Revision history [back]

This didn't work at all until recently, so I'm a little surprised. See Trac 11948 for where this happened...

Uh-oh.

sage: erf(i*55.)
1.00000000000000 + 5.64861461531350e1311*I
sage: import mpmath
sage: mpmath.erf(i*55)
mpc(real='0.0', imag='5.648614615313498e+1311')

It's pretty reliably exactly one off in the real direction for pure imaginary input. That would be a problem. Some other testing reveals that it seems to be fine for other input. This seems, on first glance, to be a bug in the upstream system (Pari) that we still use for evaluating this function.

See Trac 1173 and Trac 13050 for switching this evaluation to mpmath.