Perl 6 - the future is here, just unevenly distributed

IRC log for #pdl, 2015-04-10

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

All times shown according to UTC.

Time Nick Message
18:43 opkick [pdl] mohawk2 created travis-testinstall from master (+0 new commits): http://git.io/vehPK
18:44 opkick [pdl] mohawk2 pushed 1 new commit to travis-testinstall: http://git.io/vehXK
18:44 opkick pdl/travis-testinstall 1740ed9 Ed J: Add Travis case for testing not blib but installed
18:46 opkick [pdl] mohawk2 opened pull request #98: Add Travis case for testing not blib but installed (master...travis-testinstall) http://git.io/veh1z
19:01 opkick [pdl] mohawk2 force-pushed travis-testinstall from 1740ed9 to 26d0347: http://git.io/vehPK
19:01 opkick pdl/travis-testinstall 26d0347 Ed J: Add Travis case for testing not blib but installed
19:18 opkick [pdl] wchristian created inlinepdlp-installed-fix (+1 new commit): http://git.io/vehjJ
19:18 opkick pdl/inlinepdlp-installed-fix 925fade Craig DeForest: minimalist inline pdlpp fix
19:19 opkick [pdl] drzowie opened pull request #99: minimalist inline pdlpp fix (master...inlinepdlp-installed-fix) http://git.io/vehjB
19:20 mohawk craig, i cancelled the rest of my testinstalled travis job, so yours should get free rein
19:21 zowie joined #pdl
19:21 zowie Hey (tap tap) is this thing on?
19:23 mohawk craig, hello!
19:23 mohawk craig, i cancelled the rest of my testinstalled travis job, so yours should get free rein
19:23 mohawk PLEASE don't merge it to master until we've had a bit of a pow-wow
19:24 mohawk i'd like to integrate any such activity with my effort to repro the underlying problem using travis
19:27 mohawk having said that, your change does look nice and minimal, so that's a plus
19:34 mohawk zowie, you can see the progress of your job on https://travis-ci.org/PDLP​orters/pdl/builds/57996349
19:35 zowie Hi, thanks!
19:35 zowie OK, I won't merge to master.  I've got my own machines running with the patch, so work is getting done :-)
19:35 mohawk shweet
19:36 * mohawk fixes the problems in the inline tests breaking the "test installed" build...
19:36 zowie (Sorry I spaced out there -- this window ended up half behind the LyX window where I'm reviewing the PP tutorial I posted.  It's sufficiently old that I can get a good read now -- and I'm actually pretty pleased with it...
19:36 mohawk ha ha, awesome!
19:37 zowie The big deal with finding the bug is that you have to run an Inline script from somewhere *outside* the development tree.  Maybe the test script could copy itself to /tmp (no, verboten -- too UNIX-bigoted) and try running itself there...?
19:37 mohawk zowie, what is it you use PDL for, if i may ask?
19:38 mohawk zowie, what my updated travis config is make install clean, then runs the tests
19:38 mohawk so it *can't* use blib
19:38 mohawk (nor can it use the built version of anything)
19:39 zowie Everything.  I do solar astrophysics for a living.  Here's an example: https://www.youtube.com/watch?v=896wj9cxT50
19:39 mohawk regrettably the inline tests are a bit broken at the moment
19:39 zowie Here's some work from last year, using PDL. http://arxiv.org/abs/1404.3235
19:40 zowie re: travis:  Gotcha.  So long as blib doesn't exist, it should be OK... I think.
19:40 mohawk my goal is to actually get it to break without your fix, then work *with* it
19:41 zowie :-)
19:41 mohawk so therefore, if travis likes your fix, i'll "take control" of your commit, if you're ok with that, and work with it on my branch, then merge it when my travis (and inline test) changes work
19:41 mohawk does that work for you?
19:41 zowie I first noticed the break on MacOS Yosemite.  Don't know if it was an OS issue or an update to PDL that broke it -- I'm too lazy to go back and look if someone grubbed around in there.
19:42 zowie re: "so therefore...": Sounds fine.
19:42 mohawk holy cow, this solar storm stuff is amazing!
19:42 zowie Pretty nifty, eh?  All those movies were rendered on a laptop with PDL.
19:43 mohawk that's quite something
19:43 mohawk are the scripts to do it published?
19:43 mohawk which PDL modules are you using?
19:44 zowie They don't show the reduction process.  99.9% of the light in the data from the spacecraft camera is due to stars and interplanetary dust.  It takes a bit of scrubbing to get rid of the noise.  The scripts are published on github: https://github.com/drzowie/STEREO-pipeline.
19:44 zowie The main modules were PDL::Transform, PDL::FFTW, PDL::Graphics::PGPLOT (yecch, I like gnuplot better these days).
19:46 mohawk i'm having a bit of a fight with PGPLOT to make its test coverage be much better
19:46 zowie The reason for PDL::Graphics::Gnuplot was because PGPLOT was such a frickin' pain in the ass.  It might be easier now that PGPLOT has been GPLed and is supported by Starlink -- but for years we struggled with every new platform and gcc update.
19:47 zowie I gave up when I couldn't get anyone to pick up my RGB patch.
19:47 mohawk zowie, i hope you're liking the relatively awesome github/travis facilities sivoais (zaki) has put together? ;-)
