Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!utgpu!water!watnot!watmath!clyde!rutgers!ucla-cs!ames!ucbcad!ucbvax!sdcsvax!sdchem!tps
From: tps@sdchem.UUCP
Newsgroups: comp.lang.c
Subject: Hex Floating Point (was C and Floating point)
Message-ID: <658@sdchema.sdchem.UUCP>
Date: Fri, 3-Apr-87 19:19:26 EST
Article-I.D.: sdchema.658
Posted: Fri Apr  3 19:19:26 1987
Date-Received: Sun, 5-Apr-87 07:37:25 EST
References: <15958@sun.uucp> <5716@brl-smoke.ARPA> <1283@dataio.Data-IO.COM>
Sender: news@sdchem.UUCP
Reply-To: tps@sdchemf.UUCP (Tom Stockfisch)
Organization: UC San Diego
Lines: 63
Keywords: C Fortran Floating Point

In article <1283@dataio.Data-IO.COM> bright@dataio.UUCP (Walter Bright) writes:

>	How about allowing numbers to be specified in binary radix,
>	as in 0b00110101 ? ...
>	As a corollary, support %b in printf and scanf.

Yes, by all means.

>	Also, support specifying floating numbers in hex radix....
>	The syntax would be:
>
>		double def = 0x1324.5E678e+0x12;
>		double abc = -0x.1B4F;
>
>	def prescribes a mantissa of 0x12345E678 and an exponent of
>	(4*16+0x12).
>
>	Note that a + or - would be required before the exponent, to
>	distinguish the e from the digit e.

This is non-orthogonal with the current decimal floating point, in that
the "." is required in order to avoid ambiguity.
This is because

	0x12e+0x34

could be

	0x12e + 0x34

Also, your second example,
	
		double abc = -0x.1B4F;

is just plain ambiguous -- you must require the exponent as well, or

	0x.1b4e+0x23

could be
	
	0x.1b4e + 0x23

If you use the "longest string that could be a token *is* the next token"
disambiguating rule, then you might break some existing code which wanted
to add to hex integers instead of creating one hex float.

Given all these subtleties,
I think this format is going to be confusing to read, no matter how you
disambiguate.  Perhaps a (standardized)
include file defining a macro function could be devised made to work instead.

You might then have

		# include <hexfloat.h>
		...
		double def = Hexfloat( 0x1324.5E678, 0x12 );

where the first arg is the mantissa, and the 2nd, the exponent.  An advantage
here is that you might be able to allow octal and binary representations
as well.

|| Tom Stockfisch, UCSD Chemistry	tps%chem@sdcsvax.ucsd.edu
					or  sdcsvax!sdchem!tps