Model input parameter validation features and standards

[tariqshihadah]

Model parameter input validation will be a critical component of the framework and the development of models using the framework. Past iterations have used homebrewed, purpose-built validation classes which worked OK for basic cases but were not robust and very limited in scope. Instead, I'm initially recommending use of the open-source cerberus package which is vast, well-documented, and easily extensible, making it a strong fit for this project I think. Additionally, validation parameters can easily be serialized to JSON or .py, making it simpler to store and manage validation information outside of the main model files and is consistent with the goals of the framework to be modular and accessible for necessary editing by users with minimal python skills.

Check out the validation module to see how this is currently being implemented. At this time, it is minimal and can only be used to perform success/fail validation of all parameter inputs using the existing cerberus API and is not plugged into all branches. A few goals/hopes/dreams for this issue:

• Validation failure fallback. It would be great to add a variety of options to respond to validation failures, such as snapping numerical values to the required range, providing a default value, or throwing/documenting a non-fatal warning
• Automated parameter help documentation. Based on the different types of validation, it would be great to be able to automatically present basic text documentation for set validation specification, exposing rules to users through a help() method on model Container subclasses or similar where validation is set.