CRM114 Discriminator - Controllable Rege

On Mon, Dec 20, 2004 at 02:29:08PM -0500, Bill Yerazunis wrote:
> 
> Well, the next set of usability tweaks and requests is going into
> CRM114... including a real document (i.e. book) on the language.
> 
> Here's a bit of code I could use...
> 
> Does anyone have any code around that takes a double-precision floating
> point number, and yields the C "printf()" format that expresses it in 
> the fewest characters that doesn't lose significant information?
> 
> Such as:
> 
> 	things with no fraction as integers (no decimal point)
> 	
> 	things with a fraction that ends in a few digits, print with
> 	just enough digits that all the unprinted ones are zeroes.
 
  Or, presumably 9, in the case where it's being rounded up.  (For
example, 1.0 - 1E-20)

> 	things that are too big to express as fixed-point numbers
> 	should be printed in E-notation, again with the caveat
> 	that any digit that isn't printed is zero.       
> 	
> Examples:
> 
> 	3		--> %1f
> 	3.2		--> %3.1f
> 	3.14159		--> %7.5f
> 	-123.456	--> %8.3f
> 	6.02E23		--> %7.2e
> 	-1.987E-66	--> %10.3e	
> 
> (or something like that...)
> 
> I'm toying with how to format output with :#: and just playing with ideas.

Suggestion: 

  You might want to crib this from the Perl output format code - that's
almost exactly how Perl prints numbers.  No need to reinvent the wheel.

  -- Clifton

-- 
          Clifton Royston  --  cliftonr@... 
         Tiki Technologies Lead Programmer/Software Architect
Did you ever fly a kite in bed?  Did you ever walk with ten cats on your head?
  Did you ever milk this kind of cow?  Well we can do it.  We know how.
If you never did, you should.  These things are fun, and fun is good.
                                                                 -- Dr. Seuss

From: Clifton Royston <cliftonr@...>
   > 
   > (or something like that...)
   > 
   > I'm toying with how to format output with :#: and just playing with ideas.

   Suggestion: 

     You might want to crib this from the Perl output format code - that's
   almost exactly how Perl prints numbers.  No need to reinvent the wheel.

Actually, I can't.

Well, I can -look- at their code, but I'd have to rewrite it again.

The issue is GPL v. whatever copyrighting.

CRM114 is GPL, and I own the copyright so I know what's going on.

PERL is some other license (or I have been led to so believe) so it's 
not that easy.

(as to %g - the problem there is it _always_ prints out the
significant digits... I think it might be useful to print out, say, 3
digits no matter how many nonzero digits followed.  Or 4 digits.  Or
5...)

    -Bill Yerazunis

On Mon, Dec 20, 2004 at 03:15:50PM -0500, Bill Yerazunis wrote:
> 
>    From: Clifton Royston <cliftonr@...>
>    > 
>    > (or something like that...)
>    > 
>    > I'm toying with how to format output with :#: and just playing with ideas.
> 
>    Suggestion: 
> 
>      You might want to crib this from the Perl output format code - that's
>    almost exactly how Perl prints numbers.  No need to reinvent the wheel.
> 
> Actually, I can't.
> 
> Well, I can -look- at their code, but I'd have to rewrite it again.
> 
> The issue is GPL v. whatever copyrighting.
> 
> CRM114 is GPL, and I own the copyright so I know what's going on.
> 
> PERL is some other license (or I have been led to so believe) so it's 
> not that easy.

Perl is, if I recall correctly, dual-licensed under the recipient's
choice of Artistic License, which is pretty much a "Do What Thou Wilt
(but don't blame us)" or GPL.  This should let you use Perl, or code
taken from it, in either a GPLed or non-GPLed way.

...

Yes, I just unpacked the source for the current distribution we're
using here (5.8.5) and behold the beginning of the README:


    This program is free software; you can redistribute it and/or modify
    it under the terms of either:

        a) the GNU General Public License as published by the Free
        Software Foundation; either version 1, or (at your option) any
        later version, or

        b) the "Artistic License" which comes with this Kit.

    This program is distributed in the hope that it will be useful,
    but WITHOUT ANY WARRANTY; without even the implied warranty of
    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See either
    the GNU General Public License or the Artistic License for more details.

    You should have received a copy of the Artistic License with this
    Kit, in the file named "Artistic".  If not, I'll be glad to provide one.

