Ability to validate a jmespath expression against a JSON schema

[sebastien-rosset]

Problem Statement
It does not seem possible to validate a JMES path expression against a JSON schema.

A user writes a JMES path expression and wants to determine if the expression is syntactically and semantically valid. The JMES parser can find syntax errors but not semantic errors. For example the user types the expression foo[*].first. This is syntactically a valid expression, but this may not be valid against a JSON schema, e.g., the JSON schema may not allow the foo property.

If a JSON schema is available, it should be possible to determine whether a JMES expression refers to non-existent properties in the JSON schema. In the foo[*].first example, the validation would have to check if there is at least one match in the JSON schema.

A related problem is to identify the set of schema elements that are referred by a jmespath expression.

I understand this is not trivial. For example, to evaluate a JMESpath expression such as foo[100] against a JSON schema, multiple scenarios need to be considered with a BFS traversal of the parsed expression.

1. check if ‘foo’ is defined in the JSON schema.
   1. additionalProperties: false and all properties are defined explicitly. ==> It’s possible to continue BFS.
   2. additionalProperties: true ==> stop DFS, the property may or may not be present in the payload
   3. for oneOf, allOf, anyOf, cases, need to iterate through each sub JSON schema.
   4. Pattern properties in JSON schema. If the JMESpath identifier does not include regex, then evaluation is possible. If the JMESpath identifier includes a regex, then it may be very difficult to determine a match.
   5. “required” attribute in JSON schema could be used to provide additional semantic information about whether a match will definitely occur or may occur.
2. If the type of ‘foo’ is defined in the JSON schema and it’s not ‘array’, then the jmespath expression is not semantically valid.
3. if the type of ‘foo’ is not specified in the JSON schema, it is unknown whether a match is possible or not.
4. if the type of ‘foo’ is set to ‘array’ in the JSON schema, then it’s possible to do additional validation based on min/max constraints in the JSON schema.