OK by me, but while we're at it, I think float(x) should yield 'float(x) instead of just x, likewise for bfloat(x). (If we're going to the trouble to make (literal float)*variable work correctly with lazy evaluation, we should do the same for float(variable).) best Robert Dodier