Home Original page

Python vs. C++ Builder - speed of development

Steve Holden sholden at holdenweb.com
Mon Feb 24 07:33:24 EST 2003 
"Brandon Van Every" <vanevery at 3DProgrammer.com> wrote in message
news:_8I%9.6367$6P2.713325 at newsread1.prod.itd.earthlink.net...
> Delaney, Timothy C (Timothy) wrote:
> >> From: Brandon Van Every [mailto:vanevery at 3DProgrammer.com]
> >>
> >> C++ is not the bottleneck of my current development.  At this
> >> stage I spend
> >> far, far more time figuring out how to objectify various 3D
> >> mathematical
> >> constructs.  I would have the same problems of mathematical
> >> decomposition in
> >> any language.  Development is slow for *that* reason, not C++.
> >
> > May I suggest that you try experimenting and prototyping those
> > constructs in Python, rather than just thinking about them.
>
> Hm, which would I prefer:
>
> 1) set up a new tools environment, learn a new language, figure out how to
> integrate Python with my existing code, think for 1/2 hour, prototype a
tiny
> bit of incremental functionality with Python, decide whether to keep
Python
> or convert it to C++.
>
> 2) think for 1/2 hour, write it in C++.
>
> I don't gain any economies of scale using Python to solve problems I
already
> know how to solve in C++.  In fact, I lose weeks screwing around with new
> ways of doing things.  Those are weeks I don't have right now.  My C++
> approach is quite efficient for the problems I'm currently working on.
That
> is the crux of the matter; some people don't want to accept it.
>
Well, since you've already made up your mind about the relative merits of
time investment in each language it's hardly likely that anything as simple
as rational behavior will change it.

> > I've
> > always found that I'm better at visualising and designing things if I
> > can play with something that works or almost works. Most importantly,
> > I'm more likely to find the limitations of an approach if I'm playing
> > with it than if I'm just thinking about it or writing it down.
>
> That's what incremental development is for.  I don't know *why* 3 lines on
> my spherical triangle converge to a normal.  The paper math was too
> involved, I didn't want to spend another half day chasing that rabbit.
But
> hey, run the combos, diff the normals, notice in the debugger that they
> converge within the accuracy of float vectors.  Must be mathematically
true.
> Or at least, seems damn likely given what else I know.  10 minutes of
> thinking instead of a half day.
>
But I hardly think that Fermat thought "hey, I've tried a few numbers now,
this result must be true for all positive exponents". Deep results require
deep thought, and a feww runs through a debugger isn't a proof. The fact
that a proof exists simply allows you to check the accuracy of your
programming. It's OK to trust someone else's mathematics. But concluding
that debug output from a few test cases means something must be
mathematically true ... sheesh.

As for not wanting to spend half a day chasing that rabbit, the available
evidence is that you prefer arguing to programming.

> Before there was Python, there was functional decomposition, and building
> applications from primitives.  I don't know why people resist the idea
that
> this can be productive no matter what the language.
>
Of course it can. And if you build bottom up you can end up with a usable
toolkit in any language, even assembler. You might even say that bottom-up
modular programming was a logical precursor to XP. That doesn't negate the
fact that for many problem domains assembler isn't the most productive
labgyuage. Neither, as I seem to remember several well-qualified locutors
have pointed out, is C++. Neither, for that matter, is Python. If language
choice is a religious issue, however, any opposing point of view apparently
becomes sacreligious.

> > If nothing else, it will give you another viewpoint to work from ;)
>
> I'll get into Python when I finally hit jobs that C++ cannot do well.
>
You'll get into Python when hell freezes over. As far as I can tell you've
already spent far more time on this thread than it would take an
averagely-competent programmer to produce their first significant Python
program. I therefore conclude you're here for the ten-pound argument and
that you, sir, are a troll.

don't-tell-me,-I-just-got-plonked-ly y'rs  - steve
--
Steve Holden                                  http://www.holdenweb.com/
Python Web Programming                 http://pydish.holdenweb.com/pwp/
Register for PyCon now!            http://www.python.org/pycon/reg.html
More information about the Python-list mailing list