yvanzo: no visual (UI) differences, only that some artists which are not connected to writing at all get filtered out from the writers display. So "dedicated to" (for example) won't be shown in the Writers column in the works page nor under Writers in inline search :)
Once you approve, I will add tests and then we can head towards merging :)
antlarr has quit
antlarr joined the channel
ruaok
the delete_model function seems a bit odd. it deletes everything in that dir -- won't that nuke all the models?
pristine__: ^
pristine__
Right now we are saving only the current model in HDFS. When we will want to save more than one we can simply pass model_id to this function and delete that particular model
ruaok
I think it might be good to add that functionality now.
because when we get to the point where want to have more than one model, we need to code it, review it and merge it before we can.
pristine__
Okay
BestSteve has quit
BestSteve joined the channel
ruaok
but, lets get this PR approved, tests added, merged and then add a new PR for having more than one model, yes?
yvanzo
reosarevok: I don’t get why you don’t want this work to show up in there.
reosarevok
Because this is not a work by this artist, it's just marginally connected with this artist
yvanzo
I disagree, hiding linked data is not an improvement.
pristine__
ruaok: sounds good to me. Anyway I can add that feature while writing tests. Not much complex
reosarevok
I don't think anyone would expect it to be shown there (this wasn't even originally complained about by me, but in the forum somewhere by others)
yvanzo
reosarevok: Maybe we should wait for filters to be available.
ruaok
may not be complex, but it makes the PR even long and I have to review it yet again. I would prefer a separate PR.
reosarevok
Another option will be to have two versions of the page, a default one without this stuff, and one with all the cruft
yvanzo
No, I mean client-side filters.
reosarevok
I know, but that'd still show those by default, which is wrong
The data is not hidden, it's on the relationships page
that nickname makes it seem that last message very excited :D
ruaok
agreed.
BrainzGit
[bookbrainz-site] adithyaanilkumar opened pull request #325 (master…adiiiiiiiiiiiiii-contributors-edit): docs: Added Forking and Cloning to CONTRIBUTING.md https://github.com/bookbrainz/bookbrainz-site/p...
adiiiiiiiiiiiiii
I have added some changes to CONTRIBUTING.md as requested. Please verify it.
yvanzo: Thank you for the reply. I will look into the website.
iliekcomputers
#thisismylifenow
ruaok
i know the feeling. :)
alastairp
> 10:29 AM <ruaok> alastairp: would fixing the MLHD be one of those topics that would make an academic paper? all the work is practically done.
yes, sure. something like describing its issues, putting out the fix, and then doing an experiment that shows that the changes make it work much better
iliekcomputers
Noice!
ruaok
I mean I don't particularly care about academic papers, but if you think it would be worth doing that, and it is something you'd like to be involved in, then sure.
otherwise I'll fix the dataset and offer it back to Gabriel.
certainly use it for ourseves.
ourselves.
iliekcomputers
My contribution to academia is very negative rn, I'd like to get in on this
😂
alastairp
it's definitely something that could be written
ruaok
I thought you were trying to avoid academia.
alastairp: doesn't sound very definitive. :)
how about we fix the dataset, look at the popularity data and then see if we want to do a paper?
alastairp
just for an overview: the mbids in the dataset are bad? (because of lastfm badness?) and that means... that algorithms that use the data can't utilise all of its value?
ruaok
remember how last.fm never fully deconflated their artists?
alastairp
yeah
if you can concretely show that using the fixed version makes algorithm x work y% better, that's a paper that can be written in less than a week
ruaok
the artist popularity data iliekcomputers cacluated had obvious problems in like two or more artists called "muse"
iliekcomputers
ruaok: I'm trying to avoid mandatory academic obligations, to be accurate.
ruaok
iliekcomputers: got it.
alastairp: we should be able to quantify the results in the context of artist popularity, for sure.
not sure if that is sufficient.
alastairp
it could be interesting [ok, ok, in an academic context] to look and see if anyone else is using the data in that way
and show how their process is flawed because of x, y, z
sbvkrishna
adiiiiiiiiiiiiii: Hi ! we are currently trying to improve the documentation for new-comers at https://github.com/bookbrainz/bookbrainz-site/p... and if you are interested, you're welcome to give suggestions :)