How to Define High-precision Date type

[Donkey]

Hi,
The precision of built-in date type of C is very low. Even using long
double float type or double float type, we can only use 12 or 16 digits
after the decimal point. What can we do if we want to use
high-precision number such as the number with 50 digits after the
decimal point. Can we define a user date type?

[Walter Roberson]

In article <11************ **********@i39g 2000cwa.googleg roups.com>,
Donkey <ne********@hot mail.com> wrote:

The precision of built-in date type of C is very low. Even using long
double float type or double float type, we can only use 12 or 16 digits
after the decimal point. What can we do if we want to use
high-precision number such as the number with 50 digits after the
decimal point. Can we define a user date type?

No, if you want longer precision that is provided by the compiler,
then you either have to program the calculation routines yourself,
or else find a pre-written extended precision library and use it.

--
Programming is what happens while you're busy making other plans.

[Keith Thompson]

"Donkey" <ne********@hot mail.com> writes:

The precision of built-in date type of C is very low. Even using long
double float type or double float type, we can only use 12 or 16 digits
after the decimal point. What can we do if we want to use
high-precision number such as the number with 50 digits after the
decimal point. Can we define a user date type?

Use a library such as GNU GMP (google it).

--
Keith Thompson (The_Other_Keit h) ks***@mib.org <http://www.ghoti.net/~kst>
San Diego Supercomputer Center <*> <http://users.sdsc.edu/~kst>
We must do something. This is something. Therefore, we must do this.

[Martin Ambuhl]

Donkey wrote:

Hi,
The precision of built-in date type of C is very low. Even using long
double float type or double float type, we can only use 12 or 16 digits
after the decimal point. What can we do if we want to use
high-precision number such as the number with 50 digits after the
decimal point. Can we define a user date type?

The average distance from the sun to Earth is about 149.598e6 km.
Supose DBL_DIG to be 15. A (1e-15 * sun_to_earth_di stance) is about .15 mm.
Be careful is assuming stating that 15 significant digits ("digits after
the decimal point" is meaningless) is "very low". When 15 significant
decimal digits allows expressing the average distance from the sun to
the earth to with 0.15 mm, that is high accuracy, indeed.

The place that greater precision becomes meaningful is in intermendiate
results in some arithmetic operations. You might want higher precision
in, say, matrix inversion. There are several things to say about this:
1) long double *might* be sufficient, or
2) many packages exist for doing arithmetic with greater precision, or
3) you are making a serious mistake.
Point (3) is significant. If I found my problem required values beyond
the capabilities of long doubles (18 significant figures and a range of
1e-4391 to 1e4392 on my implementation) , I would seriously consider
whether I had pathologogical data or an algorithm that needed fixing.

[jacob navia]

Donkey a écrit :

Hi,
The precision of built-in date type of C is very low. Even using long
double float type or double float type, we can only use 12 or 16 digits
after the decimal point. What can we do if we want to use
high-precision number such as the number with 50 digits after the
decimal point. Can we define a user date type?

The lcc-win32 compiler offers the qfloat data type with 103 digits
precision and 350 bits
#include <stdio.h>
#include <qfloat.h>
int main(void)
{
qfloat s = sinq(1.0q);
printf("%118.10 9qf\n",s);
}
Output:
0.8414709848078 965066525023216 302989996225630 607983710656727 517099919104043 912396689486397 435430526958543 49

[Spoon]

Donkey wrote:

The precision of built-in date type of C is very low. Even using long
double float type or double float type, we can only use 12 or 16
digits after the decimal point. What can we do if we want to use
high-precision number such as the number with 50 digits after the
decimal point. Can we define a user date type?

http://www.google.com/search?q=arbitrary+precision

http://en.wikipedia.org/wiki/Arbitra...ion_arithmetic

http://www.mpfr.org/

[Mark McIntyre]

On Fri, 31 Mar 2006 06:45:30 GMT, in comp.lang.c , Martin Ambuhl
<ma*****@earthl ink.net> wrote:

Be careful is assuming stating that 15 significant digits ("digits after
the decimal point" is meaningless) is "very low". When 15 significant
decimal digits allows expressing the average distance from the sun to
the earth to with 0.15 mm, that is high accuracy, indeed.
OTOH, when buying a couple of tonnes of Platinum in Tunisian Dinar,
its easy to rapidly lose all 15 digits, and the discrepancy is quite
expensive... :-)
The place that greater precision becomes meaningful is in intermendiate
results in some arithmetic operations.
This is absolutely true.
If I found my problem required values beyond
the capabilities of long doubles (18 significant figures and a range of
1e-4391 to 1e4392 on my implementation) , I would seriously consider
whether I had pathologogical data or an algorithm that needed fixing.

