Here are some tips for a smooth installation experience for installing the DisCoPy Python package so that you can use it in Jupyter notebooks through Anaconda.
Links:
- GitHub README page: https://github.com/discopy
- GitHub repository: https://github.com/discopy/discopy
On the pages above it says to just run pip install discopy for a quick installation. However, when I did this I got a few error messages when running the following code (from this documentation page):
from discopy.grammar.pregroup import Ty, Id, Word, Cup, Diagram
n, s = Ty('n'), Ty('s')
Alice = Word("Alice", n)
loves = Word("loves", n.r @ s @ n.l)
Bob = Word("Bob", n)
grammar = Cup(n, n.r) @ s @ Cup(n.l, n)
sentence = Alice @ loves @ Bob >> grammar
sentence.draw(figsize=(5, 5))
The error message I got was: TypeError: Expected discopy.monoidal.Diagram, got Ty('s') of type discopy.rigid.Ty instead.
When I ran the pip freeze | grep discopy command to check which version of DisCoPy the pip install had installed, I discovered that it had installed version 0.6.0 and not the latest version (1.1.4 on the day of writing this).
I then tried to download and install the released version 1.1.4 through https://github.com/discopy/discopy/releases/tag/1.1.4, but I didn't manage to get it to work.
In the end, I found this GitHub issue and decided to install DisCoPy as suggested in the issue:
git clone https://github.com/oxford-quantum-group/discopy.git
cd discopy
pip install .
However, to be able to use DisCoPy in Jupyter notebooks through Anaconda, I had to do the following:
- Create a new Conda environment:
conda create --name discopy_env - Activate the environment:
conda activate discopy_env - Install a Python kernel in the Conda environment:
conda install ipykernel
https://github.com/hidaris/thingtalk/issues/9
https://stackoverflow.com/questions/58568175/upgrade-to-python-3-8-using-conda
https://stackoverflow.com/questions/42231764/how-can-i-rename-a-conda-environment