Date: Thu, 08 Sep 2005 10:57:13 +0900 (JST)
Message-Id: <20050908.105713.102767324.garrigue at math.nagoya-u.ac.jp>
To: andrieu at ijm.jussieu.fr
Cc: romildo at uber.com.br, lablgtk at math.nagoya-u.ac.jp
Subject: Re: Help in C interface
From: Jacques Garrigue <garrigue at math.nagoya-u.ac.jp>
In-Reply-To: <17183.24114.533709.114006 at karryall.dnsalias.org>
References: <20050907192141.GA7177 at malaquias>
	<17183.24114.533709.114006@karryall.dnsalias.org>
Mime-Version: 1.0
Content-Type: Text/Plain; charset=us-ascii
Content-Transfer-Encoding: 7bit

From: Olivier Andrieu <andrieu at ijm.jussieu.fr>

> You can use g_object_set_data_full() to tie your dynamically allocated
> arrays to the object. This function lets you register a finalizer
> function that will free the arrays when the GtkDatabox object is
> finalized. Somthg like this:

Wouldn't it be simpler to register a callback on destroy?
Since a GtkDatabox inherit from GtkObject, it has a destroy signal.
This avoids adding a "key" parameter, whose role seems unclear here.
You can also pack the two arrays in one, to use only one allocation.

So you could write:

   const gint n = Int_val(length);
   gfloat *vetx = g_new(gfloat, 2 * n);
   gfloat *vety = vetx+n;
   g_signal_connect_swapped(G_OBJECT(box), "destroy",
                            G_CALLBACK(g_free), vetx);

This should be enough in practice.
Also, the cost of copying the data is so small that there should be no
problem in the conversion approach you use. Bigarrays are interesting
only when the data is huge, or shared with C in a complex way.

Jacques