Remove support for np.exp

I am really confused of this exponentiation that I expected as exp(-1j X).

>>> np.exp(-1j*cirq.X(q))      
cirq.rx(np.pi*0.3183098861837907).on(cirq.GridQubit(0, 0))

In fact, the definition of cirq.rx® is cirq.XPowGate(exponent=r/np.pi, global_shift=-0.5). Because cirq.XPowGate(exponent=t) is [[g·c, -i·g·s], [-i·g·s, g·c]] = g ( I c - i X s) where: c = cos(π·t/2) = cos(r/2) s = sin(π·t/2) = sin(r/2) g = exp(i·π·t/2) = exp(i r/2) since the global_shift = -0.5, exp(i * pi * global_shift * exponent) = exp(i * pi * -0.5 * r/pi) = exp(-i0.5r) = g’ = 1/g, so the final equation is g’g ( I c - i X s) = I cos(r/2) - i X sin(r/2)

However, when I remember the Euler formula like identity equation, exp(-1j r X) = I cos® - i X sin®. Could you help me?

[kevinsung]

At Cirq sync we decided to drop this functionality (support for np.exp).

[dabacon]

There is no longer any exp methods on gates in Cirq, so this is now fixed as far as I can tell.