Am 30.12.2015 um 23:51 schrieb TomD <tomdiscus@xxxxxx>: (x - 0.3).abs < 1e-4 // 1e-4 = 1 * (10 ** -4) = 0.0001 Not in general, if you make a test if a number x is near 0.3, you cannot assume that x is larger: e.g. 0.29 - 0.3 < 1e-4 gives true, but it doesn't say, that 0.29 is in a "0.0001-environment" of 0.3: in fact it is not, as the "distance" is 0.01 If you look at the source code of equalWithPrecision, you see that it does exactly this kind of check (absdif is another method for taking the absolut value of the difference): equalWithPrecision { arg that, precision=0.0001; ^absdif(this, that) < precision } *Ok interesting method thanks, I might replace my choice with this but it's Indeed it is problematic to define a general threshold for all magnitudes of possible numbers. E.g. the default precision 0.0001 fails here: ((0.1 + 0.1 + 0.1) * 1e20).equalWithPrecision(0.3 * 1e20) --> false I am not aware if something exists already in SC (does it maybe ?), but a equality check with relative precision could be easily defined: // Add a sc-file to the Extensions folder and recompile +SimpleNumber { equalWithRelativePrecision { arg that, relativePrecision = 0.0001; ^absdif(this, that) < (this * relativePrecision).abs } } Then we can compare with a relative precision default of 0.0001: ((0.1 + 0.1 + 0.1) * 1e20).equalWithRelativePrecision(0.3 * 1e20) --> true Greetings Daniel |