From j.l.schonfelder@liverpool.ac.uk Tue Jun 1 12:52:02 1999 Received: from mailhub2.liv.ac.uk (mailhub2.liv.ac.uk [138.253.100.95]) by dkuug.dk (8.9.2/8.9.2) with ESMTP id MAA00689 for ; Tue, 1 Jun 1999 12:52:02 +0200 (CEST) (envelope-from j.l.schonfelder@liverpool.ac.uk) Received: from pcmail2.liv.ac.uk ([138.253.252.15]) by mailhub2.liv.ac.uk with esmtp (Exim 1.92 #1) for SC22WG5@dkuug.dk id 10om8p-0007bE-00; Tue, 1 Jun 1999 11:51:31 +0100 Received: from [138.253.102.118] (helo=pc102118.liv.ac.uk) by pcmail2.liv.ac.uk with smtp (Exim 1.92 #1) for SC22WG5@dkuug.dk id 10om8o-0002PR-00; Tue, 1 Jun 1999 11:51:30 +0100 From: Lawrie Schonfelder Reply-To: j.l.schonfelder@liverpool.ac.uk To: SC22/WG5 members Subject: (x3j3.1999-150) Re: (SC22WG5.1603) F95, String Module and problems due to Language deficiencies In-Reply-To: <3.0.1.32.19990528122806.00e428c8@postoffice.worldnet.att.net> Message-ID: Date: Tue, 1 Jun 1999 11:51:30 +0100 (British Summer Time) Priority: NORMAL X-Mailer: Simeon for Win32 Version 4.1.5 Build (43) X-Authentication: IMSP MIME-Version: 1.0 Content-Type: TEXT/PLAIN; CHARSET=US-ASCII On Fri, 28 May 1999 12:28:06 -0500 Dick Hendrickson wrote: > At 05:38 PM 5/28/99 +0100, Lawrie wrote: > [snip]> > >For a datatype like a variable precision number this becomes a real pain. > >Unless I am wrong, given a type like > > > >type NUMBER > > INTEGER :: exp > > INTEGER, POINTER :: sig(:) > >ENDTYPE NUMBER > > > >there is no way of declaring something that would be equivalent to > > > >type(NUMBER),PARAMETER :: zero = 0 > > > >no matter what circumlocutions one goes through. > > > > Can't you do > > type (NUMBER), PARAMETER :: zero = NUMBER(0,null()) This may be allowed but is only valid for the value zero and then only if a disassociated sig(:) is defined as a representation for zero. For any other value the field to be pointer assigned by the intrinsic constructor to %sig must have the TARGET attribute but it is explicitly forbidden for any declaration of an object that is both a target and a parameter. type(NUMBER),PARAMETER :: one = NUMBER(0,(/1/)) fails because the array constructor does not have the TARGET attribute. INTEGER,TARGET :: tone(0:0)=(/1/) type(NUMBER),PARAMETER:: one=NUMBER(0,tone) fails since tone is a variable and is not allowed in an initialisation expression. > > I think I've tried that and it usually works (depends on the compiler). > > If you want the type or constructor to be private then I think you'd > have to define all of the popular parameters (0, 1, whatever) yourself in > the module and make them public. This is certainly true, but I contend I can not even do it for a few selected constants in the defining module. > > I'm pretty sure that F2k will allow overloading of the type constructor, > so a person could do something like > type (NUMBER), parameter :: zero = NUMBER('do whatever you need to make & > & the value be a zero') Allowing overload of the constructor was something I proposed first for F90 and again for F95. It helps but not for PARAMETERs unless the restrictions on initialisation expressions are relaxed. It may be that we may need to introduce a CONSTANT object as distinct from a PARAMETER that can be defined by a relaxed specification expression rather than an initialisation one, but a CONSTANT once defined cannot be redefined. > > and this is supposed to (in my opinion) work even though the type is > private. The module/type author merely needs to provide overloads for > NUMBER that decode the arguments in The Right Way. > > In thinking about it I bet there isn't a good way to do parameters. The > idea of a user defined function on the right hand side of a parameter > definition will be hard to get right in the standard. Since the > likely case will be setting to zero maybe we could/should allow a > type constructor with no arguments to supply the appropriate version > of zero for everything. Whatever is used as default for EOSHIFT and NULL() > if the component is a parameter. > > Dick Hendrickson > -- Lawrie Schonfelder Director, Computing Services Dept. The University of Liverpool, UK, L69 7ZF Phone: 44(151)794 3716, Fax: 44(151)794 3759