Closed liamtoney closed 1 year ago
I also noticed it yesterday. It's weird that the commands in the notebook install "python 3.9" but the python version in the error message says it's python3.6.
I also noticed it yesterday. It's weird that the commands in the notebook install "python 3.9" but the python version in the error message says it's python3.6.
After doing some reading (e.g. https://towardsdatascience.com/conda-google-colab-75f7c867a522), it seems that we need to pin conda's Python version to be the same as that on Google Collaboratory (currently Python 3.7) instead of the Miniconda default (Python 3.9). However, I tried getting a Python 3.7 installer from https://repo.anaconda.com/miniconda/, and it still produces a DistributionNotFound error (as Liam and a forum user at https://forum.generic-mapping-tools.org/t/installing-pygmt-on-google-colab/898/12 reported). So needs a bit more work somehow.
I managed to use pygmt in the colab "installing GMT6 fresh". Take a look at https://github.com/andrebelem/pythonverse/blob/main/PyGMT_on_COLAB.ipynb and feel free to incorporate it in the try-gmt.
Should work with these commands
Two years ago, I created an example of how to use PyGMT in COLAB. I've updated the example now in the latest version of COLAB (March 2024), which you can see here: https://github.com/andrebelem/pythonverse/blob/main/Using_pygmt_in_COLAB_[version_2023].ipynb.
Essentially, you'll install Conda, and then GMT/PyGMT. Note that this should be done in parts, and the first two cells take about 3 minutes to install everything. Good luck.
PyGMT can't recognize to the libgmt.so under Google Colab. So, it is important to refer the path of libgmt.so in linux system which google are using
!ln -s /lib/x86_64-linux-gnu/libgmt.so.6 /lib/x86_64-linux-gnu/libgmt.so
What I did just adding alias of libgmt.so.6 and I reinstalled the GMT and ghost script.
It works well. I didn't try to use conda but good to know 2 ways of solving the problem
Thanks
Installing GMT using !sudo apt-get install gmt gmt-dcw gmt-gshhg
takes much longer than using conda/mamba, which causes most users to give up on the process. I myself abandoned the idea of apt install
some time ago. I recommend using conda install.
After running the cell which installs conda etc. the
import pygmt
command fails with: