Why not use python's schema library (or similar) for validation?

Python’s ‘schema’ (as well as similar libraries) can also be used to validate the structure of objects. Validating YAML is even cited as a reason on their README.

Using a schema for validation requires running the YAML through a parser first which and then taking the output (usually a data structure like a dict) and passing it through the schema.

Unfortunately there are a number of problems with this approach:

You still have the Norway Problem

If the standard YAML parser parses ‘NO’ as false or empty string as None then it doesn’t really matter if the schema says an empty string or the text ‘NO’ is okay, it will be seeing a ‘None’ or a ‘False’ which will cause a failure.

You can’t get line numbers and snippets for the validation errors

Assuming you’ve successfully circumvented the Norway problem, parsing and feeding the output to schema is still problematic. If you pass a parsed dict to schema, schema can’t tell which line number the failure happened on and can’t give you a code snippet highlighting where it happened.

Roundtripping becomes very very difficult if not impossible

Due to the loss of metadata about parsed YAML being lost when it is fed into a generic schema validator, it also becomes impossible to to change the data and serialize it without losing critical details (i.e. mapping ordering, comments or validation structures).