This release includes contributions by @waltertross. Below are his comments:
The output format was %g, which is equivalent to %.6g and is good for floats, but the code uses doubles, which have at least 15 digits of precision. Therefore the output format is %.15g now. The inverse trigonometric functions were missing but are important. In the example words file, the value of pi has been set to its full double precision, by taking it from math.h. The erf (error function) completes the set of useful mathematical functions that cannot be derived from other words. The erf is an important function in statistics, where it is linearly related to the normal (i.e., Gaussian) cumulative distribution function. See also the "68–95–99.7 rule".