Camelia, the Perl 6 bug

IRC log for #pdl, 2014-07-04

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

All times shown according to UTC.

Time Nick Message
13:01 mohawk joined #pdl
19:46 chm joined #pdl
19:47 chm mohawk: How's the G::P announcement coming along?
19:55 sivoais chm: is the mediawiki for PDL gone?
19:57 sivoais ah, it looks like all the pages got transfered to the Sourceforge wiki
19:59 sivoais chm: I noticed a while back that one of the GSoC ideas was to interface with numpy <The test fails because `overload::Method()` returns `undef`. This is not the case
19:59 sivoais with any of the other PDL overloaded operators nor do other objects that
19:59 sivoais stringify exhibit the same behaviour.
19:59 sivoais woops :-P
19:59 sivoais darn chrome with it's non-X11 selection clipboard
20:00 sivoais <https://web.archive.org/web/2013050812304​4/http://sourceforge.net/apps/mediawiki/p​dl/index.php?title=GSoC_2012_Ideas#Python.2FNumpy_.2B_Perl.2FPDL>
20:01 sivoais so the idea is to interface with numpy... I was wondering if it would also be a good idea to try to interface with Octave or MATLAB at the C level
20:01 sivoais I use MATLAB in my research way more than I use PDL or numpy
20:03 chm sivoais: Media wiki is no more.  With the sf.net upgrade to allura platform, they got native wiki support.
20:04 chm I would like to see a Mex interface to MATLAB for PDL.
20:04 chm Since matlab already has a perl command, it is just a crippled one.
20:05 sivoais yes, all MATLAB installs (even on Windows) come with a stripped down Perl
20:05 chm If we were to implement a mex bi-directional connection (0MQ?) then matlab
20:05 chm could call pdl routines directly and vice versa.  For me it would be nice to not
20:05 sivoais ah, 0MQ is the approach taken by some Python and Julia MATLAB bridges
20:06 chm have to translate simple pdl code into messy matlab code to work with colleagues.
20:06 chm 0MQ is apparently platform agnostic and the work to make iPDL based on iPython uses that.
20:06 sivoais yeah, bsxfun isn't enough to do threading
20:07 chm The good news is that since mex files can pass matlab arrays as pointers to memory,
20:07 chm the perforrmance should be pretty good---no copying needed and PDL has no problem working
20:07 chm with wierd matlab dimension orders.  :-)
20:07 sivoais ah yes, I forgot about that :-P
20:08 chm However, since I'm not a power matlab user and PDL still needs work for
20:09 chm 64bit index support so PDL2 can be put into maintenance mode, I don't
20:09 chm think I'll be getting to the matlab<->pdl link any time soon.  Should be straightforward
20:09 chm to someone who knows matlab and PDL.
20:10 chm Back to your original query and interfacing with NumPy:  the idea there was to give PDL matplot lib
20:10 sivoais I might try to tackle the MATLAB in a bit. Just need to clear out some time.
20:10 chm access which is arguably one of the key features of NumPy.  If PDL had it, then I think a lot of Matlab
20:10 chm users would be interested in trying PDL.
20:11 chm What does "+++ sivoais has given op to chm" mean?
20:12 sivoais oh, that's just I added the operator permissions to your nick
20:13 chm Ok, whatever those are.  :-)
20:14 chm Regarding MATLAB on windows: I think it should be possible to make a PAR packed mex+PDL that could
20:14 chm install into the captive perl.  Of course, if admin is not a problem then one could just substitute in the path(s)
20:16 chm sivoais: What is the issue you are seeing with the overload operator and PDL?
20:17 chm Is it a new feature/bug?  Sometimes new perls change how old stuff works (breaks...)
20:17 sivoais If I try to retrieve the stringification coderef, Perl doesn't know about it via the overload::Method() call
20:17 sivoais I tried it in v5.16, v5.18, and v5.20. Same results
20:18 chm What should the result be?
20:18 sivoais it should return \&PDL::Core::string
20:19 sivoais all the other operators that are overloaded from PDL/Core.pm work fine
20:22 sivoais I can work around it, but it would be nice to have consistency. I'm experimenting with keeping track of units when performing operations.
20:23 chm I think it is because the stringify routine is actually PDL::string and is defined in Core.pm
20:23 chm Then PDL::Core::string is aliased to that---it appears.
20:23 sivoais ah
20:24 chm Very confusing and I'm not sure why it is that way.  It is possible that it has been semi-broken but working nonetheless.
20:24 chm This would be something to clean up in the PDL3 and core-cleanup work.
20:24 chm BTW, part of the push for PDL3 is the needed full support for arbitrary data types.
20:25 chm That is needed to interface cleanly witth MATLAB, NumPy, ...
20:25 sivoais yes, that would be helpful for what I'm trying to do
20:25 chm Oops, have to go. July 4th dinner calls.  Happy Holiday!  o/
20:25 sivoais have a good one :-)

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