This project uses jupyter-book to build a book with all the notebooks of the packages in the PySAL federation. The result is available at:

http://pysal.org/notebooks

If you want to access the built book, simply head over to the URL above. If you want to build the book locally, you will need the following:

• A recent version of jupyter-book
• Unix OS: the builder relies on a few shell commands (rm, mkdir, cp, mv, cd)
• pandoc (for .rst to .md conversion)
• git
• Python libraries listed in build-requirements.txt

A recommended route is to run this repository inside a gds_env (use the gds_dev flavour) container, which is the platform used to build it. If you have it installed, you can launch it from the notebooks folder by running:

> docker run --rm -ti -p 8888:8888 -p 4000:4000 -v ${PWD}:/home/jovyan/work darribas/gds_dev:4.0

The current build process involves two main steps:

1. Pull repositories to extract the files used to build the book. This can be done by running the following command from the root folder:

   > python lib/build.py --pull

   This will generate a notebooks folder that contains all the files required to build the book.

2. Build the book from the downloaded notebooks. This can be done by running the following command from the root folder:

   > python lib/build.py --build

   This will create/update the docs folder in the root folder so it contains everything needed for the hosting server to serve the book. Commit the changes and the updated book will be available online shortly.

The built website is available on the ./docs folder and is ready to be pushed to the repository for serving on Github Pages. If you want to serve it locally, you will need to:

1. Navigate into the ./docs directory:

   > cd docs

2. Serve the site:

   > jekyll serve

   If you are serving it inside a container, fix the host so it can be accessed from the host machine:

   > jekyll serve --host="0.0.0.0"

   Then point your browser to localhost:4000/notebooks/

Additionally, the command line tool lib/build.py incorporates experimental support for testing the notebooks. Again, two options are available:

1. Test all the .ipynb files are well-formated and can be converted:

   > python lib/build.py --test_no_run

2. Execute all the .ipynb files:

   > python lib/build.py --test_run