Sergey B Kirpichev

With some refactoring of mpmath.libmp imports in the diofant (https://github.com/diofant/diofant/pull/1357 and https://github.com/diofant/diofant/pull/1361) I got this list: ``` $ python ~/s.py . ['ComplexResult', 'MPZ', 'MPZ_ONE', 'NoConvergence', 'dps_to_prec', 'fhalf', 'finf', 'fnan', 'fninf',...

> what SymPy really needs is pure functions like mpf_add that take precision as an argument rather than anything based on contexts or global precision settings. Otherwise it is just...

> It looks like you are preparing for a new release of mpmath. No. It's far, far away... Some issues I just can't solve alone, e.g.: #709 But I think...

> It looks correct on PyPI to me, so probably people are using --pre: Or add a specific version requirements. But not in this case, unfortunately. > I'm not sure...

> I am making changes to sympy to account for this (https://github.com/sympy/sympy/pull/26269) and aim to put out a sympy 1.13 release soon. https://github.com/mpmath/mpmath/pull/769 will include some compatibility code for this....

It seems, this is the libmp namespace, used both SymPy (since v1.12) and latest Diofant (``diofant_imports``, ``sympy_imports`` and ``sympy_112_imports`` are from run of ``find-libmp.py`` script on the latest Diofant, SymPy...

> In SymPy mpmath's PythonMPContext is subclassed (I would like to remove this). Good idea. After all, this class seems to be a private one. It's not documented or exported...

You can use ``np.cross()``. An example: ```pycon >>> import numpy as np >>> from mpmath import mp >>> np.cross(mp.matrix([[1, 2, 3]]), mp.matrix([[4, 5, 6]])) array([[mpf('-3.0'), mpf('6.0'), mpf('-3.0')]], dtype=object) ```

I think this doesn't make sense as an enhancement request.

FYI: numpy now deprecate poly1d with the mpmath-like coefficient ordering. I would welcome same change in the mpmath.