My latest bad idea thats actually good: what if instead of a schema generating a python module, what if the schema was a python module by hooking into the import machinery
Then theres no worrying about either getting out of sync. Detect changes and regenerate on import. If you want to directly modify the python module, no problem, its fully invertible back to the schema in a way thats totally orthogonal to any modifications youd make
So then you dont just have code thats generated from a linked data schema, you have code that IS a linked data schema. And with our recent work with arrays you get a linked data schema that you can actually use for normal data things. And with a firmer bridge, it becomes easier to virtualize eg. Tables of values as pandas dataframes with exactly the same object youd represent an individual value.
Add comment