mayhem: the recording similarity seems to work with window sizes as well. should i export and import it on wolf somehow or setup it be to inserted in main lb db/ts db?
mayhem
wolf would be good. but we need to get smarter about how we identify the runs of data that we create. last time I got lost in all the attempts. thoughts on how to do that?
lucifer
we could probably add a comment on the table in postgres, with the description of various parameters the data was generated on.
mayhem
or do we make a new dataset hoster query for each with the description properly filled out?
lucifer
we already had that last time, no?
mayhem
a bit more work, but then the end result is available to all
lucifer
specify the table name and it'll fetch the data from that table.
mayhem
yes, but we didn't have descriptions for each of the tables.
I suppose something like a google doc with table definitions and how they were generated could work
lucifer
i think we could make 2-3 fields in 1 query and name our tables in a way so that the table name can be generated from the params. then have a general description of various params on the same page.
mayhem
clever, I like it
lucifer
yes that'll work too. if we want to experiment with how we define sessions or other parts of the algorithm then it makes more sense to do it that way.
but if we are going to only change params then this table name idea should do
(params as in only changing threshold, lookahead, session duration but keeping the algorithm always fixed)
mayhem
then lets do the doc, because we will need to change the alg too
lucifer
makes sense
MRiddickW joined the channel
for sessions, we discussed 30 mins (configurable) of continuous listening iirc. i am wondering how to define it, is it the same as checking the difference between the timestamps of two listens is less than 30 mins?
mayhem
yes, that is a decent first order approximation.
a better one would be to take the duration of the last track into account.
lucifer
i see, we don't have it in spark dumps so we'll need to import MB recordings/release dump for that. let's start with the approximation then later improve.
I really appreciate how each album has its type shown. makes loads of sense.
chinmay
Yes, that was the last thing that I thought of while designing the release card
mayhem
I think "Beautify releases" -> "only show releases with coverart" would be more clear.
chinmay
Did you like the sidebar filters? No more dropdowns
mayhem
yes, but clearly it need an "all" option, no?
chinmay
Yes, the "all" option is the default option.
Like in the first mockup.
They'll only get filtered out as you start selecting the release types. In the 2nd shot, you can see that I have selected Albums and EPs only, and only that's filtered out as you can see in the release type chips of the cards..
Yes, the "beauty releases" -> "only show releases with cover arts" can be done
mayhem
ahh , I see, ok.
chinmay
Yeah, you can select and deselect multiple release types if you wish. And the "clear all" button on top will deselect all of the filters