#metabrainz

last week I helped zas deploy rakim. This didn't really require me to do very much at all, which is a good sign as it indicates the Ansible playbooks and roles are working consistently.

they do ;)

Nice!

I also added a new role to manage the Shorewall6 IPv6 firewall and deployed it on aphex, floyd, gaga and rakim

then I began the process of creating new metabrainz git repos for each role, complete with CI for lint and functional testing.

that process is still ongoing but I'm getting there.

CatQuest: go!

Meow! `😺`

Have been creating tickets for bookbrainz, and having many fruitful discussions with monkey and shubh about bookbrainz and tabviews and so on.

Have with help from reo, updated

ütőgardon https://beta.musicbrainz.org/instrument/1c3a672...

and Appalachian dulcimer https://beta.musicbrainz.org/instrument/e618d02...

as well as continuing work on recorders, pictures and adding aliases ♒

fin, go reosarevok

(People still up: lucifer, akshaaatt, alastairp, bitmap, mayhem, Freso – anyone else who wish to give review, let me know ASAP!)

I did a lot of test improving and documenting, finally merged the PR for porting edit lists to React, and worked on a few smaller things

Also started implementing small schema change tickets

(which I'll keep doing today / tomorrow)

That's about it. Freso?

Freso: aerozol is here!

🙋

atj: If they tell me they’re going to give review, I’ll add them. :)

morena!

ok, noted :)

Oh nope no review, just listening in :D

CatQuest: morena <3

Spent a bunch of time dealing with reported editors and also spent some time with forum admin’ing related to GSoC, such as hiding old applications and also updating the application template on the forum to reflect the newest one in the wiki/docs.

Other than that, being around and about etc.

fin.

lucifer: Go!

hi all!

In the first half of previous week, I worked with alastairp on upgrading our Postgres/Timescale installation on gaga. It was mostly smooth. After that, worked with mayhem on some tuning Collaborative filtering parameters for LB recommendations. other than that misc bugfixes and a couple of small UI PRs.

also, misc discussions with LB team on various upcoming features. fin. alastairp next?

hullo

I merged a few of Ansh's PRs to CB and deployed them, thanks Ansh!

I also worked on some upgrades and bug fixes in CB, during the process I found a few bits of templates that had stopped working due to backend changes so I fixed them again and also made some changes to BU MB database access to add some more data which is needed by CB and is currently missing

then as lucifer said, we upgraded postgres on gaga. it mostly involved waiting for things to finish and then either breathing a sigh of relief or panicking ;)

Hey mayhem, can I get some info on the upcoming ListenBrainz MLHD cleanup & Spotify Cache projects?

I've been working on some pretty similar stuff recently, so I'd love to dive into such projects at a larger scale :)

I learned a few things about the upgrade process that we can keep in mind in case we have to do this in the future

thanks also to atj for helping me find a fast compression tool!

akshaaatt: you're up

PrathameshG: sure, but lets us let the community meeting finish for now.

Hi everyone!

\o

Last week I continued work on the MusicBrainz revamp work and made quite some progress by fixing PRs and adding new sections.

(Only bitmap and mayhem left on my list. Last call for anyone else who wish to give review!)

On LB side of things, I continued work on the bootstrap v5 PR and made updates to get the changes made.

mayhem: sure, please ping me if possible once you're free!

I worked on implementing a new look to the android app to welcome Jetpack compose and introduce bottom navigation.

I discussed some new ideas for LB and MB Android App, most of which have now been added to the ideas page.

Also, I connected with prospective students over the week and guided them on getting ready for GSoC.

That's about it for me. fin!

Go mayhem!

hey!

last week was super hectic in real life and that too huge amounts of my time.

*took

that too took huge amounts of time

so, I did PRs and the usual background MeB work and other things...

I did create a postgres based fuzzy artist_credit recording lookup, which might be a good way to simplify our mapping. I'll continue on that.

