I haven’t received any mailed in reviews. Which means the usual suspects are up: lucifer, mayhem, reosarevok, bitmap, alastairp, yvanzo, akshaaatt, monkey, zas, CatQuest, Freso – anyone else who want to give review, let me know ASAP!
i mostly worked on migrating various parts of LB to use user ids instead of user names. we'll hopefully deploy it all this week.
also discussed some stuff about listen counts and timestamps with mayhem and worked on that.
other than that follow up on existing PRs. that's it for me.
CatQuest: BB is in node.js, LB in python so no code reuse possible there
last week I reviewed some things for lucifer - documentation stuff from the previous week + the user id stuff, and a bit of the timestamps things too
I did some further digging into automatic computation of the "mood" of audio recordings
although this turned into a few different discussions with mayhem and others about how we can use our commumity to help us get better annotations for content in MB without resorting to automatic classifiers
I also spent some time preparing classes and teaching them at the uni, so was away for a bit in the middle of the week for that
that's it for me, yvanzo next?
Thanks, hi all!
I mostly finalized scripts to repair indexing MB in Solr when broken.
Resent 880K search index requests to the indexer, most from previous failures in 2021.
Also reviewed a coupe of MBS PRs, including the spammer flag’s PR.
Fin. Go reosarevok?
(Still up: mayhem, bitmap, akshaaatt, monkey, zas, CatQuest, Freso – anyone else who want to give review, let me know ASAP!)
I mostly reviewed our Perl tests
And wrote docs for them
As in, added descriptions of what they actually do
Other than that, some code review and some finally fixing some small bugs that I hadn't found time to look into before the holidays
I've been working
this week I'll be working more!
today is both clothmass and chinese new years! 🎊
I'm also going to work on instrument documentation (with reosarevok)
finally I hope to add some more instruments
fin! next can be: monkey! 🐒
Hi everyone !
Last week I spent most of my time reviewing and merging PRs for BookBrainz and ListenBrainz.
There was (and still is!) a big backlog of PRs from the holidays that I hadn't got to yet, and a lot of general repo upkeep tasks.
Including fixing the automated linting setup for the github repo
I helped BenOckmore roll out an update from Boostrap version 3 to v4, now deployed on beta.bookbrainz.org
And as a follow-up, I'm currently reading v4 to v5 migration guides. I think we'll soon want to use the design-system package that akshaat is working on which uses Bootstrap v5.
I also helped Shubh with a couple of big impact projects for BookBrainz: notifications and import userscripts. The latter requires creating POST endpoints to pre-fill entities, as those don't exist in BB yet.
And that's most of it ! bitmap go !
last week I mostly worked on code review, plus fixing test failures and merge conflicts in open PRs
was trying to fix more test flakiness since circleci often fails for no reason, so pushed a couple changes for that that seem to help
other than that I was working on finishing some pgtap tests for the area_containment schema change I mentioned a couple weeks ago (that would improve perf of the area-related pages)
fin. mayhem go!
last week was finishing off more documentation and emails and some more peer learning network calls.
but then thursday evening inspiration struck and I realized we can use listens to calculate something akin to recording similarity. this factors into alastairp's update about using LB to generate some of the data we were hoping to product from AB.
and this realization, after doing about 5 hours of work and getting promising results, changes many things about how we should be moving forward with AB.
(Sorry, Ffx crashed. On mobile while I reboot.)
I hope that alastairp and I can chat about it more and that I can finish the proof of concept and tune the results a bit.
There's already a patch for that and deployed on beta now CatQuest :)
mayhem: let's do that chat tomorrow!
last week I worked on Picard patches & reviews, many good and big patches from phw (like support for long due eac logs)
I also upgraded discourse in multiple steps (I started on friday, but I delayed to today the rest, and I did well, it failed)
so we are now running discourse 2.8
plus usual tasks, edits, etc.. fin. akshaaatt ?
I enjoyed working last week since I got the design system up and sailing!
I think we have some discussions and work incoming for the design system.
I fixed open PRs in MB and LB, and removed the boilerplate code that my past self had added.
Some Android app work is always happening in the background as well.
I'm happy with the way February is looking!
That's it for me. fin!
Last week I started working on some community management docs, still a way to go on that.
There were a couple of other things too, but I can't access my time tracker rn. :/
I'll be getting my booster shot tomorrow (finally), so I'll take most of tomorrow and Wednesday off.
I haven't reacted too poorly to prior vaccines (covid or otherwise), so I don't expect to be out for long... But we'll see how I feel Wed/Thu/Fri.
I think that's all for reviews, right?
Thank you :)
And thank you all for your reviews!
nothing more for today, i think
No more items on tonight's agenda, so this also concluded tonight's meeting.
Thank you all for your time! Stay safe out there! 🙏
mayhem: after your work last week I've also been mulling over a bunch of things, especially around content-based computation vs user data-based and when each of them is useful
have you found useful cases for the content based approach is the not BPM?
and I was reading a bit about differences between "recommendation" vs "playlisting" (Ben F and Paul L did a well-known tutorial about this >10 years ago now - https://www.slideshare.net/BenFields/finding-a-...), I'm interested to see how this fits in with some of your ideas
I had a really quick look through some papers and other documents
one of the biggest drivers for content-based computation is "what do we do about the long tail?"
so, stuff that doesn't have a lot of people listening to it (because it's not popular, or because it was released yesterday)
ok, I can see that. deffno not low hanging fruit though
so a lot of the work on "content-based recommendation" was trying to take user-based data, which was known to be pretty good, and see if they could get similar results using only the audio content compared to the user-centric methods
which I think ties into a lot of this discussion about what we're collecting the data for, and what we want to do with it
I think it's correct to say that if we have a lot of user interaction data, for popular tracks, then that's definitely the way to go
agreed, that has become more than evident now.
I'd really like to be recommended less popular stuff too though :)
and if we want to try and work out how to do recommendation in the long tail then we have to look at a combination of things (there's been a lot of work in this area for years, since Oscar Celma's PhD), but I'm not sure what the current state of the art on that is
I think this popular vs long tail distinction is a good one for us to keep in mind.