2006-05-10 (15:28 EDT)

20.28.00 #       <Falco98> is anyone here really familiar with the algorithm by which "insert shuffled" (for, let's say, recursive folders) works?
20.28.41 #       <Falco98> i'm poking around in playlist.c, but the code is a bit too obfuscated [and my C skills likewise rusty] to get it well
20.28.59 Join    obo [0] (n=obo@82-46-82-224.cable.ubr02.trow.blueyonder.co.uk)
20.29.17 #       <hardeep> Falco98: sure, what's your question?
20.29.50 Quit    thegeek (Read error: 113 (No route to host))
20.30.50 Join    Rob2222 [0] (n=Miranda@ACB24D7B.ipt.aol.com)
20.31.39 #       <Falco98> well in a perfect world, i'd like to code up an additional folder context-menu item called something like "insert shuffled folders", etc
20.32.10 #       <Falco98> where instead of shuffling all songs, it only shuffles the order in which folders are added, but then the tracks therein will play in order
20.32.27 #       <Falco98> (basically accomplishing the "album shuffle mode" that a few people are clamoring for)
20.32.48 #       <Falco98> but i'm afraid at this point my programming skills might not be up-to-task
20.33.29 #       <Falco98> anyway, i've been reading the code, and i can't figure out how the "insert shuffled" command picks what position each track is inserted
20.33.31 #       <hardeep> Falco98: it would be a little messy because of the recursion... we don't have a complete list of all the subdirectories available
20.33.43 #       <Falco98> yeah
20.33.54 #       <Falco98> that seems to be the biggest obstacle
20.33.56 #       <hardeep> insert shuffled inserts the next track at some point between current playing track and end of playlist
20.34.04 #       <hardeep> but that's not what you'd want to use
20.34.22 #       <hardeep> instead, take a look at playlist_insert_directory()... that's where you'd want to make the changes you suggest
20.34.29 #       <Falco98> how about when you "insert shuffled" on an entire file tree?
20.35.00 #       <hardeep> same thing, it inserts randomly between 1 and N where N is number of tracks
20.35.03 #       <Falco98> i'm already in "add_directory_to_playlist", which seems to be the true recursive function
20.35.08 #       <Falco98> what, it inserts in-order?
20.35.23 #       <hardeep> insert what in-order?
20.35.34 #       <hardeep> it goes through the directories in order and inserts each track randomly
20.36.07 #       <Falco98> ah
20.36.26 #       <Falco98> for some reason i thought if i did "insert shuffled" on my root music folder, it would basically be a random-all
20.36.27 #       <Falco98> no?
20.36.28 #       <hardeep> hmmm, there might be an optimization I can do there... if we're inserting shuffled, no reason to sort the tracks
20.36.53 #       <hardeep> Falco98: yeah, it would be a random-all
20.37.14 #       <hardeep> since the tracks are being inserted randomly
20.37.16 #       <Falco98> oh.. it steps through every track (recursively) inserting *that track* into a random position in the playlist-so-far?
20.37.30 #       <hardeep> yes
20.37.33 #       <Falco98> ahh
20.37.36 #       <Falco98> that makes sense
20.37.40 #       <Falco98> hmm
20.38.05 #       <hardeep> if directory cache is enabled, you may be able to get a list of all subdirectories
20.38.18 #       <Falco98> i wonder if it would be possible to do that kind of thing for "random folder insert".. it would only need to keep index positions for where each "album" ends
20.39.12 #       <Falco98> how much more code would that take, do you think?  having it create a temporary index list, indexing the album start/end positions for the playlist it's generating?
20.39.13 #       <hardeep> yeah, that may work
20.39.31 #       <Falco98> then when it goes to insert the next folder, it picks one of those positions at random and inserts the whole folder in order...
20.39.56 #       <hardeep> hmmm
20.40.51 #       <Falco98> that would completely negate the need for any current directory listing
20.41.14 #       <Falco98> and wouldn't put any (or much?) more stress on the system than the current "insert shuffled"
20.41.15 Quit    damaki_ (Read error: 110 (Connection timed out))
20.41.38 #       <hardeep> i don't think it would be all that hard to add actually
20.42.04 #       <hardeep> just thinking if there are any corner cases you'd need to watch out for
20.42.21 #       <Falco98> maybe some really weird ones
20.42.34 #       <Falco98> but as long as the "index" idea handles "first" and "last" cases well..