🎉 real life

then I also worked with lucifer on tweaking our collaborative filtering and the results are much better already. there is much more to be done, but the results are no longer :facepalm:

fin.

bitmap: go

hey

last week I worked on migrating the redis store on yehudi to rakim, & implemented some changes to MBS to allow us to make the move without any downtime https://github.com/metabrainz/musicbrainz-serve...

this will also be useful if we want to move to a new redis/keydb cluster without downtime

bye aerozol

I also did a fair amount of code review and worked a bit more on the new artist credit editor code I mentioned last(?) week - still trying to integrate it properly in the current release editor so it keeps focus and doesn't flicker when you move between tracks, but javascript is being annoying

fin!

>javascript is being annoying

I was gonna say..

:D

:)

back to Freso I think

Yeah, no one else left on my list. :)

Thank you everyone who gave review!

And no more items on today’s agenda, so this also wraps up today’s meeting!

notbad.jpg

28 minutes, i think that's a new record

Thank you all for your time, and please remember to stay safe! 😷

</BANG>

Thanks Freso!

<3 everyone

Thanks !

Thank you!

zas: if you need any new roles for rudi, e.g. for managing haproxy then let me know

PrathameshG: Meeting’s over now. :)

thank you everyone

oh man we forgot a trick. tody is cetus

I'm guessing we'll need to create a least one new group

the one day that the starsign cetus is in the zodiac :D

Freso Appreciate the ping 👍

I'll be around tomorrow to discuss

oh, and I didn't mention that I recognise pi day under duress

wait pi day?

uh 14. 3 ?

at least pi approximation day writes the numbers correctly

that seems ackwards :D

and is accurate to the same number of decimal places

wich is pi approximation day

CatQuest: yeah, exactly. 3.14 for some people like bitmap

CatQuest: 22/7

:O cool

hmm.. 3,14159265358979 vs 3,14285714285714

didn't say it was a good approximation

heh :D

So, can someone kindly give me any details on the Listenbrainz projects that I mentioned above? (MLHD cleanup & Spotify Cache)

still find it weird that the 14th of the 3rds is somehow 3 14

I checked out this link, https://wiki.musicbrainz.org/Development/Summer... and the above 2 seemed the most appealing to me, especially since I am working on some very similar stuff right now :)

hi PrathameshG, what kind of things are you currently working on?

Hey alastairp, I am currently working on a project that aims to scrape last.fm data through APIs, clean it up, and then do some further interesting stuff with it by combining it with more metadata from the spotify API

well, that sounds very much like the MHLD projects.

oh neat. MHLD fits right in there, then - have you had a look at the dataset yet?

you know about that data, yes?

Sadly Lastfm only stores the first artist in the song, so a LOT of metadata is being missed out on at the moment. It doesn't even remotely relate to other rich data like production, composition, etc credits that could enrich the MusicBrainz DB as well.

TLDR, my plan mostly revolves around creating the best possible quality data with lastfm and spotify APIs and using it for my own analysis stuff, as well as using it to aid the musicbrainz DB

alastairp I tried taking a look at the data, but for some reasons I wasn't able to download it at all (some site issues ig)

mayhem yes, I have a bit of idea about the data as per the description on the MLHD site as well as some data dumps that you shared the last time I was active on this IRC (a few months ago)

PrathameshG34: ok, cool.

so, you know that the MHLD has some problems, yes?

Yep

namely last.fm's MBID mapping system was.... not very good.

we have two datasets that can help us:

Oh yes, I've noticed that a lot in my current project as well

1. canonical recordings: this identifies all of the recordings MB that are canonical. (the first release of a track, ignoring re-releases, compilations, etc).

2. recording redirects -- when a recording MBID is retired, we add it to a redirect table in MB.

so what we should do in this project is this:

1. Download a MLHD chunk.

2. Load it into spark.

3. Remove rows that are incomplete (e.g. no recording_mbid, toss it)

4. Check for redirects (was a recording mbid retired, if so, us the modern one)

5. Canonicalize the track. If it is not a canonical recording, replace the recording_mbid with the canonical one.

6. Compute artist and recording popularity (as a nifty by product of this page -- its a simple histogram operation)

7. Re-write new data chunks with the updated data.

Then we repeat this process for all the chunks.

Epic

Sounds extremely interesting, and very doable as well

and in the end we combine the artist/recording popularity data for each of the chunks.

ok, kewl.

now, we're not certain that this will work well.

it might work well to clean up the data. it might not.