[IndiChix] [ilugb] [x-post] [event] [BLR] Polling interest for a Julia language workshop

स्वक्ष svaksha at gmail.com
Thu Apr 18 05:32:10 UTC 2013


On Thu, Apr 18, 2013 at 8:07 AM, A. Mani <a.mani.cms at gmail.com> wrote:
> On Wed, Apr 17, 2013 at 11:30 PM, स्वक्ष <svaksha at gmail.com> wrote:
>> Not trying to start a language war, but I personally found R
>> interesting yet frustrating, especially in terms of
>> dependency/installation issues, which quickly turned my "fun" personal
>> project into a maintenance nightmare <-- Debugging R code paled in
>> comparison to remotely troubleshooting dependency hell. YMMV.
>
> R code should be checked in the interpreted way and if it is written
> in the functional /equational way then it is easy.
> Dependency handling is auto-magically manageable via 'views'
> In 3.0, this has improved further.

I was narrating my 3-yr old experience so things would have
changed/improved now.  If it was a paid gig, I would not mind the
time-sink with dependencies but not for "fun" stuff. YMMV.



> Earlier, dependency issues could crop up with deprecated packages.

I did (still do) admire the vast amount of libs available, but
unsolvable dependency hell on my debian machine changed from "using R"
to fighting to get the basic stuff working properly. A plug to all the
friendly and helpful folks on #R, but if the core package was not
maintained in sync with the distro, there is only so much they can do.
Again, this being in the past, I moved on.


>> A big plus for R (and other languages that are a few decades old) is
>> the large amount of libraries. Julia is not "production" ready as yet
>> - its getting there and I assume ver-1.0 would be the target (not
>> sure). Their users list has people from the MATLAB and R (some,
>> Fortran) world, including Ruby, Lisp, Perl and Python.
>>
>
> Nobody would want to reinvent the wheel... except those coding in
> relatively lower level languages like C.
> Fortran people will see no point in moving to anything else (small
> differences in performance would matter a lot).
> There are such people in many of institutes in different parts of the
> country including B'lore and nearby places. But you seem to be
> suggesting that Julia is not ready for them.

That is not what I said. Anyone who wants to experiment / learn stuff
with a new language stack should definitely definitely try Julia out.
And, women (students, professionals, or any one who is just curious)
who want to do the same should definitely attend the workshop when I
announce it in a few weeks.


> The entire research literature in statistics lands in R, people do not
> have time to read all the papers in their own area. ~75% of the R
> libraries are written in R. I do not think that this can be easily
> rewritten.

Exactly, but I'd like to give Julia more time to get there. TBH, one
cannot compare a just released (circa 2012, although work started in
2009) language with those with a head-start of a few decades
containing entire statistical stack support.


>> TBH, I'd choose python for any work-related projects but I can live
>> with an experimental tool for my personal projects -  atm, I'm trying
>> to model some WHO data.
>> http://julialang.org/blog/2012/02/why-we-created-julia/, has some
>> thoughts from its creators.
>
> This part is surprising.
> Coding from scratch was assumed to be possible in Julia in earlier
> discussions in R blogger, lists.

Not sure what you mean here. Most of Julia's standard library is
implemented in Julia itself built on top of the LLVM compiler
framework.
-- 
Best,  Vid  ॥ http://svaksha.com


More information about the IndiChix mailing list