Strange error



Rene Grothmann wrote:
> Why is
> 
> integrate(exp(t)*(1+t^2),t,0,a)
> 
> asking, if a is negative, zero, or positive? However
> 
> integrate(exp(t)*(1+t^2),t,a,b)
> 
> works.
> 

The short answer is that maxima needs to know to figure out how to do 
the integral. :-)

In the first case, it wants to know if it can transform the integral 
from 0 to a into the range 0 to 1, and also if the integrand has a 
certain form.  This form only works if the upper limit is positive.  In 
this particular example, the integrand has the wrong form, so the 
transformation can't be used anyway.

After that, a different method is tried, which is the same method that 
is used directly for the second integral, since the lower limit isn't 0.

I suppose the test could be changed to check for the form of the 
integrand before checking for the sign of the upper limit.  Might fix 
this, but might introduce other issues.

Ray