19:47 zowie I'm totally loving them.  To be honest, I"ve been under a bit of a rock for about a year.  What a difference that has made!
19:47 mohawk zowie, conceivably we could make an Alien::PGPLOT which includes your patches
19:48 mohawk that's something jberger could comment on best
19:48 zowie I'd be into that, in the sense that I'd be glad to deliver the patches to anyone who wants them.  But to be honest I don't have the time and energy to get back into PGPLOT.  I'd rather focus on making the gnuplot interface work better.
19:49 mohawk that seems like a sound way forward ;-)
19:49 zowie The Alien modules in general are interesting intellectual challenges, since you have to support as many platforms as possible (even ones you hate).
19:50 jberger Alien::Base is supposed to take care of a lot of that for you of course
19:50 zowie Yes, and it sort of does.  But a lot of stuff is just weird enough that you end up either trying to make an exception in Alien::Base or having to recode everything in your new one.
19:51 zowie I'd bet Alien::Base is a lot better now than it was last year...
19:51 mohawk zowie, it has come on in leaps and bounds largely thanks to sivoais, jberger and graham ollis's amazing efforts
19:52 jberger I give credit to them, I burned out pretty hard writing the first iteration of it
19:52 zowie Thanks very much for your efforts jberger!!!
19:52 zowie (heh.  I sent the wrong youtube link before.  This one has a more informative soundtrack...) https://www.youtube.com/watch?v=1kSx7AOwEco
19:53 jberger it needed bootstrapping, which I'm happy I did, but graham is much more knowledgable in the area, so I'm glad he has the lead now
19:53 zowie Alien::Gnuplot and Alien::FFTW were fun to work on, but it must've been a lot harder to carve out the intellectual landscape, starting from just a vague idea.  It was Alien::Base that convinced me I'm now officially an "old fogey" when it comes to software design.
19:54 zowie ("your Alien::Base", I should say)
19:55 jberger oh, why's that?
19:55 jberger I can't say its any kind of beautiful code
19:55 zowie It's very meta compared to the kind of data-banging I do most of the time.
19:55 jberger it has to be
19:56 zowie :-)
19:56 jberger it is so far under what is actually being used, you can barely assume anything
19:56 zowie In the community I work in, it's still hard to convince people to write tests or to consider any edge cases at all.  It was refreshing to see so much thought go into how stuff is structured.
19:58 jberger that's a really cool video btw
19:58 jberger zowie++
19:58 zowie Thanks!
19:59 jberger I did have an even grander scheme for testing Alien::Base, but I had to abandon it
19:59 jberger that is where the Acme::Alien:: dists came from
20:00 jberger too much in the toolchain was fundamental to the tests, I couldn't mimic enough of it without ruining the usefulness of the test
20:00 jberger and part of that is how I burned out :-P
20:00 zowie The best-laid plans of mice and men.  Even Tuomas Lukka, who wrote the PP engine, had to give up on explicit threadloops when it got too hairy.  Then he left entirely...
20:02 jberger I was rather proud that my big simulation module has tests: https://metacpan.org/source/JBERGER/Phys​ics-UEMColumn-0.901/t/free-propagation.t
20:02 jberger maybe not great ones, but I test the known case
20:02 zowie :-)
20:02 zowie You also spell "propagation" right!  Woot!
20:03 jberger how do you write tests for a simulation that predicts behavior that is as-yet unknown :-P
20:03 jberger that was a word I used extensively in my thesis, I got it after a while :-)
20:03 mohawk ha ha
20:03 mohawk i extracted the spelling correction from the LLDF branch and it's now on PDL master
20:04 zowie I had that problem with the fluxon code.  There are two general answers, right?  (1) you test the hell out of known cases and hope it works right in the case you're testing, (2) you do what the rest of the community seems to do, and publish the first thing to come out of the rendering engine, then publish again later to retract.
20:04 mohawk i have a crazy dream it'll one day even be in a released version ;-)
20:04 zowie Thanks for that, Ed.
20:04 mohawk de niente
20:05 mohawk zowie, which one did you choose? ;-)
20:06 zowie Unfortunately we went with (1).  I remember we had a lot of trouble getting the fluxons funded since the market was so crowded with other simulation groups, and everyone was skeptical that they'd work at all.   During that time I went to a conference where our competitors were proudly displaying their first CME simulation.  It evolved from a nearly-spherical origin event to a cubical cloud.
20:07 zowie TTFN, guys.
20:09 mohawk ciao for now
20:11 opkick [pdl] coveralls comment on issue #99:
20:21 zowie left #pdl
20:48 opkick [pdl] mohawk2 pushed 1 new commit to travis-testinstall: http://git.io/vej1x
20:48 opkick pdl/travis-testinstall a35b66a Ed J: Have t/inline* not "use blib" as test framework does that
21:04 mohawk that correctly failed, but the errors aren't beautiful yet
21:24 zowie joined #pdl
21:46 opkick [pdl] coveralls comment on issue #98:

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