yangyushi / nplocate Goto Github PK

The package is not compatible with old numba versions. I should dug out the minimum numba version number and explicitly specify it in the setup file.

Typically, the users get the error message like the following

TypeError: expected dtype object, got 'numpy.dtype[float64]'

when calling the get_gr_pyhs function.

I should spend some time to make see_slice interactive. It would be better if people can manually add some particles with a GUI.

napery might be a good tool. Don't know how hard it is to install it.

Some user with Python version newer than 3.8 might experience following issue when importing nplocate.

>>> import nplocate
Traceback (most recent call last):
 File “<stdin>“, line 1, in <module>
 File “.../python3.8/site-packages/nplocate/__init__.py”, line 4, in <module>
  from .model import GaussianSphere
 File “.../python3.8/site-packages/nplocate/model.py”, line 16, in <module>
  from nplocate import csimulate
 File “nplocate/csimulate.pyx”, line 1, in init nplocate.csimulate
ValueError: numpy.ndarray size changed, may indicate binary incompatibility. Expected 96 from C header, got 88 from PyObject

After some experiment I think this is related to the version of numpy (1.20.x). It needs to be upgraded to 1.22.x, so that the Cython compiled code csimulate.pyx could work.

So, if you see the error, try to upgrade the numpy to latest version.

However, the situation can be harder when numba is taken into consideration. The current numba version (0.54.1) does not support numpy 1.21.x +. Hopefully the numba people will support latest numpy at some point. At the moment, the way to solve the issue seems to be using an older Python (like Python 3.7).

On my own computer the code gave slightly different numerical values, with an order of 10^-4, between different version of python.

• python3.7.7 + scipy 1.5.2 + numpy 1.20.2
• python2.7.16 + scipy 0.16.0 + numpy 1.16.6

This should not be a big problem, but still worrying. The safe option is to always stick to the latest versions of relevant packages.

Supporting python2 is important because it enables people to work with colloids.

To do so I will have to support utf-8 encoding and find a work around for f string