NetNews Offline 2/NetNews Offline Volume 2.iso/news/comp/lang/c++-part1/489

Text File | 1996-08-05 | 4.8 KB | 152 lines

Path: news.sprintlink.net!datalytics!news
From: Rob Stewart <[e-mail]>
Newsgroups: comp.lang.c++
Subject: Re: ode to c++
Date: 4 Jan 1996 19:00:26 GMT
Organization: Datalytics, Inc
Message-ID: <4ch84a$[e-mail]>
References: <4bq620$[e-mail]>
NNTP-Posting-Host: pc071.datalytics.com
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
X-Mailer: Mozilla 1.22 (Windows; I; 32bit)

[e-mail] (The Crow) wrote:
>About 2 months ago I got very into Pascal (because it was the only thing on the
>network at my highschool) I wrote some pretty cool stuff with it.  I also
>messed around with VB and found it to be FAR too slow at number crunching.
>Then I finally sat down with C++ these last two days.  I rewrote a simple
>Pascal program in C++ that calculates Pythagorean triples, here is the C++
>version
>
>//    The amazing Pythagorean Triple Generator
>#include <stdio.h>
>
>main()
>{
>const     deep = 13200;
>long      a = 1,b = 1,c = 1;
>
>    for(c = 1; c<= deep; c++)
>    {
>        for(b = c/2; b <= c; b++)
>        {
>            for(a = (c-b)+1; a <= b; a++)
>            {
>                if (a*a + b*b > c * c) break;
>                if (a*a + b*b == c * c)
>                printf("%ld %ld %ld \n",a,b,c);
>            } // for a
>        }    // for b
>    }    // for c
>}    // main
>
>
>The Pascl version took 20 seconds for C to get to 500, C++ took 6 seconds! It
>blew my mind!, I'm never going back! long live C!
>
>
>--
>The Crow - [e-mail]
>"It can't rain all the time"
>-Kryptology
>

Now you know why so many people prefer C and C++.  They aren't
the easiest languages to learn, but they are screamers.  Now,
two points of order.  First, you wrote a C program.  Without
resorting to OOP, I'll show you how to take advantage of some
better things you can do with C++.  Second, you are falling
into some stylistic traps that I would like to correct.

The code below illustrates both points:

#include <assert.h>      // assert
#include <iostream.h>    // cout
#include <math.h>        // sqrt
#include <stddef.h>      // stddef.h

typedef unsigned long index;

int main(void)
{
   const int DEEP = 13200;
   const index maxIndex = sqrt(ULONG_MAX);
   for (index c = 1; c <= DEEP; c++)
   {
      for (index b = c / 2; b <= c; b++)
      {
         for (index a = (c - b) + 1; a <= b; a++)
         {
            assert(a < maxIndex);
            index a2 = a * a;
            assert(b < maxIndex);
            index b2 = b * b;
            index sumOfSquares = a2 + b2;
            assert(c < maxIndex);
            index c2 = c * c;
            if (sumOfSquares > c2)
            {
               break;
            }
            if (sumOfSquares == c2)
            {
               cout << a << ' ' << b << ' ' << c << ' '
                  << endl;
            }
         }
      }
   }
   return 0;
}

Note that I change the case of "deep."  Making it uppercase
makes its being const stylistically obvious.  Note also that I
declared and initialized the for loop indexes in the for loop
statements.  At the same time, I changed their type to size_t.
This is a standard typedef to the native type for the current
environment.  It is usually a good choice for loop indices.
(On the other hand, for some environments, it could prove too
small for your purposes.)

Note that you must be certain of the capacity of the type you
use for the loop indexes.  You used a signed type for an
unsigned value, so you immediately lost half the capacity to
the sign bit.  Furthermore, squaring the indexes could exceed
the capacity of the type you selected.  And, of course, adding
the square of a to the square of b could exceed the capacity.
As originally written, your program would simply provide
incorrect answers because the values in the conditionals would
wrap to small numbers.

I used the assert macro to report a value out of range and
terminate the program.  The assert macro conditional compiles
to a NOP if you #define NDEBUG.

I also added whitespace around all binary operators.  This
makes the code a little more readable.  I also removed the
duplicate calculations in the inner for loop by creating a
couple of temporaries.  (Do note the identifiers I chose a
reasonably indicative of what they represent.)

Finally, I wrote the if statements in the block style for
consistency.  Many old C programmers still use the (perfectly
legal, but harder to read and more error prone) style you
used.  However, as soon as the conditional requires a second
statement, you must use the block style.  Without the block
style you can get confused as to what belongs to the
conditional and what doesn't.  (This has been the subject of
PC-Lint advertisements for years.)

My stylistic comments are subjective, to be sure, but well
conceived and time-tested.

--
Robert Stewart        | My opinions are usually my own.
Datalytics, Inc.
[phone]
[e-mail]