NetNews Usenet Archive 1993 #1/NN_1993_1.iso/spool/comp/lang/cplus/18935

Internet Message Format | 1993-01-11 | 3.8 KB

Path: sparky!uunet!mcsun!julienas!sophia!xihu.inria.fr!gchow
From: [e-mail] (Gloria Chow)
Newsgroups: comp.lang.c++
Subject: Re: Passing 2-d arrys to functions
Message-ID: <[e-mail]>
Date: 11 Jan 93 13:58:08 GMT
References: <[e-mail]> <[e-mail]> <[e-mail]> <[e-mail]>
Sender: [e-mail]
Organization: INRIA, Sophia-Antipolis (Fr)
Lines: 88

In article <[e-mail]>, [e-mail] (Suresh Thennarangam - Research Scholar) writes:
|> In article <[e-mail]> [e-mail] (Rune Henning Johansen FBA) writes:
|> >In article <[e-mail]> [e-mail] (Nikki Locke) writes:
|> >
|> >  >  For a two (or more) dimensional array, to access element [m][n], the
|> >  >  compiler needs to know the address of the first element, the size of each
|> >  >  element, and the number of elements in the least significant dimension.
|> >  >  The calculation for an array declared [DIM1][DIM2] is
|> >  >  ((m * DIM2 + n) * sizeof(element)).
|> >
|> >I'm including a little program that seems to work fine. Here I'm pas-
|> >sing only the pointer to  a two dimensional array.  Can I expect such
|> >programs to work, and if so, why?
|> >
|> > - Rune
|> >
|> > = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
|> >
|> >#include <stream.h>
|> >
|> >#define begin {
|> >#define end   }
|> >
|> >void Input ( double **b )
|> >begin
|> >    b[0][0] = 1.0;
|> >    b[1][0] = 2.0;
|> >    b[0][1] = 3.0;
|> >    b[1][1] = 4.0;
|> >end
|> >
|> >void Output ( double **b )
|> >begin
|> >    cout << b[0][0] << "\n";
|> >    cout << b[1][0] << "\n";
|> >    cout << b[0][1] << "\n";
|> >    cout << b[1][1] << "\n";
|> >end
|> >
|> >int main ()
|> >begin
|> >    double **a = new double*[2];
|> >    a[0] = new double[2];
|> >    a[1] = new double[2];
|> >
|> >    Input  ( a );
|> >    Output ( a );
|> >
|> >    delete a[0];
|> >    delete a[1];
|> >    delete a;
|> >
|> >    return 0;
|> >end
|>
|> Yes, I think this program will work always .. since you have passed in
|> a double ** pointer which is a vector of double*, the least dimension
|> of the array is not required to correctly index the values.
|> Although the same syntax can be used to access 2-d arrays (no flames
|> please .. I mean statically allocated as in double arr[2][2] ) and
|> variables of type pointer to pointer ( double **), the underlying memory
|> layout need not be the same. However, the algorithm used to index either
|> one is the same.. pointer arithmetic.
|>
|> I believe( correct me if I am wrong) most or all C/C++ compilers always
|> use the same pointer arithmetic no matter whether the array passed
|> is declared as  double arr[][least_dim]  or  double **. This explains
|> how we can successfully mix pointer arithmetic and array-indexing.
|> The last dimension is required merely for setting up storage.
|>

I have always thought (could very well be wrongly so, since I am really
a "C" programmer, and am just in the process of learning C++ ).
The pointer arithmetic involves for something like double **a is as
follows:
       &( a[i][j] ) = a + sizeof( double * ) * i + sizeof( double ) * j

whereas for double a[][least_dim], it will be :
       &( a[i][j] ) = a + sizeof( double ) * ( i * least_dim + j )

I have always declared by multidimensional arrays by the first method
since it allows me to index arbitraily sized arrays without telling the
compiler specifically what the dimensions are.  I didn't even know you
can do something like "double arr[][least_dim]" ( that goes to show you
can still teach an old dog a new trick :) ).  Is "least_dim" here a
run-time variable, like in Fortran, or is it a constant??

Gloria