On Monday, 21 January 2013 at 23:42:49 UTC, Nick Sabalausky wrote:
I've heard that argument before, but I've yet to come across a
good
explanation of how this "dynamic mindset" different from the
static
one. Just that "it exists" and "it's different". Far as I can
tell so
far, the "dynamic mindset" is just "Bugs? Efficiency? Meh, I
don't care
as long as I'm piling on and shipping (easily broken) code".
So what is this "dynamic mindset" that makes using dynamic
typing
productive and non-bug-prone? How does it work, if it's not as I
described above?
Take a look at this project: https://github.com/pydata/pandas
It's a very effective library, that's easy to use in a flexible,
interactive fashion. It's fast due to judicious use of C and
cython where absolutely necessary and numpy elsewhere.
One of the strengths of the library is it's ability to make a
"best guess" as to what to do with information it's passed,
irrelevant of what type that information is. It's not the sort of
behaviour you'd want in a large production codebase, but for an
interactive toolbox it speeds things up by orders of magnitude.
Whatever you throw at it, it just works, which for a
high-pressure financial analyst (the primary target audience) is
absolutely invaluable.
This is all stuff that (with better shared library support) D
could do, making it the awesome king of all languages, but at the
moment i'll still to drafting my data analysis in python and then
writing it up in D when it's settled.
P.S.
dmd 32bit on linux produces functional shared libraries, which
can be happily called by IDL
(http://www.exelisvis.com/ProductsServices/IDL.aspx). IDL rules
the roost in several data heavy scientific arenas (fusion
physics, medical imaging, solar physics, astophysics etc) so this
could be a potential jackpot for D. By and large people don't
write their own extensions for IDL as they don't want to get
their hands dirty with C. D could be the perfect tool for the job.
I've recently started collaborating with some people from MAST
and JET (http://www.ccfe.ac.uk/), working on some better data
analysis tools for some of the fusion diagnostics, to be run
online during "shots" (running the reactor) and retrospectively.
They all use matlab (yuck) and/or IDL so naturally i'll be
writing my code as extensions to IDL, written in D!
