Je Fri, 13 Jul 2018 10:21:06 -0400 skribis Predrag: > What are you trying to do with Julia? If you are just trying to do > science it is probably a bad choice. Jeff Bezanson came here to Carnegie > Mellon University to give a talk 2 years ago and I was not too > impressed (arguably I am more interested in science than in computer > language design). They had immense momentum 5-6 years ago but I think > the enthusiasm is dissipating at least among scientist.
It seems so, but that's of course the perception we have of (fellow) scientists having a certain perception. I am using Julia for scientific computing, though nothing that couldn't be achieved with, for instance, Python. Our general circulation models (at our and similar institutes) are written in Fortran. There has been some development in using C++ for such models, but just as the best our IT department is offering on our computers is Ubuntu, ultimately we use Fortran probably for many years to come for our modelling. It is not all too bad, because most of it is Fortran 95/2003, which are probably decent enough for scientific computing. For the statistics on the model output I use Julia. This works fine. What makes Julia, or any other language, good or bad for scientific computing? Probably it is for a large part perception, and momentum. Could it be that enthusiasm is dissipating, or at least it seems so, because of a couple of negative blog posts that pop up quite high in the search engines of the web? Or is it that they sort of promissed a 1.0 version over a year ago but it has yet to emerge? (Of course one shouldn't care so much about version numbers -- surely a stable API is a good thing but maybe one cannot get a stable language out as fast as they want.) Or is it more of a system design issue, probably not the language but the packaging of Julia and all its dependencies? Except for having designed a language with surely very interesting properties, they try to incorporate everything useful for scientific computing, so they have to have, just to give an example, seemless integration of old and proven linear algebra packages. Marco
