Float underflow exception: what is it and how to cope

[tsh73]

EDIT
*I had a forum glitch making me think I lost that post. So I recreated it with more info, look below*

---

sometimes then your program calculates functions involving powers it breaks with error
"Float underflow exception"

What is it?
It is condition happening then math results goes somewhere under 1e-300.
Here how you can get one:

y=10.0^-308
print y

(powers up to 307 works OK in this case)

Is it supposed to mean something?
I dunno, some other languages just allow such small a number change to 0.0

How to fix/ avoid it?
enclose offending part in a function
catch error with on error
return 0 if it is indeed "Float underflow exception"

Here's an example
pi = acs(-1)
h=pi/300
t=25*h

'print (cos(6*t)*cos(18*t))^10 'dies with Float underflow exception
'now wrap that in a function and catch error there
print y(t)

end

function y(t)
on error goto [catch]
'y = 1/5*(cos(6*t)*cos(18*t))^10
y = 10.1^500

exit function
[catch]
if Err$="Float underflow exception" then y = 0:exit function
'else die naturally
on error goto 0
resume ' - will not do, loops
'print Err, Err$
end
end function

[tsh73]

Float underflow exception: what is it and how to cope

Sometimes program involving math (drawing wonderful plots for example) dies with "Float underflow exception".
What is it?
it is a notinn that result of calculation got out of limits presentable bu Double data type
And it goues out of limits not in "big" part but rather very very close to 0.
Here it is:
print 10.0 ^-307 'still OK
print 10.0 ^-308 'Float underflow exception

Now, does it supposed to mean something?
I dunno
some languages I tried just silently change result to 0.0
If you are trying to draw interesting plot, that
(rather then breaking the program)
would be desired behavior.

So how to avoid it?
* wrap offending calculation in a function
* catch error there
* return 0

Now, you can catch this very error and return 0, or break program (: END) otherwise.
(like in example y(t) does)
Or you can just catch error and ignore it
(like in example y2(t) does)
For *this* error, it would default to returning 0
But in case you ever get other types of error (
divide by zero, float oveflow, argument of a function out of range
)
program will just return 0. And you'll never know that
So I'd rather use longer method, just in case.


pi = acs(-1)
h=pi/300
t=25*h

print (cos(6*t)*cos(18*t))  '-0.11329165e-30, really really small

'print (cos(6*t)*cos(18*t))^10   'dies with Float underflow exception

'now, wrap that in a function, catch error there, return 0
print y(t)
print y2(t)


function y(t)
   on error goto [catch]
   y = 1/5*(cos(6*t)*cos(18*t))^10
   exit function
[catch]
   if Err$="Float underflow exception" then y = 0: exit function
   'else die
   print Err, Err$
   end
end function

function y2(t)
   on error goto [catch]
   y = 1/5*(cos(6*t)*cos(18*t))^10
   exit function
[catch]
'ignore error
   'it returns 0 by default
end function


Now, the program I found it in
justbasiccom.proboards.com/thread/916/parametric-plot-maple