Division in CAS

[matalog]

I was writing a long integer divider program, and was testing a value to see if it was working, and it was right, but it looks like the Prime was wrong, in CAS. I know this is very unlikely, and I hope someone will explain that it is down to the limited amount of information stored about the number, resulting in the wrong number being displayed after the place where E was.

I calculated 123456789123456789/127, in CAS mode. It gave the result 123456789123456789/127, then I pressed a b/c and it gave 9.72100701759E14 and I then pressed a b/c again, which resulted in the value that is not correct, which is 972100701759492.

I have just tried another 147258369147258369/127 and pressing a b/c twice again results in 1159514717694936, which is wrong by 11! The answer could have been as accurate as 1159514717694947.787401574803149606299212598425196850393700787401574803149606299​2125984251968503937007874015748031496.

Is there any way to get full length division results in CAS, just like the long results from multiplication? Or a way to get more accurate results from CAS?

[Nigel (UK)]

You can use the iquo() CAS function to get an exact integer result for a quotient of two integers. There is also a function irem() that gives the remainder.

Nigel (UK)

[Albert Chan]

[a b/c] key turning exact fraction to approximation may be a bug. (if exact mode is ON)

Number once turned approximate, may not be able to revert back to exact form.
exact() does not really mean "exact", only rationalize the approximated number

CAS> exact(approx(π))      → 4272943/1360120

Also, in CAS, approx numbers shown may not match what is internally stored.

CAS> approx(123456789123456789/127)      → 9.72100701759e14
CAS> Ans-9.72100701759e14                       → 492.

972100701759492 is stored as hex-float 0x1.ba0f6a81d4020p+49