@ecterrab   this is excellent news. I have installed the fix and it does indeed solve the problem. Thanks!  

@Carl Love   . Thanks.  I checked your code and indeed it does exactly what I needed. 
But I must confess it will take me longer to understand it.  For the time being I will 
use it verbatim. Many thanks. 

@Carl Love thanks.   Here I have a more concrete example: I need to make the substitution

\int(f(x,y)) dy = g(x)

where x and y may have different names. This means, the expression I have has terms that look 
like: 

\int(f(u,v)) dv,   \int(f(x1,u)) du,  \int(f(x2,z)) dz,  etc.

I would like them to be subtituted by g(u), g(x1), g(x2) respectively. 

How can I do this?   Thank you!

Thank you for your help. What you wrote is exactly what I needed. 

@Carl Love This is great. My post refer to a toy model of a much longer problem.  I will study your new proposal (my knowledge of Maple is still very limited). Thank you!  

@Carl Love thank you, I understand. 

@Carl Love thank you! 

@Carl Love thank you for your comment. I now understand the difference between eval and subs . 

But what I was trying to undertand in my previous post was why inserting or not the command 
define(G, ordeless)  resulted in diferent behaviors:

Without inserting define(G, ordeless)  the commands subs and eval gave the same result in my particular problem. 
After inserting  define(G, ordeless)  they gave different results.

I was wodering if there was a simple explanation for this apparent correlation between define(G, ordeless)/subs/eval. 

Max

@vv  Thanks, this tip did solve my problem. 

Is there a quick way to understand why in this example eval is correct and subs is not? 
(I am new to Maple programming and trying to understand)

Thank you! 

@Carl Love Thanks Carl, I have learned a lot today:

You are right, the define command is also problematic. My second problem was exactly related to it. 

I have made a second post under the name Unexpected behavior of define(orderless). 
It is a simple example where define gives unexpected (and unwanted) results.

I need to figure out how to define a function with the property f(x,y) = f(y,x) without using define
 

@Carl Love Thanks a lot. lprint shows very clearly where the problem is.  My next step is to find a solution!  

What would be an alternative to typematch? More generally, your advice is not to use "that whole family of commands". 
What is that family?  How can I separate "good" from "bad" commands?  (Other than hitting a problem....)

I often need to identify parts of expressions. For example, the integrand from

1/2*Int (f(x) , x)

I just learned the command GetIntegrand (from the IntegrationTools package). But this command extracts f(x) leaving the 1/2 outside. It does not work very well with double integrals either.   I thought typematch was the perfect command to extract parts of expressions.

As a matter of fact, I am having a similar problem with evalb inside Categorize: two expression look identical to 
evalb but when applied inside Categorize appear different. I know suspect that the root of the problem is the same. 
Is Categorize also a "bad" command?

Thanks a lot for your help.  Max 

@Carl Love I have included the file in the first line of the previous post. Any hekp will be welcome! 
Thanks, Max