Consider the simplest case: polynomial substitution, with polynomials that are not units (that is, their constant coefficient is zero).
We can obtain the same result by substituting the actual polynomials in, rather than the polynomials converted to power series.
We can verify that the result is correct by comparing with the result of doing the substitution entirely in the domain of polynomials.
We can do the same with unit power series.
The following example shows that simple arithmetic does not suffice to do the substitution operation: the input power series has rational coefficients, but the result has transcendental coefficients. We compute the power series for from that of .
To underscore the necessity to have analytic expressions in the presence of unit power series, consider the following example, where we omit the analytic expression for the series of .
We want to substitute for and for in . If we first substitute for , then the resulting power series doesn't know its analytic expression, and we obtain an error when we substitute for .
In the other order, the process does work.
This is the order chosen by Maple when the two substitutions are given simultaneously.