number 1 now is to add user generated mappings then :p
      • 2022-12-14 34813, 2022

      • mayhem
        getting there, certainly.
      • 2022-12-14 34849, 2022

      • lucifer
        i have some impl questions on it. let me ponder a bit and ping you again on that.
      • 2022-12-14 34824, 2022

      • mayhem
        k
      • 2022-12-14 34840, 2022

      • mayhem
        lucifer: for invoking YIM, can patches accept user_id on the command line or only user_name ?
      • 2022-12-14 34822, 2022

      • lucifer
        mayhem: i think user id because we dont have an api to convert user names to user ids.
      • 2022-12-14 34838, 2022

      • mayhem
        ok, let me change both over to user_id then
      • 2022-12-14 34850, 2022

      • lucifer
        👍
      • 2022-12-14 34822, 2022

      • mayhem
        hmmm.
      • 2022-12-14 34853, 2022

      • mayhem
        troi needs user names for desc/title. unless we take those out, but that makes the playlists harder to follow.
      • 2022-12-14 34821, 2022

      • mayhem
        the user_name column is still there, but might be inaccurate, correct?
      • 2022-12-14 34823, 2022

      • lucifer
        we can pass both to troi then
      • 2022-12-14 34837, 2022

      • mayhem
        a bit hacky, but ok.
      • 2022-12-14 34853, 2022

      • lucifer
        use user name for playlist desc but user id in query.
      • 2022-12-14 34802, 2022

      • mayhem nods
      • 2022-12-14 34817, 2022

      • mayhem
        can you please sanity check the top missed recordings query to see if we need to do the same user_name -> user_id switch?
      • 2022-12-14 34833, 2022

      • lucifer
        will do in a few mins
      • 2022-12-14 34819, 2022

      • bitmap
        lucifer: hi, do you still need a hand with the pg_hba.conf config?
      • 2022-12-14 34849, 2022

      • reosarevok
        bitmap: do I remember correctly that https://tickets.metabrainz.org/browse/MBS-12787 is a dupe of something you've already PRd?
      • 2022-12-14 34850, 2022

      • BrainzBot
        MBS-12787: Adding Instruments to Band Members Causes Duplication [Beta]
      • 2022-12-14 34845, 2022

      • Sophist_UK joined the channel
      • 2022-12-14 34848, 2022

      • bitmap
        not sure, I thought we already deployed all the attribute-related PRs to beta. I'll test it
      • 2022-12-14 34835, 2022

      • Sophist-UK has quit
      • 2022-12-14 34809, 2022

      • Sophist_UK has quit
      • 2022-12-14 34802, 2022

      • mayhem
      • 2022-12-14 34853, 2022

      • mayhem
        yeah, does look a lot better. ;)
      • 2022-12-14 34808, 2022

      • mayhem
        lucifer: I need to change the other query to user_id, no need to review the query
      • 2022-12-14 34823, 2022

      • lucifer
        mayhem: ah cool 👍. indeed looks better.
      • 2022-12-14 34827, 2022

      • lucifer
        is this sorted in any way?
      • 2022-12-14 34832, 2022

      • lucifer
        bitmap: yes and no.
      • 2022-12-14 34833, 2022

      • Sophist-UK joined the channel
      • 2022-12-14 34808, 2022

      • mayhem
        sorted feels dumb (because two tracks come back to back often if you listen to whole albums) so it is randomized.
      • 2022-12-14 34813, 2022

      • yvanzo
        zas, reosarevok, bitmap: What actually happened on Monday was about ~270 Mastodon instances trying to reach the same page in a range of ~5 minutes.
      • 2022-12-14 34818, 2022

      • mayhem
        but even that isn't great.
      • 2022-12-14 34821, 2022

      • lucifer
        makes sense.
      • 2022-12-14 34826, 2022

      • reosarevok
        Fun.
      • 2022-12-14 34800, 2022

      • mayhem
        oh fuck, mastondon as as a DDoS tool. what could possibly go wrong?
      • 2022-12-14 34813, 2022

