Perl 6 - the future is here, just unevenly distributed

IRC log for #pdl, 2017-05-27

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

All times shown according to UTC.

Time Nick Message
03:08 _sfiguser joined #pdl
03:41 jberger Oh hahaha, sorry
03:41 jberger I was tinkering on my version
03:41 jberger http://pdl.perl.org/?page=develop
03:53 mohawk looking now
03:54 mohawk what am i looking for?
03:54 mohawk jberger, ^
11:12 _sfiguser joined #pdl
12:33 jberger This should be a small page that gets progressively bigger as you answer questions
12:33 jberger Right now, at best I see all the options laid out
12:33 jberger At worst there are floating boxes etc
14:25 _sfiguser hello all, given a matrix A let's say it is 3x4, how can i add a column vector composed by all ones  ? in octave i would do B = [ ones(3,1) A ]M
14:26 _sfiguser sorry  B = [ ones(3,1) A ];
14:26 _sfiguser how can i do this in PDL ?
15:31 chm joined #pdl
15:39 chm _sfiguser: See PDL::FAQ question 6.5
15:41 chm _sfiguser: I'm assuming you've read PDL::QuickStart and are learning the PDL documentation system (help and apropos in the PDL shells or pdldoc -h or pdldoc -a from the command line).
15:42 chm _sfiguser: You can look at PDL::FAQ with perldoc PDL::FAQ or pdldoc FAQ or help FAQ in the PDL shells...)
15:43 chm _sfiguser: If you have any suggestions on docs improvements to make the answer more "discoverable" please let us know...
15:44 chm _sfiguser: Alot of this can be explored from the PDL web site http://pdl.perl.org
15:44 chm _sfiguser: In particular the left sidebar.  For example: http://pdl.perl.org/?docs=QuickStart&title=PDL::QuickStart
15:45 chm _sfiguser: At the top is "See also: How do I search for a function?"
15:49 _sfiguser chm, yeah yeah i read it
15:49 _sfiguser http://pdl.perl.org/?docs=QuickStart&title=PDL::QuickStart
15:49 _sfiguser but couldn't find what i need
15:50 _sfiguser also chm i read the faq
15:50 _sfiguser and functions like glue and so on
15:50 _sfiguser work on rows
15:50 _sfiguser it seems
15:59 chm _sfiguser: hi, are you using the PDL shell?  It is a good way to experiment/explore PDL?
16:00 chm _sfiguser: I'm in the pdl2 shell and pdl> ?append gives "append two or more piddles by concatenating along their first dimensions" which would be rows
16:01 chm _sfiguser: pdl> ?cat gives "concatenate piddles to N+1 dimensional piddle" and more....
16:02 chm _sfiguser: ?glue gives "Glue two or more PDLs together along an arbitrary dimension (N-D append)" and more
16:02 chm _sfiguser: It seems that glue is what you are looking for.  The FAQ gave all three as possibilities.
16:03 chm _sfiguser: I guess my question now is why doesn't glue do what you need?  Here is where trying in the PDL shells is nice...
16:04 chm _sfiguser: The rows/cols between PDL and matlab/octave can definitely get confusing!
16:05 chm _sfiguser: In your initial question you have a 3x4 matrix, so that is a [4,3] shape pdl (3 rows, 4 cols)?
16:07 chm _sfiguser: Could you give an example of what you want in octave that we could convert to equivalent in PDL?
16:14 chm _sfiguser: Playing around in the pdl2 shell, it appears that if $A->shape is [4,3], then ones(1,3)->glue(0,$A) appears to do what you want...
16:15 chm _sfiguser: Hope this stream of thought helped.
16:15 _sfiguser chm, thanks... let me try
16:15 _sfiguser wait
16:16 chm _sfiguser: It seems to me that glue() is pretty close to what the octave cat() command does.
16:17 _sfiguser chm yes yes
16:17 _sfiguser thanks a lot for the help... i just tried now...
16:17 _sfiguser it is that i didn't understand how to use ones in that way
16:17 _sfiguser i mean... i didn't understand i had to do ones(1,3)->glue(0,$A)
16:17 mohawk "you've taken your first step into a larger world" :-)
16:17 _sfiguser i think it comes from my lack of knowledge of perl
16:17 _sfiguser hahaha thanks mohawk
16:18 _sfiguser thanks a lot chm
16:18 _sfiguser just a personal opinion... these days most of the people use scipy or numpy... do you think that in terms of performance they are much greater with respect to PDL or perform more or less the same ?
16:23 chm _sfiguser: I think from high level, all the matlab/octave, numpy/scipy, perl/PDL computation tools are roughly equivalent.
16:23 chm You can use them all to do similar types of things.
16:24 chm Alot of differences are in the user base and support environment.
16:25 chm Personally, I use PDL because I started with perl and numpy was barely functional at that time.
16:25 chm I would prefer to be able to use whichever tool works best for a given problem.
16:26 chm I would also like to collaborate with others without requiring that we all use the same tool(s).
16:27 chm Currently, most of the tools are stove piped in that there is not an easy way to interoperate with others.
16:29 chm One of the goals of PDL::NG ("Next Generation") is to provide a C function interface so that PDL could be used from any language that supports a C Foreign Function Interface---most all do.
16:29 chm Then I could use PDL from matlab or octave and call numpy routines from PDL,...
16:31 chm _sfiguser: good luck and happy PDL-ing!
16:33 _sfiguser ok thanks a lot chm
16:33 _sfiguser you have been very helpful
16:34 _sfiguser i'm actually using PDL for machine learning... if i will be proficient enough... i could write a module for Machine Learning for PERL
16:39 chm _sfiguser: The current PDL-2.x is the official release.  PDL::NG is to improve computation and usability.
16:40 chm _sfiguser: Please feel free to keep an eye out for any ideas for what could be improved for PDL to support ML.
16:47 mohawk _sfiguser, if you have ideas for how to do machine learning stuff and want to use PDL for that and/or make PDL-connected modules for it, come onto this channel and say :-)
17:40 _sfiguser ok thanks mohawk you even are a machine learning scientist ?
17:40 mohawk i am not
17:40 mohawk but there are some serious PDL experts here and it's a powerful system
19:16 _sfiguser joined #pdl
22:04 chm joined #pdl
22:12 chm Anyone interested to meetup for PDL during YAPC::NA 2017 this June?  https://sourceforge.net/p/pdl/mailman/message/35855482/

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