Or you work in astronomy or finance...
Mark McIntyre
--
"Debugging is twice as hard as writing the code in the first place.
Therefore, if you write the code as cleverly as possible, you are,
by definition, not smart enough to debug it."
--Brian Kernighan

[Martin Ambuhl]

Mark McIntyre wrote:

On Fri, 31 Mar 2006 06:45:30 GMT, in comp.lang.c , Martin Ambuhl
<ma*****@earthl ink.net> wrote: [...]

If I found my problem required values beyond
the capabilities of long doubles (18 significant figures and a range of
1e-4391 to 1e4392 on my implementation) , I would seriously consider
whether I had pathologogical data or an algorithm that needed fixing.

Or you work in astronomy or finance...

Nonsense. Astronomers are often ecstatic with data of much less
significance than is guaranteed even for a float. Consider that the
relative error for G, the Newtonian gravitational constant, is 1.5e-4 in
the 2002 CODATA values. Astronomy is an extremely poorly chosen example
for your purposes.

Nonsense. No sane person is using floating point types for finance, in
the first place. Such a person will be fired before the day is out,
anyway, so the number of significant digits is irrelevant.

Let's assume a less than sane person did use long doubles of my
implementation for finance. It is unlikely that such a person would
deal in numbers larger than those for the United States government. The
estimated revenues for 2011 (the largest numbers available) are
$3.034861e12 ($2.233286e12 on budget) and estimated expenditures are
$3.23879e12 ($2.648669e12 on budget). These numbers are all expressible
to the penny in 15 decimal digits. They would need to be 1000 times as
large to cause concern with 18 significant digits. Of course, the Bush
experience has taught us that these numbers might not even have one
significant digit. Finance is an extremely poorly chose example for
your purposes.

[Mark McIntyre]

On Fri, 31 Mar 2006 22:55:53 GMT, in comp.lang.c , Martin Ambuhl
<ma*****@earthl ink.net> wrote:

Mark McIntyre wrote:

Or you work in astronomy or finance...

Nonsense. Astronomers are often ecstatic with data of much less
significance than is guaranteed even for a float.

Oh, sure. Apparently you don't do much spectroscopy or
astronavigation .
Consider that the
relative error for G, the Newtonian gravitational constant, is 1.5e-4 in
the 2002 CODATA values.
This is a /physically observed/ constant. Its by definition not
accurate.
Astronomy is an extremely poorly chosen example
for your purposes.
Apparently you're not an astronomer.
Nonsense. No sane person is using floating point types for finance, in
the first place. Such a person will be fired before the day is out,
anyway, so the number of significant digits is irrelevant.
Go tell that to the thousands of people using it every single day in
the City, New York .Tokyo, Frankfurt. etc.

Sorry, but you really have absolutely no idea what you're talking
about. I've worked in the City for 15 years now, and I do.
It is unlikely that such a person would
deal in numbers larger than those for the United States government.
Rolls on the floor laughing. A few words to play with: leverage,
options, exchange rates. Do you have any idea the total value of FX
and FX options traded last year?
Finance is an extremely poorly chose example for your purposes.

Actually, since I was pointing out that often 15sf isn't enough
precision, I think its a very good example.

Mark McIntyre
--
"Debugging is twice as hard as writing the code in the first place.
Therefore, if you write the code as cleverly as possible, you are,
by definition, not smart enough to debug it."
--Brian Kernighan

[stathis gotsis]

"Mark McIntyre" <ma**********@s pamcop.net> wrote in message
news:io******** *************** *********@4ax.c om...

On Fri, 31 Mar 2006 22:55:53 GMT, in comp.lang.c , Martin Ambuhl
<ma*****@earthl ink.net> wrote:

Mark McIntyre wrote:
Consider that the
relative error for G, the Newtonian gravitational constant, is 1.5e-4 in
the 2002 CODATA values.

This is a /physically observed/ constant. Its by definition not
accurate.

Well, many physical constants are meausred through experiments. I think the
relatively low accuracy in the measurement of the gravitational constant is
due to the fact that gravitational forces are rather weak.

Astronomy is an extremely poorly chosen example
for your purposes.

Apparently you're not an astronomer.

Nonsense. No sane person is using floating point types for finance, in
the first place. Such a person will be fired before the day is out,
anyway, so the number of significant digits is irrelevant.

Go tell that to the thousands of people using it every single day in
the City, New York .Tokyo, Frankfurt. etc.

Sorry, but you really have absolutely no idea what you're talking
about. I've worked in the City for 15 years now, and I do.

I apologize for my ignorance: City==London?