Jeronimo Pellegrini

Hello! It's not a specific problem I'm working on. I was actually using Maxima to verify some math doing in another system, and realized that it wasn't returning the (mathematically) correct answer when the second arg is float. So I reported it. But since you (Stavros) mentioned a wider view of how float math is handled, then I agree! It should either work all the time, or not. Sorry about the noise then.

I had forgotten about rationalize. What I think is that Maxima could be "gentle" enough to detect a bignum in the first argument, then do rationalize on the second, get the result, and turn it back into a float. You'd get 19.0, which is a mathematically correct answer (although not verifiable through IEEE arithmetic, as you mentioned).

I had forgotten about rationalize. What I think is that Maxima could be "gentle" enoigh to detect a bignum in the first argument, then do rationalize on the second, get the result, and turn it back into a float. You'd get 19.0, which is a mathematically correct answer (although not verifiable through IEEE arithmetic, as you mentioned).

Yes - but there probably are ways to get more precise results (that is what I was trying with that reasoning up there). I'll see if I get a MWE for a possible implementation

mod gives incorrect results for bignums

That would likely require using rational arithmetic for large values (so as to be precise enough). Using Taylor expansion, perhaps.

Range reduction for periodic trig functions

Some more updates!