Serialize and Deserialize Python Objects From Different Modules While Never Importing Both at the Same Time
I had a Kubeflow pipline where module size was limited. One can either revert to type unsafe data conversion or depend on a third module for data schemas.
We define functionality and data in separate Python modules. While functionality gets split into different modules, data definitions are placed in a common module. The size of code that needs to be sent to pipeline workers. If data_models
becomes to big, e.g. if there are more classes that are can be converted into each other, it can be split up into foo_data_model
and bar_data_model
.