Context Validation

Flow

• New or modified context nodes are validated
  • Invalid context trees are not saved
  • The node's natural language representation are computed and saved for future reference
• Existing nodes are re-validated upon use
  • Invalid nodes are automatically flagged ignore=true to prevent returning no or bogus results
  • Validation errors will be attached onto the node
• Ignored nodes can be forced to be used by setting ignore=true
  • Until the node is changed or removed, this flag will remain so the node does not get redundantly validated

{
    "concept": 4,
    "field": 39,
    "value": 54,
    "operator": "gt",
    "nulls": false,
    "nlang": "PTA is greater than 54",
    "ignore": true,
    "errors": ["Value out of range"]
}

Errors

• VALUE_OUT_OF_RANGE - Value out of range
  • Applies to numerical values that are greater than the max or less than the min
• VALUE_NOT_A_CHOICE - Value not a choice
  • Applies to enumerable or key-based data where the value is not in the set
• INVALID_OPERATOR - Invalid operator
  • The operator supplied is not valid for the field
• INVALID_OPERATOR_OR_VALUE - Invalid operator or value
  • The combination of operator and value are invalid, e.g. between query with a single value

Implementation

• settings

  • AUTO_REMOVE_CONDITIONS
    • auto-flag conditions that were once there or valid, but are no longer

• serialization

  • stringify and humanize conditions in plain text
    • ensures original representation and meaning does not change for warning messages
  • warning messages at the concept-level (if one exists)

• changes

  • hard validation occurs for new conditions
  • soft validation occurs during restore and use
    • this involves flagging conditions that are no longer valid
  • ignore flagged conditions rather

Client Considerations

• show warning messages per concept/field
• provide remediation options
  • confirm removal
    • this will remove the condition permanently from the tree
  • revoke removal
    • certain use cases may deal with volatile data and the data may appear eventually