      • alastairp
        in interesting
      • 2022-12-14 34829, 2022

      • alastairp
        since it's federated, every instance needs to get their own preview text?
      • 2022-12-14 34841, 2022

      • mayhem
        yep
      • 2022-12-14 34815, 2022

      • mayhem
        soon we'll see SRE's throttling mastondon instances and the mess starts unraveling.
      • 2022-12-14 34830, 2022

      • yvanzo
        There has been a separate similar request earlier so I guess that someone did the search and posted the link on Mastodon and their post got replicated among instances, then each instance tried to make an overview of the page. (I don’t know what Mastodon is doing about posted links really.)
      • 2022-12-14 34842, 2022

      • alastairp
        it's a shitty solution, but I wonder if we could do less sql queries on metadata scrapes (based on user agent sniffing)
      • 2022-12-14 34805, 2022

      • alastairp
        but yeah, that way leads to a slippery slope of configuration options
      • 2022-12-14 34810, 2022

      • lucifer
        270 requests to 1 page in ~5 mins should be fine in general though?
      • 2022-12-14 34817, 2022

      • yvanzo
        But it may also have revealed a potential issue on our side: Since it was an MB direct search, that is, it queries Postgres instead of Solr. I don’t if there is any cache to handle this kind of search.
      • 2022-12-14 34801, 2022

      • yvanzo
        lucifer: ~800 requests (each instance made up to 3 attempts) but yes, was my though too.
      • 2022-12-14 34816, 2022

      • lucifer
        ah ok, makes sense.
      • 2022-12-14 34833, 2022

      • alastairp
        what are you doing to poor gaga
      • 2022-12-14 34843, 2022

      • yvanzo
        bitmap, reosarevok: Do you think that caching direct search can be improved and/or that it should be unlinkable (e.g. using ws/js)?
      • 2022-12-14 34845, 2022

      • mayhem
        alastairp: that's me sorry.
      • 2022-12-14 34856, 2022

      • jivte joined the channel
      • 2022-12-14 34805, 2022

      • mayhem
        recalculating data for playlists for yim
      • 2022-12-14 34802, 2022

      • alastairp
        mayhem: I'm listening to my daily mix
      • 2022-12-14 34817, 2022

      • mayhem
        thanks!
      • 2022-12-14 34819, 2022

      • alastairp
        as expected, it's relatively well known tracks of "safe" artists around my tastes
      • 2022-12-14 34834, 2022

      • alastairp
        very good as a "turn on and put in the background" list
      • 2022-12-14 34838, 2022

      • mayhem
        thats the goal. easy listening for the post-muzak age.
      • 2022-12-14 34843, 2022

      • mayhem
        yay!
      • 2022-12-14 34852, 2022

      • mayhem
        !m team LB
      • 2022-12-14 34853, 2022

      • BrainzBot
        You're doing good work, team LB!
      • 2022-12-14 34803, 2022

      • alastairp
        so i"m really interested to see how it'll change as I work through the well-known songs
      • 2022-12-14 34816, 2022

      • mayhem
        got 4-5 weeks to find the answer?
      • 2022-12-14 34826, 2022

      • alastairp
        that's the plan
      • 2022-12-14 34811, 2022

      • jasje joined the channel
      • 2022-12-14 34851, 2022

      • reosarevok
        alastairp: my understanding is that it will start replacing them with Running up That Hill
      • 2022-12-14 34856, 2022

      • mayhem
        now if I could get zas to listen to daily jams for a week and tell me what he thinks....
      • 2022-12-14 34805, 2022

      • mayhem
        and on april 1st...
      • 2022-12-14 34812, 2022

      • monkey
        Oh yes
      • 2022-12-14 34816, 2022

      • alastairp
        reosarevok: tbh, I would't be surprised if it turns up on my list before long
      • 2022-12-14 34832, 2022

      • monkey
        Running Down That List
      • 2022-12-14 34837, 2022

      • mayhem
        rickroll has already shown up in my daily jams.
      • 2022-12-14 34838, 2022

      • alastairp
        mayhem: actually, I can see where that sqrt trick comes in handy
      • 2022-12-14 34850, 2022

