问题
I'm trying to convert Single to Double while maintaining the original value. I've found the following method:
Single f = 5.2F;
Double d1 = f; // 5.19999980926514
Double d2 = Double.Parse(f.ToString()); // 5.2 (correct)
Is this practice recommendable? I don't need an optimal method, but the intended value must be passed on to the double. Are there even consequences to storing a rounded value in a double?
回答1:
The conversion is exact. All the Single values can be represented by a Double value, because they are "built" in the same way, just with more possible digits. What you see as 5.2F is in truth 5.1999998092651368. If you go http://www.h-schmidt.net/FloatConverter/IEEE754.html and insert 5.2 you'll see that it has an exponent of 2^2 (so 4) and a mantissa of 1.2999999523162842. Now, if you multiply the two numbers you'll get 5.1999998092651368.
Single have a maximum precision of 7 digits, so .NET only shows 7 digits. With a little rounding 5.1999998092651368 is 5.2
回答2:
You could use "decimal" instead of a string.
float f = 5.2F;
decimal dec = new decimal(f);//5.2
double d = (double)dec; //5.2
回答3:
If you know that your numbers are multiples of e.g. 0.01, I would suggest that you convert to double, round to the nearest integer, and subtract that to get the fractional residue. Multiply that by 100, round to the nearest integer, and then divide by 100. Add that to the whole-number part to get the nearest double representation to the multiple of 0.01 which is nearest the original number.
Note that depending upon where the float values originally came from, such treatment may or may not improve accuracy. The closest float value to 9000.02 is about 9000.019531, and the closest float value to 9000.021 is about 9000.021484f. If the values were arrived at by converting 9000.020 and 9000.021 to float, the difference between them should be about 0.01. If, however, they were arrived at by e.g. computing 9000f+0.019531f and 9000f+0.021484f, then the difference between them should be closer to 0.02. Rounding to the nearest 0.01 before the subtract would improve accuracy in the former case and degrade it in the latter.
来源：https://stackoverflow.com/questions/29492878/convert-float-to-double