omega() method does not respect precision of object

[pjbruin]

In the following example, the precision of E.omega() is too low:

>>> from cypari2 import Pari
>>> P = Pari()
>>> P.set_real_precision_bits(128)
>>> E = P('ellinit([0, 0, 0, 1.0, 0])')
>>> E.omega()
[3.708149354602743838, 1.8540746773013719187 - 1.8540746773013719187*I]

Also, elliptic curves over number fields are not supported:

>>> E = P('ellinit([0, 0, 0, 1, 0], nfinit(y^2 - 2))')
>>> E.omega()
Traceback (most recent call last):
...
cypari2.handle_error.PariError: the PARI stack overflows (current size: 8000000; maximum size: 8003584)
You can use pari.allocatemem() to change the stack size and try again

Both problems can be fixed by using the PARI function member_omega instead of ellR_omega to define the omega() method.

The only disadvantage of this solution is that the precision parameter will be ignored. This means that one has to set the global precision to compute periods of elliptic curves over number fields (like one has to do in GP) instead of passing the precision parameter to omega().

[jdemeyer]

The only disadvantage of this solution is that the precision parameter will be ignored.

That's strange. I wonder whether there is a fundamental reason for that.

[jdemeyer]

More generally, we don't have a good solution for dealing for member functions so far, see #51

[jdemeyer]

The only disadvantage of this solution is that the precision parameter will be ignored.

Not quite ignored. It will use the precision set by ellinit(), which is reasonable.