NetNews Usenet Archive 1992 #27/NN_1992_27.iso/spool/comp/lang/c/16714

Text File | 1992-11-17 | 1.7 KB | 53 lines

Newsgroups: comp.lang.c
Path: sparky!uunet!utcsri!newsflash.concordia.ca!IRO.UMontreal.CA!CC.UMontreal.CA!sifon!thunder.mcrcim.mcgill.edu!mouse
From: [e-mail] (der Mouse)
Subject: Re: 2-D Arrays, Revisited.
Message-ID: <[e-mail]>
Organization: McGill Research Centre for Intelligent Machines
References: <[e-mail]>
Date: Wed, 18 Nov 92 10:26:04 GMT
Lines: 43

In article <[e-mail]>, [e-mail] (Fahad A Hoymany) writes:

> I am writing a function that should accept a 2-D array (hopefully of
> any size), [...the argument array must be a 2-D array, not done with
> an array of pointers...].

> What should the formal parameters line look like?

This cannot be done without knowing the size of all but the first
dimension at compile time.  To use your "int a[2][5]" example, you have
to know the 5 at compile time.  The formal parameter line can be
deduced easily enough:

- argument is of type int [2][5].
- argument is an array in an rvalue context, so it decays to a pointer
   to its first element; new type is thus int (*)[5].
- formal parameter type must match (modulo a couple of quibbles that
   I'd rather not go into here).

Thus, the formal parameter must have type int (*)[5]:

    sometype dosomething(int (*array)[5])
    {
     ...
    }

or, if you want to be portable to pre-prototype compilers,

    sometype dosomething(array)
    int (*array)[5];
    {
     ...
    }

Notice the 5 must be specified.  If you try to write "int (*array)[]"
instead, the resulting pointer cannot really be used; the only value
for the first subscript that makes any sense is 0, since the size of
the pointed-to thing is not known.

                    der Mouse

                [e-mail]