Import new Base images from media library

[josecelano]

Right now the current workflow only uses the images that are already inside the data folder.

The data folder contains tif images that were manually copied to this repo using [dvc](https://dvc.org/)

data/
├── 000001
│   └── 32
│       ├── 000001-32.600.2.tif
│       └── 000001-32.600.2.tif.dvc
└── 000002
    └── 32
        ├── 000002-32.600.2.tif
        └── 000002-32.600.2.tif.dvc

We got the images from this media library: https://github.com/Nautilus-Cyberneering/chinese-ideographs

You can deploy the application using the script: npm run deploy. That command executes the script ./scripts/copy-images.sh which copies the images from the data folder to the public\images folder.

public/images/
├── 000001-32.600.2.tif
└── 000002-32.600.2.tif

Those are the images that are shown on the website: https://nautilus-cyberneering.github.io/chinese-ideographs-website/

We want to update the website when new images are added to the source library. For this first iteration we are going to use this solution:

1. We create two different workflows: deploy.yml and import-base-images.yml
2. The new workflow (import-base-images.yml) is going to git clone the library and auto-commit new Base images in public/images/ folder.
3. The current workflow (deploy.yml) will be modified to only build and deploy the website using gh-pages.

The deploy workflow will:

1. checkout the website repo
2. build app
3. deploy to gh-pages

And the import-base-images.yml workflow will:

1. checkout the library repo in a subfolder (gitignored). It will be the current data folder, but maybe something like: library\data\...
2. dvc pull all images in the sub git project (we are not going to use git submodules or git subtrees for the time being, although maybe it worths considering it).
3. copy images from the library to the repo folder (public\images)
4. auto-commit new images
5. Trigger deploy.yml workflow to publish the new version of the website on gh-pages. If this step is complicated for the POC we can simply repeat the steps needed to publish the website on this workflow.

The reasons to have two workflows are:

• We want to have a simpler workflow if we only change the website source code (JS, CSS) but not the images.
• In the future, we want to improve the import workflow. We want to trigger it from the library repo and we want to pass it an event payload, which does not make sense for the standard deploy.

Future improvements:

• Run import-base-images.yml workflow on-demand only when Base images change in the library repo. We could trigger a workflow on the website repo from the library repo using GitHub API: https://github.com/marketplace/actions/repository-dispatch
• Avoid pulling all images with dvc. In the previous trigger, we can set a payload for the event with the list of changed Base images. So we only need to process those files.

[josecelano]

I'm using git submodules and it seems to work fine in this case but I do not have a list of prons and cons yet versus using subtrees.