20.42.55 #       <hardeep> easiest way would be to limit the number of directories to a few hundred so that you could just use the stack for storage
20.43.14 #       <Falco98> the big problem i can think of is it would require fucking with the current way the recursive function determines a random position.. but then again, i'm new enough to this code so it may be easier than i think
20.43.21 #       <Falco98> yeah
20.43.33 #       <Falco98> well the H140 says it only handles 2000 folders anyway
20.43.37 #       <hardeep> you wouldn't want to use the insert_shuffled option
20.43.45 #       <Paul_The_Nerd> You don't even need to save end-indexes for albums
20.43.46 #       <Falco98> and the "index" would only thus be 200 or so integers
20.43.48 #       <Paul_The_Nerd> Just star-indexes
20.43.50 #       <Paul_The_Nerd> start
20.43.54 #       <Falco98> yeah
20.43.58 #       <Paul_The_Nerd> Then, when randomly inserting, you do it at "random start-1"
20.44.09 #       <hardeep> instead, you would specify an exact position
20.44.14 #       <Falco98> well when i mention start-and-end i am only being complete :-P
20.45.00 #       <Paul_The_Nerd> But all ends are going to be start-1 of the next one anyway. Save on a bit of storage. ;-) The only problem case there is if it's the start of the first album, or if you want to place it after the last one.
20.45.10 #       <Falco98> indeed
20.45.40 #       <Falco98> really all you need is a list of numbers separated by .. arbitrary non-integer (whitespace, newline, comma)
20.45.51 #       <Falco98> 0,12,25,etc
20.45.54 #       <hardeep> i think an array would be easier
20.46.14 #       <Falco98> oh i guess the index list would change every time something is inserted in the middle though
20.46.29 #       <hardeep> keep track what albums have been inserted where
20.46.30 #       <Falco98> yeah an array makes more sense
20.46.32 #       <klrspz> ahhh
20.46.37 #       <klrspz> my settings were wiped :(
20.46.40 #       <hardeep> and then randomly choose a "slot" for the next album
20.46.50 #       <Falco98> at this point i'm still thinking visually instead of code :-P
20.46.56 #       <hardeep> (somewhere between beginning and last album)
20.47.35 #       <Falco98> yeah
20.47.54 #       <Falco98> so what would the array contain.. just the number of tracks on each album?
20.48.23 #       <hardeep> index where either the first or last track of the album is located
20.48.28 #       <hardeep> in the playlist
20.48.40 #       <Falco98> and when you choose a "slot", you tell it to insert the tracks at array[0] + array[1] .... +array [i]
20.48.54 #       <klrspz> does the latest CVS build wipe configs?
20.48.56 #       <Falco98> hmm
20.49.06 #       <hardeep> no, it would be array[slot]+1, array[slot]+2, etc
20.49.23 #       <Falco98> klrspz: i think so.. you should save your settings by doing "write .cfg" before you update to latest build
20.49.43 #       <klrspz> yeah, i had an older one luckily.. but stil gotta make some changes... oh well.. 
20.49.55 #       <Falco98> hardeep: but then when a new album is inserted somewhere in the middle, wouldn't all the other array members need to be shifted?
20.50.12 #       <hardeep> Falco98: yes, that's the idea
20.50.47 Join    Bagder [0] (n=daniel@1-1-5-26a.hud.sth.bostream.se)
20.51.02 #       <hardeep> keeping a count wouldn't save you much, either way you would need to count all elements
20.51.08 #       <hardeep> both ways work fine though
20.51.10 #       <Falco98> wouldn't it be less processor intensive to have the array value be the number of tracks in that album, and do a summation of 0 -> i to get the index value for insertion?
20.51.17 #       <Falco98> yeah
20.51.22 #       <Falco98> i suppose you're right
20.51.28 #       <Falco98> hmm
20.51.37 #       <Falco98> i'm thinking of putting a bounty on this :-P
20.51.50 #       <Falco98> since i don't think i could ever quite handle the programming job myself
20.52.12 Join    DrMoos [0] (i=DrMoos@m75.net81-66-158.noos.fr)
20.52.54 #       <Falco98> BTW i've submitted it to flyspray, an hour or so ago
20.52.55 #       <Falco98> http://www.rockbox.org/tracker/task/5338
20.53.01 Quit    Moos (Read error: 104 (Connection reset by peer))
20.53.08 Nick    DrMoos is now known as Moos (i=DrMoos@m75.net81-66-158.noos.fr)
20.53.41 #       <Falco98> ooh someone has commented too :-D
20.54.52 #       <Falco98> Falco98: testing
20.57.53 #       <Falco98> hardeep: do you think you could take a stab at it?  i'd be unbelievably grateful
20.58.19 Join    mirak [0] (n=mirak@AAubervilliers-152-1-3-120.w82-121.abo.wanadoo.fr)
20.59.11 #       <hardeep> not right now but maybe after the 3.0 release
20.59.52 #       <Falco98> cool