import rtree leads to `OSError: Could not load libspatialindex_c library` on aarch64 about rtree HOT 6 CLOSED

I have an m1 mac. if you can add a Dockerfile and its build command that demonstrates the issue, I would be much more likely to attempt to address it.

These kinds of errors have been so hard to address. if you're using conda, you should use the conda-forge Rtree package instead of pip-installing the wheel. Maybe things are getting confused there.

from rtree.

A very reasonable request!
Here's a minimum working example of the issue. I suspect it will run fine if run on x86 architecture, but I have no way to confirm:

Dockerfile contents (example taken from @here for anyone who is new to this - the filename is Dockerfile.)

FROM python:3.9-slim-buster

RUN pip install rtree

CMD ["sh", "-c", "python -c 'import rtree'"]

Run the following commands from the directory containing the dockerfile

docker build -t rtree_error .
docker run rtree

The error:

Traceback (most recent call last):
  File "<string>", line 1, in <module>
  File "/usr/local/lib/python3.9/site-packages/rtree/__init__.py", line 9, in <module>
    from .index import Index, Rtree  # noqa
  File "/usr/local/lib/python3.9/site-packages/rtree/index.py", line 17, in <module>
    from . import core
  File "/usr/local/lib/python3.9/site-packages/rtree/core.py", line 74, in <module>
    rt = finder.load()
  File "/usr/local/lib/python3.9/site-packages/rtree/finder.py", line 118, in load
    raise OSError("Could not load libspatialindex_c library")
OSError: Could not load libspatialindex_c library

from rtree.

These kinds of errors have been so hard to address. if you're using conda, you should use the conda-forge Rtree package instead of pip-installing the wheel. Maybe things are getting confused there.

I didn't fully address this - I would actually much prefer the pip-solution to work if possible. The only reason I created a conda environment is that I find the environment-functionality very convenient there (and I feel that venv / asdf aren't quite as user-friendly). The reason is that I'm using geopandas to build a geographical calendar at work, and at work everyone is using linux-in-docker via pip for reproducibility.

But I totally understand if this level of debugging is outside your scope! I can probably find a different way.

from rtree.

pip install Rtree in this case only installed the source for the Python library, not the libspatialindex library that's also included with the wheels. That's because none of the wheels on https://pypi.org/project/Rtree/#files match the sys_tags.

In short, the wheel didn't get installed.

python  -c "import packaging.tags; print(list(packaging.tags.sys_tags()))"

In this case, just do apt-get install libspatialindex-c5 -y to pull that in and things should work.

from rtree.

@hobu The point of wheels is that all necessary libraries are included, right? At the very least, if the library is not found, the installation should fail, not only at runtime.

from rtree.

@simaoafonso-pwt Please stop spamming the repository. If you need 32bit wheels you will have to build them yourself.

from rtree.