numpy's design is to work with arrays and carries that much overhead, that is numpy needs to detect whether the argument is an array or an integer.

A quick benchmark shows that, for single numbers, math.sqrt is faster, but numpy.sqrt is a lot faster when dealing with arrays. If I'm not too tired, I see a factor of 10. Sadly many places within https://github.com/RedFT/Hexy/blob/1057e5df5962627c52c5bf92742ebc94204a6b9d/hexy/hexy.py uses numpy's function on simple integers, including my recent contributions which were based on existing code.

In [1]: import math

In [2]: import numpy as np

In [3]: %%timeit N = 10000
   ...: for i in range(N):
   ...:     z = math.sqrt(i)
   ...: 
1.45 ms ± 25.1 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each)

In [4]: %%timeit N = 10000
   ...: for i in range(N):
   ...:     z = np.sqrt(i)
   ...: 
16.8 ms ± 103 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)

In [5]: %%timeit R = [i for i in range(10000)]
   ...: np.sqrt(R)
870 µs ± 981 ns per loop (mean ± std. dev. of 7 runs, 1000 loops each)

The install with pip works perfectly 👌 , one issue we're running into is the number of dependencies.

We develop a library, and use hexy as well as a library for its work with coordinates, hex cells.
Not being the end-user, we don't need any of the graphical stuff.

I figured there's an option in setup.py to split dependencies to "required" and several named "extras"
https://stackoverflow.com/questions/41678951/ignoring-some-requirements-when-installing-pip-requirements

• core:
  for hexy.py, we found only numpy suffices (maybe some more for the tests?)

• visualizations:
  pull the large game and maybe some more.

Would you consider splitting your setup to the

• base library pip install hexy
• and providing extras? for example:
  • tests: pip install hexy[tests]
  • visualizations: pip install hexy[visual]
  • and altogether: pip install hexy[full]
    ?

Thank you,
Mark

Hi,
first of all, this is a very nice software, thank you! 👍
We are going to use your lib in our oss library too,
it would be great if you could package your lib in PyPI so it can be easily installed with
pip install hexy

Cheers,

it is a known and documented problem that the dict object possess in-line optimizations that will make any subclass behave in a weird way. For example, calling HexMap.pop(k) will not go through your customized __delitem__, and would result in buggy behavior down the line. If the pop or update methods aren't welcomed in your API, consider wrapping an actual dict in your HexMap object instead.

Example:

class HexMap:
    def __init__(self):
        self._map = {}
    def __setitem__(self, k, v):
        self._map[k] = v
    def __delitem__(self, k):
        del self._map[k]

It it, from my understanding, recommended to use the abstract objects from collections.abc to obtain a dict-like behavior

More over, in the linked code bellow, it is seemingly impossible to both initialize and populate a HexMap at the same time.

It looks like there have been new features added in the last few years, but the latest version on PyPI is from 2019.

Thanks for the nice library!

Just a heads up, with the latest release, you've jumped from 1.3, 1.4, to 1.41.
Likely you've meant 1.4.1.
This will cause issues with updating, as 1.5 < 1.41, can you fix/remove the wrong release number from pip?