Perl 6 - the future is here, just unevenly distributed

IRC log for #pdl, 2013-08-02

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

All times shown according to UTC.

Time Nick Message
01:57 Meiermann joined #pdl
07:08 MeierM joined #pdl
12:06 chm joined #pdl
12:07 chm run4flat, Hithaldu: To add some facts to the speculation from a few days ago, http://mailman.jach.hawaii.edu/pipermai​l//pdl-porters/2013-January/005226.html
12:09 chm My recollection was there was no follow through from anyone past that list discussion.
12:09 chm As for me, I was working on getting PDL-2.006 out the door.
12:09 chm Joel was working on his dissertation.
12:11 chm Also, I believe some of the confusion is the fact that #pdl discussion does not equal perldl or pdl-porters list discussions.
12:11 chm Mithaldu: perldl is the list to use for general PDL discussion and help.
12:12 chm pdl-porters is the list for PDL development, internals,...
12:12 Mithaldu noted, thanks :)
12:14 chm Mithaldu: I was looking at the bool overload in Core.pm and I think by use of better warnings, we could have
12:15 chm the check for "bad code" David was calling out and still avoid surprises for generic perl modules.
12:16 chm I still think the biggest problem is the inconsistency and not the actual difference from the naive assumption that an object does not overload bool
12:16 chm Unfortunately, getting people to grind through code to sort out these types of problems rarely (never) gets many takers.
12:17 chm On github or off.  :-)
12:18 chm Did you get things working with the quaternion PDL::PP implementation?
12:18 Mithaldu chm: you mean throw a warning and return 1, instead of dieing?
12:18 Mithaldu also, nope, haven't looked at that yet
12:18 Mithaldu been busy with other things :)
12:19 Mithaldu (specifically, a copypaste detector)
12:19 chm I was thinking throw a warning and return the object.  The result at that point is whatever is "standard"
12:20 Mithaldu wouldn't that result in the overload being triggered *again*?
12:20 chm According to the overload docs, and a test of my own, that does the right thing (overload explicitly does not trigger again)
12:20 Mithaldu also, this implies removing the bits that return the value if a pdl has only one value, right?
12:20 Mithaldu oh cool
12:21 Mithaldu if you're happy with that i'll make a branch over the next week with which you can test it out
12:22 chm Unfortunately, I realized why the scalar piddle option works, it prevents a scalar value in a piddle from not acting the way a perl scalar would.
12:23 Mithaldu can you rephrase that? i don't understand
12:23 chm Definitely worth a list discussion on the topic.
12:24 chm In PDL operations, perl scalars get promoted to piddles automatically as required.
12:24 chm You don't want the operations on that value (if now "in" a piddle) to surprise the user.
12:24 Mithaldu oh i see
12:24 Mithaldu tricky
12:24 chm As you can see, surprise may be the least bad option here.  Warnings might help.
12:25 Mithaldu yes
12:25 Mithaldu it will need to be documented though
12:25 Mithaldu i'll add that when i make the branch
12:25 chm On the scale of things, I think a more consistent PDL with clear documentation would be a much better fix in the long run.
12:26 chm FYI: here is a list of inconsistencies that could be fixed: random naming and arg order for IO routines, random use of Test, Test::More or nothing
12:27 chm in the tests, FFT and FFTW support but inconsistent and not interoperable, TMTOWTDI for a lot of functionality through out PDL
12:27 Mithaldu chm: random thought: maybe chunk stuff like that onto questhub?
12:29 chm I don't know.  An important concern is to keep PDL working and usable.  I run into a lot of modules (mature API) that
12:30 chm re-invent themselves periodically meaning one has to constantly "fix" things that wasn't broken formerly.
12:31 chm PDL is being used for research and while I want things improved, I don't want to break it for users along the way just for perl-purity.
12:31 Mithaldu right
12:31 Mithaldu keep in mind:
12:31 Mithaldu my original request was only for a way to configure that behavior
12:32 chm Right, that that is problem what will end up here.  It would be easy to implement a state variable to determine the behavior used.
12:33 chm Then the die could be used in the PDL shell for better debugging.  Or it could do the "object is only a scalar ref in bool" for others.
12:34 chm That should read, "then that is probably what we will end up with here"
12:35 Mithaldu haha, cellphones :)
12:35 Mithaldu i think i'm fine with the die being the default
12:36 Mithaldu honestly, i'll need to think it over
12:36 Mithaldu i haven't got my head in the game on this topic at the moment
12:40 chm My interest is to get back to POGL2 for modern OpenGL support.
12:41 * Mithaldu nods
12:41 chm Not hard, just tedious and uninteresting to do---except in the result.
12:41 Mithaldu weren't you going to document the tedious parts so you could offload them?
12:41 chm I had hoped that PDL3 would be possible this summer but no-one has really stepped up
12:41 Mithaldu not having to remember where to put my ARBs would be nice
12:42 chm and I've had less time to contribute than I had thought.  It will be another PDL-2.x release
12:42 chm but at least the 64bit index support and some other niceness things might make it in.
12:43 chm I really wish folks would get behind implementing Alien::XXX modules for the PDL dependencies.
12:43 chm That would make things much cleaner and generally usable.
12:44 chm Re: document the tedious parts, that takes time too.
12:45 Mithaldu reminds me i need to play with 64 bit perl on windows more
12:45 Mithaldu also, chm: i strongly suggest you chunk tasks onto questhub.io
12:45 chm It is definitely nice having the large memory mapped file support.
12:45 chm What is questhub?
12:45 Mithaldu http://questhub.io/
12:46 Mithaldu it's a cross of a game and a task tracker
12:46 Mithaldu you can put things on there you want to do
12:46 Mithaldu and others can tell you which of those things they like
12:46 chm Thanks, I'll take a look.  Later, o/
12:46 Mithaldu and you can just open them and let others take them
13:31 pdurbin they talked about questhub at http://perlmaven.com/neil-bowers
21:30 chm joined #pdl

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