Path: utzoo!utgpu!jarvis.csri.toronto.edu!cs.utexas.edu!yale!slb-sdr!saito
From: saito@sdr.slb.com (Naoki Saito)
Newsgroups: comp.lang.c++
Subject: An array of objects again!
Message-ID: <1990Feb9.201223.14332@sdr.slb.com>
Date: 9 Feb 90 20:12:23 GMT
Reply-To: saito@slb-sdr.UUCP ()
Organization: Schlumberger-Doll Research, Ridgefield CT
Lines: 34


	A long time ago, there was a discussion about the initializing the
array of objects.  My question is slightly different from that.

Suppose I want to declare an array of object Cell.  There are two ways to
declare this.

(1) array of pointers to objects
 Cell** ppc = new Cell*[N];   // null constructor.
 for (int i = 0; i < N; i++)	
   ppc[i] = new Cell(i);      // constructor with some argument.

(2) array of objects
 Cell* pc = new Cell[N];      // null constructor.
 for (int i = 0; i < N; i++)
   pc[i] = *(new Cell(i));    // constructor with some argument.

Now, the differences of these two approaches look like
o The approach (1) needs N x (sizeof(Cell*) + sizeof(Cell)) memory space
whereas (2) only needs N x sizeof(Cell).
o The approach (2) creates temporary objects and calls assignment operator =
i.e., pc[i] = *(new Cell(i)); is equivalent to
Cell temp(i);
pc[i] = temp; // assignment operator.
~temp();
whereas (1) simply copies the address to the pointers.

So these are the trade-off.  Am I missing something here?  Does anyone has
an idea which is better than which?  Should the decision be made based on
whether memory space or computation time is important?

-- 
Naoki Saito (saito@sdr.slb.com)
Schlumberger-Doll Research