Anaconda环境中导入gdal模块报错问题求助

Hey, I’ve dealt with this exact GDAL import issue in Anaconda before—let’s get you sorted out! The error you’re seeing happens because of mismatched dependencies or conflicts between pip-installed GDAL and Anaconda’s environment setup. Here’s what worked for me:

Step 1: Clean up existing GDAL installations #

First, we need to remove any broken or conflicting GDAL packages from your environment:

• Uninstall pip-installed versions:

  pip uninstall -y gdal osgeo
  

• Force-remove any conda-installed GDAL to start fresh:

  conda remove --force gdal
  

Step 2: Install GDAL via Conda (the reliable way) #

Conda handles GDAL’s tricky system dependencies way better than pip. Follow these steps:

1. Add the conda-forge channel (it has up-to-date, pre-compiled GDAL packages):

   conda config --add channels conda-forge
   conda config --set channel_priority strict
   

2. Install a GDAL version compatible with your Python 3.6 environment (GDAL 2.4.x is a safe bet for Python 3.6):

   conda install -c conda-forge gdal=2.4.4
   

   Or let conda pick the best compatible version automatically:

   conda install -c conda-forge gdal
   

Step 3: Verify the installation #

Activate your environment and test the import:

conda activate python3
python

Then in the Python shell, use the recommended import method (the old import gdal is a legacy wrapper that can cause issues):

from osgeo import gdal
# Test if it works
print(gdal.__version__)

Step 4: If issues persist, check your PATH #

Make sure your Anaconda environment’s bin directory is at the start of your PATH to avoid conflicts with system-installed GDAL. On macOS, run this after activating your environment to confirm:

echo $PATH

You should see /Users/username/anaconda3/envs/python3/bin listed first.

The core issue here is that pip-installed GDAL often fails to properly link against Anaconda’s libraries, leading to the swig import error. Conda’s pre-built packages are tailored to work seamlessly with Anaconda environments, so this should resolve the problem.

内容的提问来源于stack exchange，提问作者Stefano Potter