Perl 6 - the future is here, just unevenly distributed

IRC log for #pdl, 2014-08-28

| Channels | #pdl index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
00:19 chm joined #pdl
00:20 chm sivoais, vicash: slice with a 0-D or 1-D piddle will dice the corresponding axis
00:21 chm $g = sequence(10,4); $keep = pdl(0..1,3..9); $g->slice($keep)
00:42 mohawk chm, if you're interested, i propose that more of the PDL "core cleanup" stuff get merged and dev-released
00:42 mohawk it would be vaguely achievable if i had push access to the PDL-code repo
00:47 sivoais chm: thanks, did not know that :-)
00:47 sivoais mohawk: release it piece-by-piece?
00:53 vicash thanks chm. i will try it out..
00:54 mohawk sivoais, at the moment it's stalled and the recent dev release change (with my humble contrib) feature code that's not in the repo at all
01:01 sivoais mmhmm
01:07 mohawk ?
01:10 sivoais !
01:11 mohawk answer the implied question, british-ish pig-dog
01:14 sivoais lol. More releases does sound good. Also, I noticed that BioPerl has been working on experimental future code in parallel with the mainline BioPerl: <https://github.com/cjfields/biome>.
01:15 sivoais that's the work they've been doing of moving BioPerl to Moose
01:16 sivoais oh, PerlRDF is also doing the same right now. <https://github.com/kasei/attean>. Also moving to Moose.
01:16 mohawk i want someone (possibly me) to get a grip of the code changes, with a bias towards getting them out in dev releases, and with approval from On High, released as 2.008, 009, etc
01:17 sivoais in my (not so important) opinion, I would like to see the same. The advantage of CPAN is that old versions can still be installed if you accidently upgrade.
01:18 sivoais but I haven't really contributed much to PDL yet, so.... haha
01:20 mohawk for now, there's a wasted resource - the CPAN smoke-test network
01:21 mohawk and the "wait until this giant change is done which i don't have time to work on" strategy is a missed opportunity to have small, incremental improvements
01:26 mohawk it seems to me there is a choice being made at the moment between the big improvement of PDL3, and effectively none at all
01:26 mohawk that seems to be a false choice
05:08 drrho joined #pdl
13:19 chm joined #pdl
13:21 chm mohawk, sivoais:  It would be great if you would participate in pdl-porters discussion: http://mailman.jach.hawaii.edu/pipermail//pdl-porters/2014-August/006362.html
13:22 chm That will keep the discussion and topics in thread for context.
13:24 * sivoais nods. I can at the end of this day. I've got a lot going on today.
13:25 chm This AM's discussion feels (to me) like "gossip" rather than constructive work to get PDL going...
13:27 sivoais yeah, that's why I said "in (my not so important) opinion". :-) I didn't feel like my thoughts carried enough weight without doing the work.
14:23 mohawk sivoais, have you heard from the alien guys?
15:01 vicash hi. how do i take a small pdl like [2, 3, 5, 13, 21, 34, 55, 89] and repeat it such that it becomes a pdl of size 100
15:01 vicash so that it looks like [2, 3, 5, 13, 21, 34, 55, 89, 2, 3, 5, 13, 21, 34 ... ]
15:01 vicash do i have to keep creating multiple objects and keep adding them until i get to the count or is there a smarter perl way to do that
15:13 vicash for now I am converting the PDL to an array, using a while loop to add the array to itself and then splicing it at the right point and converting it back to a PDL object
15:13 vicash that seems to work and is fast enough
15:14 vicash any other elegant ways are appreciated
16:39 sivoais hmm, I would do that with repmat() in MATLAB, but I'll have to think about how to do it with PDL & Perl
16:41 sivoais I think you can use dice() and a list repeat (0..12) x int(100/13)
16:41 mohawk i think the only tweak might be some kind of $newpdl = $oldpdl x 100/$oldpdl->size, instead of a for loop
16:41 mohawk snap!
16:41 mohawk ha ha!
16:42 mohawk and i seriously doubt "size" is a method, that was just for illustration
16:44 sivoais mohawk: heh, but PDL overloads the x operator to do matrix mult. :-P Useful in that context.
16:49 mohawk oh dear
16:50 mohawk fair enough
16:53 chm joined #pdl
16:54 chm vicash: The key is to use so-called "stupid indexing tricks"
16:55 chm To replicate a piddle along a dimension, add a dummy dimension the size of the multiple, then clump()
16:55 chm Threading takes care of the rest: $a = pdl [1, 2, 3]; $aaa = $a(,*3)->clump(0,1) => [1, 2, 3, 1, 2, 3, 1, 2, 3];
16:56 sivoais chm: I wanted to apologise for the "gossip" earlier. I'm trying to clear some time to help with the test coverage. I really would like to see PDL extended for more uses.
16:56 chm NOTE: I've used PDL::NiceSlice but ->slice() and/or ->dummy() can be used
16:57 vicash chm: let me try that out. that looks nice. i am using NiceSlice in my code anyway
16:58 sivoais I have been playing with how an API that allowed extending PDLs using Moo would look. I have made an enum/categorical variable PDL extension <https://github.com/zmughal/data-frame-experiment/blob/master/lib/PDL/Factor.pm>
16:58 chm You can implement repmat() that way.  On my nice-to-have list is a set of
16:58 chm MATLAB routines to simplify porting of MATLAB code to/from PDL syntax
16:59 sivoais chm: last time we talked about that, you were saying that you had some, but you had to get legal to clear releasing them?
16:59 sivoais iirc
17:10 sivoais here was the previous conversation <http://irclog.perlgeek.de/pdl/2013-12-10#i_7983053>
17:30 mohawk sivoais, do you have any idea when you'll be able to make tests as discussed there?
17:35 sivoais mohawk: I can probably start this weekend. Slowly, but surely. I have a number of tasks I have to work on for my research that take priority now. :-)
17:36 sivoais I will need lots of code review since I don't know all the PDL idioms.
18:57 mohawk grin
18:57 mohawk damn real life

| Channels | #pdl index | Today | | Search | Google Search | Plain-Text | summary