I have dreamt of this for a long time. It is a minor optimization to be honest, but I previously had to choose between ugly code (what I had just prior to this) or looping through the data twice (slow). This parses the data and puts it into my intermediary type directly, along with relevant information about whetehr or not the scrobble is valid, and then I only need to loop over it once with "nice enough" code. There is still more I can do. My ultimate goal is to remove the looping entirely, and verify the data as it's being parsed, queuing up much smaller jobs that handle the individual entities (albums, artistsongs, etc.) as they come, rather than collecting it all and running it at once. I can get over the problem of needing to wit for all the LastFM responses that way as well. Furthermore, the data checking in the function expects a rather rigid structure that I'm not certain I can guarantee, but I'm pretty sure it'll stay that way. In any case, it may behoove me to make it more dynamic at some point. In any case, I am very excited about this change, and I hope I can continue improving upon it.
This was way easier than I expected, but I am rather unhappy with some things now. In particular, the GET page is pretty gross. I think there are some things I can do, but I'm not 100% confident. Maybe I'll bring some things up to bob once I have a better picture, but I really want to try to clean up my code
Album reviews would ideally allow you to rate tracks at the same time, so we'll have to work on that next. Also, disambiguation pages are becoming more and more necessary (Little Talks in inaccessible atm) Preferably, we start working on the `INDEX` for `/ratings` as well, and maybe use a unified language for these things (is it review, rating, rating_text, score,...?)
They don't exist yet, but I was trying to decide in my head if there was a meaningful difference between emrging two songs, and a SongGroup, and I decided they are indeed different, but really only in a small way
If no ratings are present, provide a textbox to make a rating. If a rating is present, show the rating. Eventually, there will be a button that allows an additional rating to be made, and the ability to delete ratings
This is actually fantastic, I'm really happy with how this has worked so far. My only concern for the future is how posting reviews from the `/ratings` path might work, since it's currently designed around posting reviews from the song page itself, but I think some HTMX and/or JS wil alleviate any problems I run into
This uses the zmpl data as a hash map to check if we've already checked the db for some song/album/artist/etc. and now only checks once per entity/assoc. table entry to speed things up. Previously, for each scrobble, we checked if its metadata appeared in the respective table, regardless of whether or not we've scrobbled that albumsong before. So, a song like Starless had to be checked (at the time of writing) ~180 times, but is now only checked once. Similarly, Wilco was checked ~3000+ times, as Hurry Up, We're Dreaming was cheked ~700 times.
The only problem now is the way it was implemented. Obviously, copying and pasting those huge chunks of code isn't very nice looking. ATM, I don't really care, and I'm more happy about the overall speed increase, as well as the readability increase of the job. However, I don't want to leave it like that. The way I see it, I have two options: either create a funcion which does this, or I can do something even better, which is create a jsonParse function, which, if my thought process works, would remove the need for an intermediary source type, meaning we no longer need to switch on that type, which means we can just have one for loop that does everything, which would mean we just need to have that code in one place.
Also not entirely happy with the code concerning all the conversions to i64s and []const u8s, but I think I have to.
Whether or not a song is covered, there was an original artist who originally performed the song. The only issue is that an Artistsongs table will almost be the exact same as the Albumsongsartists table, since most songs aren't covered. So, it may be better not to populate that table by default, and then if two albumsongs with different artists share the same song, then fill the Artistsongs table.
The birthday paradox is a real problem with the size of our datasets. i64 is the largest numerical value we can use, and there's a 0.1% chance of collision with ~2,000,000 values, so I feel pretty comfortable with this
This can (will) be easily replicated for the other views, I just first tested it on songs. I think this looks much nicer, and I'll probably roll with a layout similar to this for the other views, with some minor adjustments for each particular view.
This also makes the entity_info struct very similar to the UnifiedResult struct, so we'll probably see a merge at some point. Would be nice if we used the fields from the entity_info result more commonly.
One of the largest components that makes zuletzt unique - implementing groups the way MusicBrainz has (release groups in particular). I thought for a while that I would just connect songs via a shared ID, but for remixes and such, I don't think they should be so tightly coupled. This also gives the user freedom for how they want to do the grouping (a remix can be included in the group if they choose to, or it may not). This will allow someone to see a combined scrobble number for an album with, for example, a regular release, a deluxe release, and an anniversary release, in addition to the individual releases. This will complicate SQL queries rather significantly I imagine, and I'm not sure what the interface for creating/deleting groups will be (although it will likely be easier when I have full use of TS), but it's a necessity for the project.
Adds datestreak query, provides the number of songs/albums when relevant, and provides timescale with all years, regardless of the number of plays (defaults to 0)
If I can figure out a way to get an array of a union instead of a union of arrays, we're in business to make this even better, but this is fine right now. The inline else was just a dumb way to keep the for on the outside
