I think, programmer, should write calculator himself
First of all, it is interesting and you can add features, required by your calculation style (e.g. reverse notation, if you love it), second, it is not a very difficult task.
p.s. surely, I wrote my favorite myself, here is list of features, maybe someone will find them useful to implement in their programs:
- almost any number lengths are supported, i.e. 128-bit numbers, 256-bit, etc.
- floating point numbers can be with different mantissa and exponent size, so one can just get 128-bit floating point compatible with IEEE-754, or play with 256-bit floats to see if they will help.
- all bit operations are supported, including popcount and cyclic shift even through carry bit (to emulate CPU instructions), but one can cyclic shift even 256-bit number too
- surely, hex, oct, bin radixes are supported, including limiting representation size, so one can always know how many bytes will be required to represent current number in memory (including floating point format)
- binary complement format is supported, one can change signed and unsigned representation on the fly, this is usefull to check, that e.g. -16=240 in byte types.
- full SNaN, QNaN, machine zero, epsilon, +Inf, -Inf and rounding schemes support for IEEE-754.
- some exotic functionality also presents: e.g. representing floating point number as a continuos fraction, checking for primeness, finding gcd/lcm, etc.
and so on... even if I missed smth required for my work, I just edit sources and add this functionality
p.p.s. not attached here, since calculator is a bit ugly and has no built-in help, and one should remember all its commands.