[bookbrainz-site] 14tr1ten opened pull request #731 (03master…fix#BB-604): Fix[BB-604]: making datepicker and link button responisve on small screen devices. https://github.com/bookbrainz/bookbrainz-site/p...
hmm, submitting a listen to listenbrainz with "additional_info": {"artist_mbids": [""]} results in a different recording_msid being assigned to the listen vs if artist_mbids is omitted entirely. is this expected behavior? looking at the API code the empty string will be filtered out, but just the presence of an artist_mbids key seems to affect things. (the end result is that on the
listenbrainz website the listen doesn't have a link to the artist mbz page)
akshaaatt: that just looks like your DB is down - how do you usually connect to it?
santiagofn: ¡Hola! Hay cosas que hacer aún con React pero muchas son cambios enormes
Pero seguro que encontramos algo :)
Una opción es pensar en qué encuentras molesto al editar MusicBrainz, y mirar si es posible cambiarlo - así empecé yo, al menos :)
alastairp: I cc'd you on someone who wants to change their MB/LB username. Less than 1k listens, I understood for small numbers you had a simple way of doing it by hand or something?
lucifer
nbin: i checked and see what you mean, there are 2 related issues here. while "" is removed from the list of artist_mbids, the json retains the then empty list [] for artist_mbids. the frontend checks whether an artist_mbid was submitted by the user, since it was (even though its an empty list) it tries to use it. if that artist_mbids field wasn't submitted, it'll try to use a mapped one.
nbin: did more tests, we actually try to handle this case in submission already but the check is incomplete. i'll fix it.
nbin
cool, thanks. noticed the odd behavior with scrobbles submitted from a go-based client since it just marshals the artist_mbids to an empty string if it's empty. wanted to check before updating the code to omit the key
if the fix is in the submission i assume that means existing scrobbles will still retain the empty key?
lucifer
yes existing one will still retain the empty key. at least till we manually run a cleanup on the db.
we can also try to improve the frontend display to try use the mapped mbids if the user submitted empty ones.
nbin
gotcha. not a huge deal there. not sure if there's any other clients that would submit the empty list, but this one would've only started sending scrobbles within the last 2 weeks so there's probably not many
ruaok: one more please, 8f3471b5-7e6a-48da-86a9-c1c07a0f47ae ;)
ruaok
gawd damn you, lucifer.
good one!
!m lucifer
BrainzBot
You're doing good work, lucifer!
ruaok
or did you actually want that playlist??
lucifer
i guess doesn't hurt to have it. might come useful sometime 😈
ruaok
done. see my playlist page
lucifer
monkey: hi! the BP animation on hovering on the video is preventing me from cancelling ads when using youtube player. anything we can do here?
thanks!
ruaok
wow, track 2 on the GFCCSW. #thingsdutchieslove
monkey
Oh wow. Forgot about ads entirely.
Well, apart from suggesting a browser that blocks them i guess it'll have to wait until I finish the BP redesign that's in progress
lucifer
makes sense 👍
monkey
Maybe some CSS could help not crop the video, but it's a band-aid
akshaaatt
lucifer is it possible to provide other apps/websites the ability to integrate ListenBrainz to their project and query users based on similarity? Example, instagram providing a means to connect your lb account and providing user suggestions/lb stats to them on their platform?
I would assume it is possible to do so but can they use their exisiting database and just associate it to the lb username for gathering the results?
lucifer
sure almost everything on LB is available through an API, if you can do something on LB you can do that elsewhere too.
akshaaatt
Can this be a potential business model? I see dating apps connecting Spotify for showcasing a better profile
lucifer
re similariry, you can query top 25-50 similar users of a user because that's what LB stores itself. general similarity between every two users is not retained.
akshaaatt
But can't these apps be using LB for providing better user recommendations?
Okayy
lucifer
¯\_(ツ)_/¯, the data is there. we know its not always great but probably good enough. its upto one how they want to use it.
akshaaatt
I really believe it can be a hit in that market. I think what we can do from end is provide examples of how that could be done.
Our end*
lucifer
the yim stuff, top missed tracks are an example of combining various stuff and probably a good example.
ruaok: i see the playlists generated for the first mbid are both named gfccsw. probably a typo, do you know which one is which?
done. stupid terminal text wrap issues -- been problematic for 30 years now.
lucifer
lol thanks
nbin
is the "Recent listens" total cached? i just cleared my listens and it went down by a lot, but with no listens it still shows 4k there
akshaaatt
reosarevok: thanks for the heads up. Seemed to have been a problem with my setup. Working fine now 👌
nbin
i also noticed that it'd decrease when "deleting" a nonexistent listen as well
ruaok
nbin: yes, indeed, all cached.
lucifer
decreasing on deleting nonexistent listens is a bug. shouldn't be much difficult to fix.
nbin
gotcha. lmk if you want me to open an issue for it
lucifer
sure, a ticket sounds good
nbin
will do
lucifer
CatQuest: i was just able to import the scrobbles of last 3 months from your LFM to a dummy LB account. can you try importing again?
monkey: while looking at LFM importer, i see it is making 4 requests to LB per LFM page. seems a bit too much. can we reduce the number of requests to LB by making each payload heavier? i doubt one LFM is so large that we can't fit it in 1 request.
Still needs genre, similar artist or year filtering to produce listenable results.
lucifer
ruaok: how do troi, inputs and outputs work? does each element produce only 1 output and what's the number of inputs a element can accept?
ruaok
The list of inputs is an unordered list of types (in a list) it can accept.
The output lists what it actually outputs.
The whole system isn't fully hashed out yet, so changes might still come.
It is possible to return more than one item from an element, but it's never been used nor do I have a use case for it. Still uncertain.
Take the set functions for instance... They can take more than one input to combine data from multiple input streams.
I'm going to try to combine mfccsw results with gfccsw results to see if there is any overlap.
alastairp
ruaok: cool, interesting first step
you're right that it definitely needs another filtering step to pull together tracks that make sense
lucifer
ruaok: makes sense. thanks.
ruaok
alastairp: I was considering the typical test -- is this better than randomly picked tracks, but its clear that it is.
alastairp
it's interesting that there seems to be this rap/hiphop cluster in the mfccs one. so obviously there's some kind of similarity match going on there, but not exactly sure what - especially because of the others
so you can place things in 2-n dimensions instead of just 1
ruaok
interesting
alastairp
that is, is there a way of characterising this similarity in a way that makes it clearer _what_ is being made similar
he's moved on to more detailed similarity (deep learning, again...), so I'm not sure how much of what he's doing is directly applicable to our data, but...
lucifer: I'm not very familiar with the LastFm importer code, not sure what's making multiple queries. Can you expand on that, or provide some logs?
Ah, it also helps if I read your question right. So 4 queries to the LB API per LFM page.
Let me look at the code to see how we chunk things
Hmm, not sure how it's making 4 requests, I don't see that in the code
lucifer
monkey: i closed the site so don't have logs currently. but looking at code, i think it was a false alarm. i don't think we chunk at all. 4 api queries were probably submit listens, submit ts, retrieve ts.
monkey
Ah yes, here it is. `MAX_LISTEN_SIZE`
lucifer
oh right, the chunking code is in APIService not in LastFMImporter.
