True, but a more sophisticated DB structure almost certainly speeds up the SQL-like backends.
Normally one would design the DB to reflect the song structure.
Thus having one table with the “Song” related info, and a connecting table to albums and artists.
That way one could parse by albums, neglecting the artist join, and vice versa. Or just browse by songs.
Compilations would be a hybrid link table and join to songs, artists and albums.
Sure memory load would eventually become an issue, but then again each table would hold less data, effectively reducing memory load.
And a join with a single key over 2 tables really shouldn’t be that hard on the slug.
I agree with the GDBM thingy though.
Hm, would be nice to poke around with SQLite a bit on the slug and try to compare one approach vs the other…..
0.02€