Convert Double to String

[ccubed]

Okay, so I have this code to convert Doubles to Strings.

string stringify(double x){
             
         ostringstream o;
         o << x;
         return o.str();
             
      }

My question is two fold. One, assume the double was originally 1.2E+003. I convert it to a string and i'll get something like XXXX.XX, but convert it back to a double via istringstream and I get 1.2E+003. Why?

Two, is there any way to get all the decimal numbers from the double and not round when it converts to string? I'm trying to create a nice round function so that I can convert a number that could have 20+ decimals to 2 decimals. The process would be really simple if i could get it into a string form and then just loop through each number behind the decimal and see if it's higher than 5.

[cilu]

"was originally 1.2E+003" -> "I get 1.2E+003"
So what's the problem? Isn't that what you are supposed to get?

"number that could have 20+ decimals to 2 decimals"
Double precision is only 15 digits. See this FAQ: http://www.codeguru.com/forum/showthread.php?t=323835. For more you should use a BigNumber implementation, with arbitrary precision.

[Lindley]

It doesn't make sense to talk about scientific notation versus fixed-point notation of a double. When you convert it to a string or output it, that's the *only* time that discrepancy matters. The double itself is simply a value, completely independent of display format.

[olivthill]

Read http://en.wikipedia.org/wiki/IEEE_754 where is given a good explanation of a double.
If you have 0.1 in binary, this is 0.5 in decimal ; 0.01 is 0.25 ; 0.001 is 0.125 ; etc. Sometimes it is not convenient to convert into a decimal value without rounding, e.g. with a value of 1 and an exponent of -1023, the corresponding decimal value will have too many digits to fit on the line of a letter.
Usually applications requiring a great degree of precision are not using doubles.

Edit: Perhaps, I am off-topic. Maybe all you need is sprintf(s, "%.2f", d + 0.005);

[ccubed]

Read http://en.wikipedia.org/wiki/IEEE_754 where is given a good explanation of a double.
If you have 0.1 in binary, this is 0.5 in decimal ; 0.01 is 0.25 ; 0.001 is 0.125 ; etc. Sometimes it is not convenient to convert into a decimal value without rounding, e.g. with a value of 1 and an exponent of -1023, the corresponding decimal value will have too many digits to fit on the line of a letter.
Usually applications requiring a great degree of precision are not using doubles.

Edit: Perhaps, I am off-topic. Maybe all you need is sprintf(s, "%.2f", d + 0.005);

I can't use sprintf. I wish I could.