I've been wanting a way to run the Swift kernel locally, but using my own Jupyter instance not in a Docker container. So I created a descendent Dockerfile that adds Jupyter Kernel Gateway to the swift-jupyter image and runs it on container startup. Details are in the readme.
This seems like a big improvement in terms of local S4TF workflow, though there are definitely other ways this could be handled. (Two that come to mind: a separate project for the kernel gateway Dockerfile; or including kernelgateway in the regular image but allowing for runtime choice of notebook or kernelgateway.)
I also did some tidy up of the Dockerfile (to reduce cache busting and reduce number of image layers), added some files to .dockerignore (to reduce image size, albeit inconsequentially), and made a recommendation to use Docker Buildkit (to improve the building process). Those all seem good and relevant to me, but aren't necessary to creating the kernelgateway image.
marcrasi
commented
4 years ago
I've been wanting a way to run the Swift kernel locally, but using my own Jupyter instance not in a Docker container. So I created a descendent Dockerfile that adds Jupyter Kernel Gateway to the swift-jupyter image and runs it on container startup. Details are in the readme.
This seems like a big improvement in terms of local S4TF workflow, though there are definitely other ways this could be handled. (Two that come to mind: a separate project for the kernel gateway Dockerfile; or including kernelgateway in the regular image but allowing for runtime choice of notebook or kernelgateway.)
I also did some tidy up of the Dockerfile (to reduce cache busting and reduce number of image layers), added some files to .dockerignore (to reduce image size, albeit inconsequentially), and made a recommendation to use Docker Buildkit (to improve the building process). Those all seem good and relevant to me, but aren't necessary to creating the kernelgateway image.
Let me know how this looks. Thanks!