On 2013-05-22, Barton Willis <willisb at unk.edu> wrote:
> (%i1) tellsimp(b+1,z);
> tellsimp: warning: putting rules on '+' or '*' is inefficient, and may not work.
> (%o1) [+rule1,simplus]
> Not OK -- b+1 isn't automatically simplified to z:
>
> (%i4) b+1;
> (%o4) b+1
>
> (%i5) expand(%,0,0);
> (%o5) z
The code generated by tellsimp for "+" and "*" doesn't make use of PART+
and PART*, respectively, which treat them as commutative operators, and
tellsimp rules are applied before built-in rules, so terms have not yet
been reordered. So the rule generated here is sensitive to the order in
which you enter the terms -- try 1 + b instead of b + 1.
I suppose it is plausible that "+" and "*" cannot be considered
commutative until the built-in rules have been applied. Maybe, maybe not.
From time to time I think I should rework tellsimp to generate the same
stuff for "+" and "*" as tellsimpafter (which applies PART+ and PART*).
An easier fix in the short term is to clarify the warning message --
"may not work" --> "operator is treated as noncommutative".
best
Robert Dodier