Perl 6 - the future is here, just unevenly distributed

IRC log for #pdl, 2014-12-16

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

All times shown according to UTC.

Time Nick Message
12:16 drrho joined #pdl
13:13 drrho joined #pdl
13:40 drrho joined #pdl
14:28 drrho joined #pdl
14:54 drrho joined #pdl
15:13 drrho joined #pdl
16:15 drrho joined #pdl
18:14 sivoais I'm writing this here before I forget. I'll write something to the PDL mailing list later.
18:14 sivoais But
18:15 sivoais has anyone thought about using RPerl for PDL3?
18:16 sivoais I notice that some tools in the Python world try to convert code/expressions into something natively compiled
18:16 sivoais in order to bypass GIL problems
18:16 vicash that seems like a restriction for python
18:16 sivoais <https://code.google.com/p/numexpr/>, <http://numba.pydata.org/>, <http://cython.org/>
18:17 sivoais yes, but I'm thinking it would be easier to think in terms of different computation engines if PDL3 built up expressions in the background and passed those to the actual backend
18:18 vicash haha.. in the end everyone re-invents Lisp
18:18 mohawk i have yet to find RPerl a useful idea
18:18 sivoais mohawk: that's because you can just write the C :-P
18:19 vicash i think that is the main point
18:19 sivoais and then Inline::C it
18:19 vicash with Inline::C and Pdlpp one does not need various compute engines
18:19 vicash you can use Roles to define which compute engine is picked but each compute engine will have their idiosyncrasies and then there will be workarounds leading to excessive complexity
18:20 vicash does PDL have that big a user base and also a use case to warrant multiple compute engines ? That is a question that should be asked i think
18:20 vicash i dont disagree with the multiple compute engines point per se, but PDL3 is falling under the "Second system effect" like Perl6
18:21 sivoais compilers are usually like that no? :-P I think LLVM did an interesting take with how they abstracted things
18:21 sivoais yeah, that's what I'm worried about
18:21 vicash Yes compilers have different backends... but all of them start with 1 backend
18:21 vicash LLVM also started with only x86 as a backend.. having an ARM backend now has taken 10+ years
18:22 vicash i would rather PDL3 come out so that features that are required are available.. the backend should be swappable as a role or something later if necessary or if there is a user demand. but that's what is missing.. chm pointed out earlier that we only know the users based on downloads from sourceforge
18:23 vicash we do not know how they use PDL and if they are seeing performance issues or need a separate backend...
18:25 sivoais aye, that's a problem...
18:25 sivoais Do we want to actively go find people? I can help with that.
18:26 vicash well.. from my perspective, if PDL3 takes too long and PDL2 is stuck at 2.007 folks who need something in Perl will just move their software to use R or NumPy or whatever in the interim.. once that is done, it is over
18:26 vicash most likely PDL2 will solve all their problems
18:27 vicash but do they know that ?
18:27 sivoais *sigh* marketing
18:27 vicash i think we need to take Julia's website examples and say you can do all this with Perl too... and enumerate each thing
18:28 vicash the fact is as i have noticed is that most Perl devs are either Web developers (and dont use scientific stuff much) or regex/parsing type users who use perl for what it does best
18:28 sivoais good idea! I have a plan for a wiki. I can start next week.
18:29 vicash the scientific community is small and folks need to be told
18:29 sivoais there are also NLP, networking researchers, and bioinformatics people
18:29 vicash yes that too.. do the NLP guys use Perl to do their statistical work ? i dont think so.. none of the modules on cpan seem to do that.. they are using it for text management.
18:30 vicash BioPerl is different. i know that is big.
18:30 vicash then there is Mithaldu and the OpenGL crowd that i have no clue how big it is..they are a good candidate too.. PDL + OpenGL will make for some great simulations
18:30 sivoais yeah, I haven't seen it used as much for stats than for thinks like word segmentation and writing parsers
18:30 vicash i would love to learn how to do that
18:31 sivoais s,thinks,things,
18:32 sivoais and infosec...MetaSploit used to be written in Perl. Now they moved to Ruby.
18:32 vicash sivoais: i am trying to use PDL in a finance project (although development has stallled since October) but that is a good area to apply it as well
18:32 vicash yea... if only they knew how to use Roles ;)
18:32 vicash sivoais: we should write a project that takes Metasploit, rips its info out and creates a Perl library out of it
18:33 vicash although it will be pointless.. majority of infosec folks are users
18:34 sivoais heh, well...we could make it easier to use, right?
18:34 vicash yes we could.. metasploit is dog slow to load
18:34 sivoais also, we could say it is more portable
18:34 sivoais metasploit on Android...?
18:36 vicash have you tried Perl on Android?
18:36 vicash do you have to jail break the Android first ?
18:36 sivoais I don't recall having to
18:37 sivoais Hugmeir is the expert on that. He's in #p5p sometimes
18:46 sivoais "Notwithstanding this decline, APL finds continued use in certain fields, such as accounting research."
18:46 sivoais from Wikipedia
18:46 sivoais links to "Knowledge of a computer language such as Fortran, Pascal, or APL is sufficient." <http://www.gsb.stanford.edu/programs/phd/academic-experience/fields/accounting/requirements>
18:47 vicash the key word is "sufficient"
18:48 vicash Morgan Stanley invented this language called A+ that is still used by their quants to develop algorithms
18:51 sivoais someone made an array language on top of Perl : <http://lang5.sourceforge.net/tiki-index.php>, <https://metacpan.org/pod/Array::APX>
18:53 sivoais and here's a paper on it <http://archive.vector.org.uk/art10500710>
20:40 sivoais interesting. Perhaps <http://p99.gforge.inria.fr/p99-html/group__generic.html> can help Pdlpp ?
20:40 sivoais C11 has generic type expressions and P99 uses the preprocessor and C99 features to emulate the feature
20:46 sivoais also interesting: "Adding classes to C via libffi & P99" <https://github.com/richardjrossiii/CFFIClasses>
20:47 * vicash sticks to using C++98 for all C++ work
20:54 sivoais I'm reading the libtool docs and a common theme is..."oh, you could do it this way...but no, that's too simple and doesn't work everywhere"
20:56 jberger I was able to add method modifiers to my insane Class::Anonymous last night
20:56 jberger https://github.com/jberger/Class-Anonymous/blob/master/t/utils.t
20:59 sivoais hehehe, nice...
21:00 * sivoais reads the Utils.pm
21:03 sivoais this might be interesting to look at relative to Alien::Base <https://github.com/magnolia-k/Enbld>
21:04 sivoais but I don't know if it works on Windows :-(
21:05 vicash bah who cares about Windows, the cool developers use Apple Macs :)
21:05 sivoais then they deploy to Linux :-P
21:06 vicash i actually worked for a venture funded company where two of the teams literally did that
21:06 vicash and then we would hear excuses of how shit works on the Mac but not on Linux
21:06 jberger I don't trust Perl software that isn't on CPAN
21:06 vicash predictably the company died in the end
21:06 vicash if the perl module is not on CPAN then the user doesn't want others to use it
21:07 vicash that's how i see it..
21:07 sivoais jberger: it's on CPAN <https://metacpan.org/release/Enbld>aa
21:07 jberger that and/or they aren't in touch with the perl community (which probably means pre-modern perl code too)
21:09 jberger POPO objects
21:09 jberger LOTS of globals (and worse, global hashes)
21:11 jberger but perhaps not terrible
21:28 sivoais hmm, OpenBSD has a libtool replacement written in Perl <http://cvsweb.openbsd.org/cgi-bin/cvsweb/src/usr.bin/libtool/>

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