Hi,
i would prefer to raise an error promoting the use of change_ring (same
for matrices), which allows more tuning on the base ring.
By the way, note that the best approx with floating-point coefficients
(e.g. for the sup norm on [-1,1]), is not necessarilly obtained with the
best approx on each
Dear devs,
Is it fine if I implement numerical_approx on polynomials by calling
it recursively on coefficients? For example, we would have
sage: x = polygen(QQ)
sage: p = 1343235439458/23 * x^2 - 234 / 143425*x + 1432/6512
sage: p.numerical_approx()
5.8401540846e10*x^2 - 0.00163151472895241*
