I think you've indirectly answered Williams question, "when is it time to start a new thread."

I think you've indirectly answered Williams question, "when is it time to start a new thread."

Why do you think it isn't working? It looks correct to me. You might change the axes to boxed so that you can see the path from the origin to the point (-1,0). Here's the same plot using Re and Im:

plot([(Re,Im)(F(t)),t=0..2], axes=boxed);

Why do you think it isn't working? It looks correct to me. You might change the axes to boxed so that you can see the path from the origin to the point (-1,0). Here's the same plot using Re and Im:

plot([(Re,Im)(F(t)),t=0..2], axes=boxed);

No, I didn't see any more. With a builtin function, which max is, that is usually all you are going to see. I did mention that max calls evalr. You can see that by doing

restart;
printlevel := 30:
max(Pi,3);

Assign printlevel a larger integer to see more details. See ?printlevel.

No, I didn't see any more. With a builtin function, which max is, that is usually all you are going to see. I did mention that max calls evalr. You can see that by doing

restart;
printlevel := 30:
max(Pi,3);

Assign printlevel a larger integer to see more details. See ?printlevel.

How is that different from max? It also eliminates all but the largest realcons.

How is that different from max? It also eliminates all but the largest realcons.

Sounds reasonable. I'm trying to figure out what it is you want to accomplish. Seems like the following might be problematic:

agree(100,99) -> 0
agree(100,199) -> 1
agree(100,109) -> 2

The number of digits that agree is *not* inherently related to the error.

Sounds reasonable. I'm trying to figure out what it is you want to accomplish. Seems like the following might be problematic:

agree(100,99) -> 0
agree(100,199) -> 1
agree(100,109) -> 2

The number of digits that agree is *not* inherently related to the error.

What should be the result for agree(1e3$2)? How about agree(1e-3$2)?

What should be the result for agree(1e3$2)? How about agree(1e-3$2)?

I missed the part about that being an example, but you figured out the idea, which was my point. The flat option speeds things up if you don't have to recurse into a structure, that is, if the outermost match to the type is what you want. Here's a simple quiz, replace (actually construct) your operator with a call to curry. A trickier quiz is, use rcurry instead of curry.

I missed the part about that being an example, but you figured out the idea, which was my point. The flat option speeds things up if you don't have to recurse into a structure, that is, if the outermost match to the type is what you want. Here's a simple quiz, replace (actually construct) your operator with a call to curry. A trickier quiz is, use rcurry instead of curry.

The help page type/numeric defines a numeric type:

  - The type(x, numeric) function returns true if x is an integer, fraction, or
    finite floating-point number (float).

Pi is none of those, hence not of type numeric. However, it is of type realcons, so you could change the type specification in your Max procedure to realcons and avoid the type matching error. However, that doesn't solve your problem:

  if Pi < 3 then yes else no end if;
Error, cannot determine if this expression is true or false: Pi < 3

You could use evalf in the conditional to evaluate the relation, however, doing so can introduce floating-point round-off issues. The builtin function max handles that by using evalr, which uses range arithmetic.