      • mayhem
        already?
      • 2022-12-14 34801, 2022

      • reosarevok
        yvanzo: unlinkable as in, rather than have https://musicbrainz.org/search?query=test&typ…, you'd have a copyable URL that is always an indexed one and you can only pick Direct from there?
      • 2022-12-14 34802, 2022

      • alastairp
        especially for this kind of sudden-popularity-due-to-tv spike
      • 2022-12-14 34832, 2022

      • mayhem
        not 100% sure sqrt will help in this case.
      • 2022-12-14 34843, 2022

      • jasje
        hey lucifer! i wanted to ask you about to_ts(to_timestamp right?) what does the numbers mean (difference is 86400 ie day in seconds) but what is up with the extra stuff?
      • 2022-12-14 34855, 2022

      • alastairp
        is that for smoothing the number of listens of a user-track pair?
      • 2022-12-14 34857, 2022

      • yvanzo
        bitmap, reosarevok: I compared MB website logs with Postgres logs, it isn’t cached at all. 1min cache seems to be reasonable.
      • 2022-12-14 34809, 2022

      • mayhem
        yep
      • 2022-12-14 34814, 2022

      • jasje
      • 2022-12-14 34822, 2022

      • alastairp
        I guess I was kind of thinking of it in terms of the "beatles" issue
      • 2022-12-14 34829, 2022

      • alastairp
        if everyone listens to kate bush for 3 weeks
      • 2022-12-14 34836, 2022

      • alastairp
        then she becomes similar to everything
      • 2022-12-14 34841, 2022

      • bitmap
        yvanzo: 1m is probably fine
      • 2022-12-14 34847, 2022

      • mayhem
        yeah, but I think that is statistally unlikely to happen
      • 2022-12-14 34853, 2022

      • mayhem
        statistically, even.
      • 2022-12-14 34855, 2022

      • alastairp
        is there a way to surpress tat similarity a bit?
      • 2022-12-14 34800, 2022

      • alastairp
        sure, that's possible
      • 2022-12-14 34812, 2022

      • mayhem
        yes.
      • 2022-12-14 34813, 2022

      • monkey
        jasje: They're the timestamps for the beginning and end of a period (in this case for a day)
      • 2022-12-14 34820, 2022

      • mayhem
        I need to post the damn videos. I keep forgetting when I am home.
      • 2022-12-14 34820, 2022

      • alastairp
        though it's likely to be more of an issue the shorter our window is
      • 2022-12-14 34825, 2022

      • monkey
        We use those to render a graph
      • 2022-12-14 34841, 2022

      • yvanzo
        reosarevok: It was just an idea in the air, but it is probably much more complicated than caching it at first. Roughly: if the user selects the direct search option as a fallback from a search results page, then the search would be made through a Node component querying ws/js so that the "method=direct" is not used in the page URL.
      • 2022-12-14 34826, 2022

      • monkey
        Well, actually I don't think we directly use those timestamps to render the graph
      • 2022-12-14 34853, 2022

      • reosarevok
        yvanzo: I mean, we could just send everything via post :) But I guess we do want to still allow linking to searches
      • 2022-12-14 34805, 2022

      • reosarevok
        Testing a cache seems fine to me
      • 2022-12-14 34800, 2022

      • yvanzo
        reosarevok: Yes but direct search is only a fallback available when indexed search is broken.
      • 2022-12-14 34818, 2022

      • lucifer
        jasje: to_ts is ending timestamp of the period. from_ts starting timestamp. say you asked for this_week stats, from_ts will be timestamp of the start of the week and to_ts of the end of the period.
      • 2022-12-14 34816, 2022

      • jasje
        monkey: lucifer just for curiosity how do you determine the number assigned to first day of the year, i.e., 1st jan?
      • 2022-12-14 34842, 2022

      • lucifer
        jasje: the listens users submit to LB are timestamped (when they listened to a particular track is available in LB). so a simple grouping of each time period and count number of listens/timestamps in that period.
      • 2022-12-14 34824, 2022

      • jasje
        so are these timestamps same for every user?