On Mon, Mar 24, 2008 at 7:01 PM, Richard Fateman
<fateman at cs.berkeley.edu> wrote:
> Re-examining 'great' and friends should be done with extreme caution....
> Changing the actual ordering relations (as opposed to making them faster to
> determine) could have very subtle and wide-ranging effects.
I agree in general that 'great' should be modified with extreme caution.
However, in the specific case of ordering between variables (whether
subscripted or not), I believe great can be modified with impunity,
though of course it will have implications for display (which is
normally in reverse sort order). That is, sort([x, y, x[1], y[1] ])
can return any of the 24 possibilities as long as it is consistent. On
the other hand, constants like %e, %pi, etc. must continue to sort
before variables.
Beyond that, modifying great's functionality for compound expressions
(e.g. a vs. a^2 vs. a^3) is a very delicate matter and any simple
interventions are likely to break something elsewhere in the
simplifier or other existing code.
-s