a) SAS uses missing values to represent NaNs. In statistics, an observation can be a NaN if it is missing for a statistical reason (for example, a survey question was left unanswered) or for a mathematical reason (the log of a non-positive number). SAS supports 28 missing values: see http://analytics.ncsu.edu/sesug/2005/TU06_05.PDF for details.
Some people assign the .P missing value to mean "positive infinity" and .N to mean "negative infinity," but most don't. To me a NaN is a NaN.
b) MATLAB licenses (or bought?) the Maple symbolic algebra engine. No, SAS does not include symbolic integration or differentiation. Some SAS procedures support automatic differentiation in order to compute the gradients and Jacobians of objective functions, but that feature is not universal.