Given an expression, I want to do an operation each time the pattern  f(arg1)+f(arg2) is found by replacing it by f(arg1+arg2). Regadless of how many there are. For example

f(A)+f(B) -> f(A+B) and  f(A)+f(B)+f(C) -> f(A+B+C)  and so on. But here is the catch, there could be anything else in the expressions. These will be left unchanged. 

So f(A)+f(B)+x -> f(A+B)+x

I can do it in Maple only when the input is exactly f(A)+f(B)  when the input is f(A)+f(B)+f(C) and so on.

But this is not practical as I need to make new type for each case.

I need a general way that will work for any expression like in the above example.

I am now using evalindets, but I do not know how to tell it the type for the general pattern of  f(n1)+f(n2)+.....+f(nn) to replace these with f(n1+n2+....nn).

For reference, this is code in Mathematica I am trying to translate to maple.

expr = Sin[x] + f[A] + f[B] + 10*Exp[x]/13 + Cos[x] + f[c] + f[10*c];
expr //. f[a_] + f[b_] :> f[a + b]

In the above //. means repeated replacement. So it will keep replacing the same pattern over and over and this works regardless of where f(a)+f(b) show up. They can be anywhere in the sum.

I wish I can the same in Maple using evalindets. I tried patmatch also, and same problem. Which is how to make it general. This is what I tried

expr:=f(A)+f(B);
evalindets(expr,`&+`('specfunc(f)','specfunc(f)'),F->f( op([1,1],F) + op([2,1],F) ) );

Which works

But to detect expr:=f(A)+f(B)+f(C) it would need new code

expr:=f(A)+f(B)+f(C);
evalindets(expr,`&+`('specfunc(f)','specfunc(f)','specfunc(f)'),F->f( op([1,1],F) + op([2,1],F) + op([3,1],F) ) );

Obviously this approach will not work. It will also fail once a new term is added in between. 

But how to extend this to the general case of

expr:=sin(x)+f(A)+f(B)+10*exp(x)/13+cos(x)+f(C)+f(10*C);

Is there a way to tell Maple to apply the pattern over and over like with Mathematica so it works for general case? I need to try to do this using either patmatch or evalindets. Ofcourse I can do it the hard way, by iterating over the expression and collecting all the f() and add their arguments one by one each `+` subtype. But that is now what I looking for.  

There should be something similar to how it is done in Mathematica, but using Maple command. Notice that the Mathematica example will work regadless of where the f(a)+f(b) shows up.

expr = Sin[x] + f[A] + f[B] + 10*Exp[x]/13 + Cos[x] + f[c] + 1/(f[10*c] + x + f[99])
expr //. f[a_] + f[b_] :> f[a + b]

ps. may be I need to use subsindets['nocache'] need to look more into it.

Are there rules of thumb to follow to decide to use evalindets vs. subsindets?  They seems to do the same thing, but I did not read every details of the help pages.  Is the difference similar to difference between using eval vs. subs? i.e.

          eval(expr,A=2)

vs.

         subs(A=2,expr) ?

For example

expr:=ln(A)+ln(B);
evalindets(expr,`&+`('specfunc(ln)','specfunc(ln)'),f->ln(op([1,1],f)*op([2,1],f)));
subsindets( expr, `&+`('specfunc(ln)','specfunc(ln)'), f->ln(op([1,1],f)*op([2,1],f)));

Both give

in my program, I keep assumptions in a set. Sometimes this is empty if no assumptions are used. This never caused a problem before (at least I do not think so, else I would have seen it) when using empty {} in assuming, except for now.

Here is one example below. Is this a known problem? I noticed when changing {} to [] the error goes away. I am not sure why, and if this is known issue. But will change from a set to a list to avoid this. 

Maple 2022.2 on windows 10

Download feb_14_2023.mw

What is the logic behind this?

expr:=(c[2]+x)^3;
simplify(expr)

Which is what expected. But 

expr:=(c[2]+x)^3+a;
simplify(expr)

gives

Which does not look simpler to me. I expected it to be the same as before but with "a" added.  This is what Mathematica gives for comparison

I know I can use simplify with size option. But my question is, how did Maple decide that x^3 + 3*x^2*c[2] + 3*x*c[2]^2 + c[2]^3 + a is "simpler" than (c[2] + x)^3 + a ? It must use some logic which I am trying to understand.

Maple 2022.2 on windows 10

I've probably asked about this long time ago but I do not remember now. But it is still a problem any way in 2022.2

Compare these two results

code is

restart;

alias(seq(c[k] = _C||k, k = 0..10));
c[1]*h(x);
Int(c[1]*h(x),x);


restart;
c[1]*h(x);
Int(c[1]*h(x),x);

I tried with typesetting level as extended and standard and same result.

Is this a know issue? It is not a big deal as it only affects display but it will be nice to find why it happens when using alias.