    You should also have received a copy of the GNU General Public License
    along with this program in the file named "Copying". If not, write to the
    Free Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
    02111-1307, USA or visit their web page on the internet at
    http://www.gnu.org/copyleft/gpl.html.

...

  You can read the rest of the preamble, but I think that right there
is what you're looking for - being offered under the GPL, you are
entitled to use code from it in other GPLed software. 

  -- Clifton

-- 
          Clifton Royston  --  cliftonr@... 
         Tiki Technologies Lead Programmer/Software Architect
Did you ever fly a kite in bed?  Did you ever walk with ten cats on your head?
  Did you ever milk this kind of cow?  Well we can do it.  We know how.
If you never did, you should.  These things are fun, and fun is good.
                                                                 -- Dr. Seuss

At 21:15 2004-12-20, you wrote:
>(as to %g - the problem there is it _always_ prints out the
>significant digits... I think it might be useful to print out, say, 3
>digits no matter how many nonzero digits followed.  Or 4 digits.  Or
>5...)

Then I may not have understood the original problem.

"%.17G" (where 17 is the maximum number of significant digits representable 
in a double) sounded close to what you wanted based on the examples.

/Johan

From: Johan Petersson <crm114@...>

   At 21:15 2004-12-20, you wrote:
   >(as to %g - the problem there is it _always_ prints out the
   >significant digits... I think it might be useful to print out, say, 3
   >digits no matter how many nonzero digits followed.  Or 4 digits.  Or
   >5...)

   Then I may not have understood the original problem.

   "%.17G" (where 17 is the maximum number of significant digits representable 
   in a double) sounded close to what you wanted based on the examples.

Almost, but NOT QUITE.

I want to get the format string that prints string A with no loss of
significant information...

... and then print string B with that information.

So, string A is a "format by example", and string B is a "the real data".

Is that clearer?

    -Bill Yerazunis

On Mon, Dec 20, 2004 at 02:29:08PM -0500, Bill Yerazunis wrote:
> Does anyone have any code around that takes a double-precision floating
> point number, and yields the C "printf()" format that expresses it in 
> the fewest characters that doesn't lose significant information?

Seems to me you could binary search for this.

Generate a format string, format, parse, compare?  Is the significant
part there?   Adjust format string accordingly and try again.

Not very time efficient, but do you need it to be?

-- 
Raul

At 21:45 2004-12-20, you wrote:
>    "%.17G" (where 17 is the maximum number of significant digits 
> representable
>    in a double) sounded close to what you wanted based on the examples.
>
>Almost, but NOT QUITE.
>
>I want to get the format string that prints string A with no loss of
>significant information...
>
>... and then print string B with that information.
>
>So, string A is a "format by example", and string B is a "the real data".
>
>Is that clearer?

Very much so. You want the intermediate result of the %G logic expressed as 
a format string.

How about first sprintf:ing with a %G format, then use %f unless the result 
includes "E" and count characters in the string to determine the numeric 
parts of the format string?

/Johan

From: Johan Petersson <crm114@...>

   >
   >I want to get the format string that prints string A with no loss of
   >significant information...
   >
   >... and then print string B with that information.
   >
   >So, string A is a "format by example", and string B is a "the real data".
   >
   >Is that clearer?

   Very much so. You want the intermediate result of the %G logic expressed as 
   a format string.

   How about first sprintf:ing with a %G format, then use %f unless the result 
   includes "E" and count characters in the string to determine the numeric 
   parts of the format string?

That's close....

%G20 would give it in max precision...

Check for an "E" - if so, use e-format
Then start counting digits; work from there.  
But still a bit messy.

It seems almost as much work as doing it in random C code...  

    -Bill Yerazunis

At 22:35 2004-12-20, Bill Yerazunis wrote:
>    How about first sprintf:ing with a %G format, then use %f unless the 
> result
>    includes "E" and count characters in the string to determine the numeric
>    parts of the format string?
>
>That's close....
>
>%G20 would give it in max precision...
>
>Check for an "E" - if so, use e-format
>Then start counting digits; work from there.
>But still a bit messy.
>
>It seems almost as much work as doing it in random C code...

It is messy, inefficient and ugly. But probably not as messy and 
error-prone as doing it directly in C would be; formatting floating-point 
values is a bit hairy.

/Johan