| [ease-dev] Jupyter kernel |
Hi,
I was working on the execution requests for the Jupyter kernels this week. I hope that I get them running tomorrow and will keep you posted here.
Once the execution requests are handled, I will move on to sending data back from the kernel to the clients. This will probably be the most interesting part for developers as this will allow you to post data to be displayed.
The general principle for posting data is to send a dictionary from mime type to actual (byte) data. Clients can then decide what they want to display. It is possible to send multiple mappings at once. A console could then display the text representation while another client would choose the jpg format (as an example).
There are a couple of possibilities how to handle this from the developer side. A simple solution would be to have a method postData(Map<String, Byte[]> data) that already receives the correctly formatted data. Besides that we could add some convenience methods for posting strings and numeric types. This is very generic and would cover all cases. It lacks a bit of usability as you would need to convert the data to its byte representation. I would also like to add a new interface IPostable that helps converting objects to their dictionary representation. This would add an additional abstraction layer for module developers to simplify publication for script developers. Do you find this to be too complicated? If so, do you see any possibilities to simplify things?
Best regards,
Martin