[00:48] *** leah2 left
[02:25] *** hulk joined
[02:26] *** kylese left
[03:15] *** hulk left
[03:15] *** kylese joined
[03:34] <iwuvmika> do you know what a 'meme' is?

[03:39] <[Coke]> Who are you asking?

[03:39] <[Coke]> (me? yes)

[04:19] *** human-blip left
[04:20] *** stanrifkin_ joined
[04:23] *** stanrifkin left
[04:42] *** mantle joined
[05:40] <japhb> .oO( Two copies of a 'me' )

[06:20] *** hurufu joined
[06:57] *** hurufu left
[06:58] *** hurufu joined
[07:22] *** hurufu left
[07:24] *** hurufu joined
[07:49] *** leah2 joined
[08:00] *** samebchase left
[09:01] *** snonux left
[09:15] *** snonux joined
[09:53] *** Sgeo left
[11:10] *** mantle left
[11:15] *** mantle joined
[11:24] *** human-blip joined
[11:37] *** Sgeo joined
[12:06] *** samebchase joined
[12:08] *** Sgeo left
[12:28] *** belluzj joined
[13:06] *** belluzj left
[13:28] <Voldenet> meme i a funny picture you laugh at even despite laughing at it earlier?

[13:45] *** tgt joined
[13:50] *** tgt left
[13:50] *** tgt joined
[13:51] *** tgt left
[13:59] *** belluzj joined
[14:15] *** leah2 left
[14:29] *** leah2 joined
[14:56] *** hurufu left
[14:56] *** hurufu joined
[15:07] <disbot> <antononcube> Meme is "a unit of cultural transmission, or a unit of imitation."

[15:14] *** belluzj left
[15:33] *** hurufu left
[15:34] *** hurufu joined
[15:57] <Voldenet> I see

[16:07] <SmokeMachine> Now we have a Red driver for DuckDB! https://usercontent.irccloud-cdn.com/file/h6IYkNan/image.png

[16:23] <Voldenet> nice, how does it compare to Text::CSV?

[16:28] *** hurufu left
[16:32] <SmokeMachine> idk, I have not compared... but duckdb does much more than only CSV...

[16:33] <SmokeMachine> But comparing to Text::CSV, I suppose it will be much slower, because it has to create the model at read time...

[16:56] <Voldenet> I'll make some benchmarking myself because I'm curious

[16:57] <Voldenet> I'd guess that Text::CSV would be slower

[17:16] *** swaggboi left
[17:58] *** swaggboi joined
[19:04] *** belluzj joined
[19:17] *** hurufu joined
[19:28] *** hurufu left
[19:33] <Voldenet> hmmmm, DB::DuckDB raw is very fast, but it's very tricky to use Red in this case

[19:33] *** mantle left
[19:33] <Voldenet> since normally I'd use duckdb with `group by`

[19:34] *** mantle joined
[19:36] <Voldenet> though doing :materialize on read-csv makes very fast queries

[19:36] <Voldenet> it becomes 4min vs 1s comparison, which is pretty good

[19:51] <SmokeMachine> You can make group by with .classify

[19:55] *** Sgeo joined
[19:59] <Voldenet> here's what I did (iot-sensor-100.csv has 100 lines) https://i.tanomi.eu/cr1eg.txt and it takes a lot of time

[19:59] <Voldenet> …16s, which suggests a lot of work being done, potentially scanning the whole file 100 time

[20:03] *** mantle left
[20:40] *** leah2 left
[20:48] *** stanrifkin_ left
[20:55] *** leah2 joined
[20:57] <SmokeMachine> I just added $*RED-DEBUG to that Driver (already released)... here is what that's doing:  https://www.irccloud.com/pastebin/qRAzNtzl/

[21:11] *** belluzj left
[21:12] <Voldenet> yes, that's what I expected with that timings, good old n+1

[21:28] <Voldenet> I'm betting there's some form of classify I don't know

[22:02] <SmokeMachine> I’m still working on that… :)

[22:03] *** vasko4535586169 left
[22:04] *** vasko4535586169 joined
[22:35] <SmokeMachine> On my current tests, it's looking like this:  https://www.irccloud.com/pastebin/tGfP7d72/

[23:19] <Voldenet> I wonder if there's way to express average

[23:21] <Voldenet> but it looks a lot better, fits primary use for duckdb

[23:23] <Voldenet> and it looks a lot better than raw duckdb

[23:32] *** abraxxa left
