Re: reading a float?

To: info-dylan@ai.mit.edu
Subject: Re: reading a float?
From: brucehoult@pobox.com (Bruce Hoult)
Date: Thu, 20 Apr 2000 18:45:03 -0400 (EDT)
Organization: The Internet Group Ltd
References: <brucehoult-1904001153420001@bruce.bgh> <hX8L4.6255$QX1.232123@typhoon.ne.mediaone.net> <brucehoult-1904001604070001@bruce.bgh> <8F1BE6261chrisdoubleconz@202.27.184.33> <ravr9c1vx51.fsf@telnet.math.fu-berlin.de> <brucehoult-2104000026580001@bruce.bgh> <38FF2C46.D8580EF8@miqs.com>
Xref: traf.lcs.mit.edu comp.lang.dylan:11889

In article <38FF2C46.D8580EF8@miqs.com>, Scott Ribe <sribe@miqs.com> wrote:

> Bruce Hoult wrote:
> > 
> > 1) punt and do whatever the C library on the given platform does
> 
> It's been implied that this may be somewhat limited and not cover all
> formats? If so, I would think the answer is obvious: Dylan code to
> reduce the input string to a canonical representation, followed by C
> library conversion.

Right.

I just took a look at the CodeWarrior library source code (which they
kindly give you) to see what they do in atof.  Basically they've got a
state machine that parses the float and puts it in a canonical form into a
MacOS "Decimal" structure...

struct decimal {
    char sgn;                           /*sign 0 for +, 1 for -*/
    char unused;
    short exp;                          /*decimal exponent*/
    struct{
        unsigned char length;
        unsigned char text[SIGDIGLEN];  /*significant digits */
        unsigned char unused;
        }sig;
};

... and then uses the MacOS str2num function.  Which I'm quite sure will
Get It Right -- Apple has cared about precise FP with the SANE (Standard
Apple Numeric Environment) right since the Apple ][ ProDOS days.

-- Bruce

References:

reading a float?

From: brucehoult@pobox.com (Bruce Hoult)

Re: reading a float?

From: "Scott McKay" <swm@mediaone.net>

Re: reading a float?

From: brucehoult@pobox.com (Bruce Hoult)

Re: reading a float?

From: chris@double.co.nz (Chris Double)

Re: reading a float?

From: David Lichteblau <lichtebl@math.fu-berlin.de>

Re: reading a float?

From: brucehoult@pobox.com (Bruce Hoult)

Re: reading a float?

From: Scott Ribe <sribe@miqs.com>