New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Failing voronoi example with the new 0.2.2 release #69
Comments
The example is
from https://geopandas.readthedocs.io/en/latest/gallery/plotting_with_geoplot.html |
I assume the build is using This looks like an unrelated failure...that codebase hasn't changed in the intervening time, so perhaps there was an update in |
Yeah, didn't yet look into detail. It's just that the readthedocs build passed few days ago, and then failed with the above error the day before yesterday, and after checking that there was a new release of geoplot in that timeframe, I assumed that might have been the cause. |
@jorisvandenbossche What version of |
The log output of readthedocs is a bit annoying, in the sense you cannot see the exact environment that is created (truncated log), so I am not fully sure. |
I tried to recreate a similar local environment using the same environment file as we are using on readthedocs (which installed numpy 1.15.4), but I cannot also not reproduce it .. I triggered a re-build on readthedocs, let's see if the issue is persistent. |
OK, now the build succeeded .. Sorry, not sure what was going wrong before, but so ignore this! |
It's theoretically possible there is a non-deterministic error in the algorithm (I remember debugging this exact error message in certain cases while I was working out the procedure). If someone else reports this or if it comes up again I'll have to take a closer look. |
OK, figured out the reason why I only sometimes saw the failure and couldn't consistently reproduce it: it is because the example in the geopandas gallery is doing a And with the following random state, it can be reproduced on my laptop:
resulting in
|
And depending on the random state, you can also have a different error:
|
…rors (#933) See ResidentMario/geoplot#69: depending on how the data are subsampled, we stumble into a potential bug in geoplot. So fixing the random state with a working one to prevent this.
…rors (#933) See ResidentMario/geoplot#69: depending on how the data are subsampled, we stumble into a potential bug in geoplot. So fixing the random state with a working one to prevent this. (cherry picked from commit 6dd50b8)
Will look at this on the weekend. The short term fix is to just set a seed for the docs build that is known to work...which, really, I should have always been doing. I see you've already done that. Sorry about that. The long term fix is to figure out what the conditions for this failure are. That's going to be more involved because it has to do with the way the voronoi polygon construction algorithm works in certain edge cases. I was aware that the algorithm had some issues, but I thought it was constrained to edge cases involving very small samples...clearly if it's failing this much on 1000-point samples that's simply not true. 😦 |
Fixed in |
The geoplot release seems to have broken the geopandas examples (the voronoi one). I am getting the following error on our readthedocs build:
The text was updated successfully, but these errors were encountered: