XPost: alt.comp.os.windows-10, alt.comp.microsoft.windows   
   From: nospam@needed.invalid   
      
   On Thu, 12/18/2025 4:43 PM, Winston wrote:   
   > Marian  posted both:   
   >>>   Longitude: -93.81759643000001   
   >  ...   
   >>>   Longitude: -93.81759643000001   
   >   
   > and:   
   >> Does anyone have any idea why the highly insecure Apple WPS database   
   >> contains GPS entries to this illogically numerous set of decimal places?   
   >   
   > Probably just the result of a 64-bit (double) floating point calculation   
   > that they didn't bother to round off.   
   >  -WBE   
   >   
      
   It could be a closest-representable-number problem.   
      
   You should really code a thing like this up. Use the GMP library   
   to do the multiply of mantissa and exponent part. I've only   
   used this once, and not for floats. You can code for C or C++   
   and doing it in C++ is slower (for the temporary copies of   
   numbers the code makes, and when there are a lot of digits in play).   
   The C++ code is easier to read (unlike the C you write in that case).   
      
      https://en.wikipedia.org/wiki/GNU_Multiple_Precision_Arithmetic_Library   
      
   You're right, that the double looks like a reasonable candidate   
   to contributing to the problem. Just need a better calculator to   
   see how close it is to an exact representation.   
      
   https://gregstoll.com/~gregstoll/floattohex/   
      
   -93.81759643   
      
   Single          Mantissa         ExponentContribution   
   0xc2bba29b  1.465899825 0961304    64                  ?   
   0xc2bba29c  1.465899944 30542      64                  ?   
   0xc2bba29d  1.465900063 5147095    64                  ?   
      
   Double                   Mantissa           ExpC   
   0xc05774537ffa0b48  1.465899944 2187497      64     -93.81759642999998   
   0xc05774537ffa0b49  1.465899944 21875        64     -93.81759643000000  <====   
   need more careful conversion   
   0xc05774537ffa0b4a  1.465899944 2187501      64     -93.81759643000001   
      
   1.4658998250961304 * 64 =  \   
   1.46589994430542   * 64 =   \___ Need a calc that takes that many digits   
   1.4659000635147095 * 64 =   /    to verify the single precision representation   
   "?"   
                                    My Microsoft "calc" gives "Invalid Input".   
      
   *******   
      
   It's sloppy workmanship :-)   
   We knew that much before investigating.   
   If your input numbers coming from somewhere are eight digits   
   after the decimal, then the tool output should also be   
   eight digits or so. Printing that many extra digits is silly,   
   as the database collecting these numbers, isn't "doing math" on   
   them. It is just storing a GPS coord as reported (from NMEA   
   message) and puking it back out on demand.   
      
      Paul   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)