RuntimeWarning when using pyshp 2.1.1

[hashstat]

We've been using pyshp 2.1.0 successfully in our large project for months now. Today, with the introduction of version 2.1.1, I am now seeing this runtime warning when I run our application:

/home/user/.pyenv/versions/3.7.7/lib/python3.7/importlib/_bootstrap.py:219: RuntimeWarning: numpy.ufunc size changed, may indicate binary incompatibility. Expected 192 from C header, got 216 from PyObject
  return f(*args, **kwds)

It threw me off because pyshp doesn't depend on numpy at all, but downgrading pyshp eliminates the warning. Have you ever seen this before? Any thoughts on how pyshp could be involved in this issue?

I'm able to reproduce this on both Linux and Windows. I haven't tried on Macos, but I'm guessing it will be an issue there as well.
I'm using 64-bit versions of Python 3.7.7. My applications depends on Pandas, numpy, scipy, and a host of other packages. I'll dig into the issue more, but thought you could provide some insight, or at least a starting point.

Thanks!

[karimbahgat]

Yes, that is indeed weird. But come to think of it I noticed some similar numpy errors when I ran some unrelated code that also used pyshp.

The only thing I can think of that may cause this is that as part of the recent changes I added warnings.simplefilter('always') upon import, because I wanted it to raise a warning each time pyshp encountered invalid polygons, not just the first time which I think is the warning module's default. I suspect that maybe that changes the warning behavior in all your other codebase as well, thus making visible some scipy/pandas/numpy warnings that were already there but that perhaps were suppressed in the past.

Removing the warning simplefilter line fixed it for me, so this must be the culprit. Pushing a new v2.1.2 to fix the issue.

[karimbahgat]

Could you confirm that the newest changes I just pushed to GitHub fixes the warnings issue in your case?

[hashstat]

Just confirmed that your fix quiets the warnings. Built and installed v2.1.2 from master and everything is working as usual.

Thanks for your quick response to this issue!

[karimbahgat]

No worries, trying to improve my response time, and that warnings issue would likely be a serious nuisance to many. Thanks a lot for pointing it out so fast!

New v2.1.2 out on PyPI now.