[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

# Re: summation problem

```Paul van Delst wrote:
> Jaco van Gorkom wrote:
> > Manish wrote:
> >  ...
> > > I have a program which produces an array of values of sunlight flux during
> > > the day.  Unfortunately, it produces crazy numbers before the sunrises and
> > > after it sets(as expected).  I'm summing the values throughout the day to
> > > get a total integrated day flux, but here's the problem - is there a way of
> > > telling the TOTAL function to ignore negative numbers and NaN numbers?
> >
> > IDL> print, total(test>0,/nan)
> >       7.00000
> > % Program caused arithmetic error: Floating illegal operand
> >
> > The 'illegal operand' error appears to be harmless, caused by comparing test>0:
>
> Harmless maybe, but I for one don't like seeing illegal operand errors, let alone ignoring
> them. What if you add some code that tries to take the log of a -ve number, see the same
> error and shrug it off?

That is *exactly* what I'd do: shrug it off! The log of a -ve number gives me
-Inf as a
result, which will propagate through any calculation to be NaN, Inf, or -Inf.
Just another
missing data point.

> loc_finite = WHERE( FINITE( test ) EQ 1, count_finite )
> IF ( count_finite GT 0 ) THEN \$
>   sum = TOTAL( test[ loc_finite ] > 0.0 ) \$
> ELSE \$
>   sum = !VALUES.F_NAN  ; Or some other suitable flag

Of course I see your point. I would probably implement a FINITE() check if I
were writing a
full program, but the annoying special-case programming for when WHERE() returns
-1 held me
back here.
On second thought: WHERE() could only return -1 for a day without sunrise...

groetjes,
Jaco

>  ...
> Oh, and make sure you set !EXCEPT = 2 in your idl setup file. That'll learn ya to remove
> errors from your code :o)
Geez! Thanks for pointing that out! This solves everything Manish: !EXCEPT = 0.
```