I’m using Literate.jl to convert example scripts to markdown (md) and Jupyter Notebook (ipynb) for the documentation. The scripts are in examples, and I have a script in docs/generate.jl that gets run as part of docs/make.jl to convert the examples/*.jl files into md and ipynb files in the docs/examples/ folder. I’ve copied this workflow from Ferrite.jl. When I execute docs/make.jl, I see that (as expected), the output folder docs/build/examples contains the html file rendered from the md, as well as the ipynb file for each example.
That is, docs/build contains
.
├── [...]
├── assets
│ └── [...]
├── examples
│ ├── index.html
│ ├── simple_state_to_state.html
│ ├── simple_state_to_state.ipynb
│ ├── state_to_state_rwa.html
│ └──state_to_state_rwa.ipynb
├── index.html
└── [...]
The output from my Github Pipeline seems to indicate that these files also get properly generated in the CI build folder.
However, the .ipynb files do not get copied over to the gh-pages branch during the CI deployment. The gh-pages branch contains the following files after deployment:
.
├── dev
│ ├── api
│ │ └── index.html
│ ├── assets
│ │ └── [...]
│ ├── examples
│ │ ├── index.html
│ │ ├── simple_state_to_state
│ │ │ └── index.html
│ │ └── state_to_state_rwa
│ │ └── index.html
│ ├── index.html
│ └── [...]
└── versions.js
I’ve also tried this with the prettyurl option for Documenter.jl set to false, which doesn’t help (it just replaces, e.g., simple_state_to_state/index.html with simple_state_to_state.html)
I haven’t found any documentation for what exactly Documenter.jl does during the deployment to gh-pages. Clearly, it does not simply copy over the entire docs/build folder, as I would have expected. Also, the deployment of the .ipynb files seems to work for Ferrite.jl as well as for Literate.jl itself. I don’t see that I’m doing anything different from those two packages. Does anybody have any experience or ideas on how to get Documenter.jl to deploy .ipynb files from docs/build to gh-pages?