LOGEXPAND[TRUE] - causes LOG(A^B) to become B*LOG(A). If it is set to ALL, LOG(A*B) will also simplify to LOG(A)+LOG(B). If it is set to SUPER, then LOG(A/B) will also simplify to LOG(A)-LOG(B) for rational numbers a/b, a#1. (LOG(1/B), for B integer, always simplifies.) If it is set to FALSE, all of these simplifications will be turned off.
LOGSIMP[TRUE] - if FALSE then no simplification of %E to a power containing LOG's is done.
LOGNUMER[FALSE] - if TRUE then negative floating point arguments to LOG will always be converted to their absolute value before the log is taken. If NUMER is also TRUE, then negative integer arguments to LOG will also be converted to their absolute value.
LOGNEGINT[FALSE] - if TRUE implements the rule LOG(-n) -> LOG(n)+%i*%pi for n a positive integer.
%E_TO_NUMLOG[FALSE] - when TRUE, "r" some rational number, and "x" some expression, %E^(r*LOG(x)) will be simplified into x^r . It should be noted that the RADCAN command also does this transformation, and more complicated transformations of this ilk as well. The LOGCONTRACT command "contracts" expressions containing LOG.
(C1) 2*(A*LOG(X) + 2*A*LOG(Y))$ (C2) LOGCONTRACT(%); 2 4 (D3) A LOG(X Y )
If you do DECLARE(N,INTEGER); then LOGCONTRACT(2*A*N*LOG(X)); gives A*LOG(X^(2*N)). The coefficients that "contract" in this manner are those such as the 2 and the N here which satisfy FEATUREP(coeff,INTEGER). The user can control which coefficients are contracted by setting the option LOGCONCOEFFP[FALSE] to the name of a predicate function of one argument. E.g. if you like to generate SQRTs, you can do LOGCONCOEFFP:'LOGCONFUN$ LOGCONFUN(M):=FEATUREP(M,INTEGER) OR RATNUMP(M)$ . Then LOGCONTRACT(1/2*LOG(X)); will give LOG(SQRT(X)).