ah yes, this one is only used by compare_to_track()
amCap1712
ok got it
I feel now I have a basic understanding of how picard works
thanks zas
zas
all those comparaisons are quite tricky
amCap1712
yes
zas
and as you can see, we don't have proper tests for all of them
amCap1712
i will keep going through the code
one more thing
zas
good luck, this app has a loooong history
amCap1712
yeah it does, I found files which were coded by ruaok as well
ruaok
oh dear. really?
wow.
zas
ahah ;)
amCap1712
zas: do you think it is feasible to make tagger for only single tracks using lookup
within a month
i mean your advice on it on whats a feasible target from your experience
zas
it depends, if you are already able to extract metadata, and query musicbrainz+acoustid, you can have something working pretty quick, single tracks matching can be very easy (acoustid) or very hard (no metadata/no acoustid), but for files with a proper title/artist you can easily find matches
then you have to decide what's the "best match", Picard let the user choose in general
amCap1712
ok, i'll have a look on acoustId implementation next to get a better idea
currently i target to get both methods working for single tracks
can you point to me to how picard does the matches using acoust id
I'm still waiting for some error output from you. got any?
pristine--
2019-03-13 11:47:22 ERROR AsyncEventQueue:70 - Dropping event from queue appStatus. This likely means one of the listeners is too slow and cannot keep up with the rate at which tasks are being started by the scheduler.
Yesterday's error was solved by restarting. It was missingBlockException.
ruaok
yep, that seems sufficiently obscure...
pristine--
But AsyncEventQueue is weird because versions >2.3.0 should not face this
I will shoot an email then :)
ruaok
good plan
pristine--
I assume load_data would have taken a lot of time when run in past
ruaok
a while, yes.
pristine--
because of the rdd stuff. I ran a file today, simple queries took around 50 sec, and the one involving rdd took 240 sec.
ruaok
building stuff on the cluster was super slow.
the VMs that we use for this are not throughput guaranteed.
microsoft agreed to give use azure credits for hosting our own cluster.
let me ping them on that.
pristine--
cool :)
ruaok
but even here is a neat little lesson.
extra steps are a severe penalty
pristine--
Yes. Good Life Lessonsπ
reosarevok
"Move as little as needed"
Sounds about right
ruaok
everyone, especially iliekcomputers and zas: I'm going to try and mess with the socketio stuff on production for a bit.
expect noisy errors and such.
ferbncode joined the channel
silverstar
Hey guys, I had posted my SoC proposal in the forums (https://community.metabrainz.org/t/gsoc-2019-ga...), there are a couple of things that I need opinion/thoughts on. So I thought I'd post this here to gather some attention (especially the coins system that I mentioned in the proposal)
pristine--
ruaok: why do we want to build a recommendation system?
ruaok
a very good question.
most of the music services all live in silos. like spotify, apple and google. as long as you play in their silo, everything is ok.
but interoperability with others is out of the question.
so, if you're in the silo, all is good. but if you're not, you're screwed. you're never going to make any money as an artist.
and inside the silo, they recommend more stuff in the silo and more stuff the major labels want you to listen to.
not what *you* want to listen to.
that sucks.
building recommendation systems is hard work -- you need good data going in in order for them to work.
which is why these don't exist in open source for music.
pristine--
there must be other open source recommendation systems?
ruaok
thus, we're going to change that. we have the data, let's build recommendation systems, so that we can hopefully build more open music systems in the future.
there are none. that is the problem.
pristine--
so basically it will help promote all artists
amd not just a sect
and*
ruaok
that. exactly that.
especially artists who are not even in the west.
small artists the world over.
pristine--
okay. Good initiative :)
D4RK-PH0ENiX has quit
pulkit6559 joined the channel
adhawkins has quit
silverstar has quit
D4RK-PH0ENiX joined the channel
adhawkins joined the channel
yvanzo
samj1912: Have mb-rngpy and sir to use the same version of Python?