#metabrainz

nope. when I click on the track that was skipped it is fully playable.

oh

Mr_Monkey: how to import the lobes file you linked

It's already imported, you should just be able to use the variable in your less file

`NameError: variable @listenbrainz is undefined`

Hm.

Maybe something like `@import "./path/to/lobes/less/theme.less";

Not sure where lobes is in LB

I think `@import "./theme/theme.less";`

But… I don't see the @listenbrainz variable anywhere, oddly.

Not sure what the deal is

maybe lobes isnt in lb

nada

Hm. OK. Then ignore my remarks :)

We'll probably want to refactor that at some point to avoid having colors defined in multiple places.

colors.less?

Something like that yes

Imported at the very top of main.less

hm

I will make it in next PR

_lucifer: I split the search bug report since not all issues can be addressed at once, can you please make your PRs/commits match the new tickets?

will do yvanzo

yvanzo: saw your comment about gender-id. what would be the process to add that to indexing and are there any deployment concerns around that ?

alastairp: ping

hey

5 minutes?

sure!

_lucifer: there is no deployment concern afaik but I just cannot test integration without changes to the indexer.

ah ok! i saw you already assigned that to yourself. thanks!

_lucifer: I'm here

hi!

what were we talking about? stats and graphs on AB?

yes!

cool

let me pull up some stuff

I think I showed you this, right? https://github.com/MTG/acousticbrainz-labs/tree...

yes, right

so one thing that we're trying to do is make the site look interesting

personally, I'd love to see this data update in real-time

so the question to answer is to work out what graphs describe the data in the most interesting way

interesting question

you'll see things like https://github.com/MTG/acousticbrainz-labs/blob... are pretty terrible

I think this one is interesting https://github.com/MTG/acousticbrainz-labs/blob...

we don't want to show most of these, because they don't make any sense

yeah, I like the features/year and features/genre one

it's a lot better to do more objective graphs - features and years are pretty good

whereas if we start showing genre graphs and say "all music falls into one of these 8 categories", I'm sure that people will start complaining :)

right, makes sense

we currently do not have a pipeline to create these graphs right?

no

well, we have the code used in these notebooks to generate the graphs

right, we need to integrate these with ab database

however, we also have this: https://github.com/metabrainz/acousticbrainz-se...

by real time you mean like updating whenever a recording is submitted ?

the problem with integration into the database is that it's too slow to query all of the data, even if we did it periodically

right, perhaps not that often, but say once a week

that's doable

we have much of this information in the `similarity.similarity` table

and it's much smaller than the lowlevel table

that's nice!

so perhaps we could have a periodic task that we run that summarises this table

if not, we could definitely also create another statistics table, although there is a question about what data we should add there

we could make some initial tables, and load data, and then if we need more data for more graphs, we add those at a later stage

for example - the similarity table doesn't have years, so we'd have to get that separately

then say for example we wanted to compare year to loudness, we'd need some kind of table that allowed us to join this info together

the genre or mood tables are much easier, because we just need categories and counts

+1

let's focus on the following charts:

genre rosamerica, feature/genre, feature/year, key estimation, genre mood (at the end of mood)

awesome!

on bono, you can `psql -U acousticbrainz acousticbrainz_big` (not inside docker)

that has a full lowlevel table, and full `similarity.similarity` table

I think we should create a new postgresql schema (call it `statistics`), and for each graph, make a new table in this schema that stores just the information that we need for that graph

yeah that's a great place to start

then we can see how easy it is to 1) get the data from similarity.similarity, or 2) get the data from lowlevel as the data comes in

okay, will be needing to use saprk ?

I don't think so

this isn't really analysis, it's just loading and transforming data

if we wanted to use it, we'd have to load all of the necessary data into hdfs, which I suspect would be really annoying

okay, yeah right. the similarity table is smaller and we can probably process it directly

that's what I'm hoping

we can use spark without hdfs but that's a thing to consider for afterwards

oh? how would that work?

in some cases it might make sense to use spark for machine learning in AB, we should look into it as future option

> Access data in HDFS, Alluxio, Apache Cassandra, Apache HBase, Apache Hive, and hundreds of other data sources.

Spark home page says this

i had also read an article on the same but cannot find it right now

this one sums it up

PostgreSQL has provides a JDBC plugin to allow spark to connect to it directly

[mb-solr] yvanzo merged pull request #39 (master…SEARCH-611): SEARCH-628: 'primary-type-id' field is missing from JSON release group search results https://github.com/metabrainz/mb-solr/pull/39

SEARCH-611: Incorrect content in JSON version of release group search result https://tickets.metabrainz.org/browse/SEARCH-611

SEARCH-628: 'primary-type-id' field is missing from JSON release group search results https://tickets.metabrainz.org/browse/SEARCH-628

alastairp: by the way, why do we use HDFS ?

_lucifer: can you please update #43 too?

yes, yvanzo i am just testing it locally and will push the changes soon

btw, status-id change requires indexer changes too. I will update ticket accordingly and work on sir.

oh ok! thanks

I’m reviewing oxml cleanup and Java 11 preps next :)

Great! :D

_lucifer: did you use specific commands for #42?

yvanzo: no, why?

It could have helped with rebasing.

"Auto format" sounds like something automated though :)

oh! that, i just had my ide format that file to 4 space indent

rest is poor choice of words 😅

can you advise how I could have made rebasing easier?

If that was a command in the commit message, one would just have to run it again.

oh! makes sense. that i can do with the ide again. i'll rebase and drop the existing commit.

the clustering is the one that will have to be done manually and take some time

Java 11 PR looks good overall, but there are a few deployment concerns, will not merge for the upcoming release.

👍

Can you also remove unrelated commits from #37? (since they have been copied to separate PRs)

yeah sure

it was bound to happen. my own code (and pristine___s) rickrolled me.

lol 😂😂

ruaok: welcome to kapilaland

I actually think this is an evil plot by pristine___ to get back at me

Hey quick question. I currently run a musicbrainz slave server via the docker image. Is there a way to turn off indexing completely so all local queries go directly to the database?

ruaok: save yourself

K says: "Never gonna give listenbrainz up, never gonna let listenbrainz down, never gonna turn around and hurt listenbrainz!"

yvanzo: ^^ see JoshDi's query

Hey

I find even with SIR tweaks, live indexing daily updates of the slave , take like 12 hours to finish. When full reindexing takes 3 hrs

ruaok: I introduced a lil of troi to ppl

And they were like damn. Dynamic playlists

I only use this server for some local processes so its not like my musicbrainz server is very busy.

They felt really excited

Any ideas?

I dont know, but yvanzo will. hang tight for him to return and he'll sort you out. (he is around)

[solr]uri = http://search:8983/solrbatch_size = 5000[sir]import_threads = 10index_limit = 2000000live_index_batch_size = 5000process_delay = 15query_batch_size = 20000wscompat = onprefetch_count = 2000

Im now running this on a machine with 128gb of ram and 40 threads at 3.1ghz.... so it should be much faster

ruaok: can I ask an off topic doubt

are you giving JMVs ram to use? by default they may not be using enough, making things slower

shivam-kapila: why do you keep asking if you can ask a question?

Anyways do we have ryzen processors in prod?

my memory settings are: shm_size: 4g and SOLR_HEAP=4g

shivam-kapila: yes

JoshDi13: why not try 8g and see what happens?

postgres -c "shared_buffers=4GB" -c "work_mem=128MB" -c "shared_preload_libraries=pg_amqp.so"