Very Computer

by **Rupen She** » Sun, 02 Aug 1992 05:13:13

Hi:
Env: Sun Sparc.

In trying to convert a string to a floating number, I don't always
get the exact value.

For example:
float value;
char str[10]; /* Say, str = "+0.02" */

sscanf(str, "%g", &value);
OR
sscanf(str, "%f", &value);

value is 0.0199999995529651642

Is there a strtof() function that can convert a string to a float?
If not, is there another way I can convert a string %f or %g to a
float without these rounding errors?

I am sure there is a way around this that I am not aware of.
Please help !!!
Respond by email to the address below.

Thank you.
---
Rupen Sheth
-------------------------------------------------------------------------------

2550 Beckleymeade Ave. MS 4004 Phone: (214) 708-3344
Dallas, TX 75237 FAX: (214) 708-4827
-------------------------------------------------------------------------------
--
Rupen Sheth
-------------------------------------------------------------------------------

2550 Beckleymeade Ave. MS 4004 Phone: (214) 708-3344

by **John Ellithor** » Sun, 02 Aug 1992 09:02:03

> Nntp-Posting-Host: falcon.ssc.gov
> Hi:
> Env: Sun Sparc.
> In trying to convert a string to a floating number, I don't always
> get the exact value.
> For example:
> float value;
> char str[10]; /* Say, str = "+0.02" */
> sscanf(str, "%g", &value);
> OR
> sscanf(str, "%f", &value);
> value is 0.0199999995529651642
> Is there a strtof() function that can convert a string to a float?
> If not, is there another way I can convert a string %f or %g to a
> float without these rounding errors?
> I am sure there is a way around this that I am not aware of.
> Please help !!!
> Respond by email to the address below.
> Thank you.
> ---
> Rupen Sheth
> -----------------------------------------------------------------------------

> 2550 Beckleymeade Ave. MS 4004 Phone: (214) 708-3344
> Dallas, TX 75237 FAX: (214) 708-4827
> -----------------------------------------------------------------------------

Well, I think one of your key problems is that the value you are printing out
has far more significant figures than is real. For floats, you only have
6 or 7 significant digits, which would then give you 0.02. I would suggest
using doubles. I tried this on my SPARC and it seemed to work fine anyway
I did it, always giving me exactly 0.02. Sorry I couldn't be more help.

John
--

===============================================================================

Dept. of Physics, Rm 26-349 | Phone : (617) 253-3074 Office
Massachusetts Institute of Technology | (617) 253-3072 Lab
Cambridge, MA 02139 | (617) 236-4910 Home
===============================================================================

by **LIU CHEUNG WAH STEPH** » Tue, 04 Aug 1992 02:32:22

How about using scanf() to do the job of strtof()?

by **D'Arcy J.M. Ca** » Mon, 03 Aug 1992 23:23:18

> In trying to convert a string to a floating number, I don't always
> get the exact value.
> [...]
> char str[10]; /* Say, str = "+0.02" */

Not surprising. Since most (all?) floating point formats in C are based
on binary then you would have to create a number based on the reciprocals
of the powers of two. That means you need to build the number by adding
up selected numbers in the last column in the following table:

1/(2 ** 0) = 1.0000000000000000000000000
1/(2 ** 1) = 0.5000000000000000000000000
1/(2 ** 2) = 0.2500000000000000000000000
1/(2 ** 3) = 0.1250000000000000000000000
1/(2 ** 4) = 0.0625000000000000000000000
1/(2 ** 5) = 0.0312500000000000000000000
1/(2 ** 6) = 0.0156250000000000000000000
1/(2 ** 7) = 0.0078125000000000000000000
1/(2 ** 8) = 0.0039062500000000000000000
1/(2 ** 9) = 0.0019531250000000000000000
1/(2 ** 10) = 0.0009765625000000000000000
1/(2 ** 11) = 0.0004882812500000000000000
1/(2 ** 12) = 0.0002441406250000000000000
1/(2 ** 13) = 0.0001220703125000000000000
1/(2 ** 14) = 0.0000610351562500000000000
1/(2 ** 15) = 0.0000305175781250000000000
1/(2 ** 16) = 0.0000152587890625000000000
1/(2 ** 17) = 0.0000076293945312500000000
1/(2 ** 18) = 0.0000038146972656250000000
1/(2 ** 19) = 0.0000019073486328125000000
1/(2 ** 20) = 0.0000009536743164062500000
1/(2 ** 21) = 0.0000004768371582031250000
1/(2 ** 22) = 0.0000002384185791015625000
1/(2 ** 23) = 0.0000001192092895507812500

etc ... depending on how many bits available in the mantissa of your format.
Without an infinite mantissa there are lots of real numbers that cannot be
represented accurately. 0.02 (or 1/50) happens to be one of these.

Quote:> Is there a strtof() function that can convert a string to a float?
> If not, is there another way I can convert a string %f or %g to a
> float without these rounding errors?

strtod() is what you think you are looking for but as I explained this
won't help you. If you need 100% accuracy you need to store the numbers
in a different format. If this is supposed to be money, for example, you
can just use ints or longs with an assumed decimal position. There is
also [shudder] BCD.

--

D'Arcy Cain Consulting | There's no government
Toronto, Ontario, Canada | like no government!
+1 416 424 2871 DoD#0082 |

Very Computer

Is there a strtof() function (str to float)?

Is there a strtof() function (str to float)?

Is there a strtof() function (str to float)?

Is there a strtof() function (str to float)?

Is there a strtof() function (str to float)?