Struct ValidationExecutor 

pub struct ValidationExecutor { /* private fields */ }

Compile-execute-validate pipeline for TensorLogic expressions.

Example

use tensorlogic_ir::{TLExpr, Term};
use tensorlogic_oxirs_bridge::{ValidationExecutor, ValidationExecutorConfig};

let expr = TLExpr::pred("knows", vec![Term::var("x"), Term::var("y")]);
let executor = ValidationExecutor::new(ValidationExecutorConfig::default());
let result = executor.execute_rule(&expr).unwrap();
let report = executor.generate_validation_report(&result);
assert!(report.conforms);
let rdf = executor.export_as_rdf(&result);
assert!(rdf.contains("@prefix tl:"));

Implementations

impl ValidationExecutor

pub fn new(config: ValidationExecutorConfig) -> Self

Create a new executor with the given configuration.

pub fn config(&self) -> &ValidationExecutorConfig

Return a reference to the active configuration.

pub fn execute_rule( &self, expr: &TLExpr, ) -> Result<ExecutionResult, ValidationExecutorError>

Compile expr to an tensorlogic_ir::EinsumGraph, execute it with placeholder input tensors via Scirs2Exec, and return an ExecutionResult.

Errors

• ValidationExecutorError::Compile — compiler rejected the expression.
• ValidationExecutorError::EmptyGraph — compiled graph is trivially empty.
• ValidationExecutorError::Execute — forward pass failed.
• ValidationExecutorError::TensorTooLarge — output exceeds configured limit.

pub fn generate_validation_report( &self, result: &ExecutionResult, ) -> ValidationReport

Generate a SHACL-style ValidationReport from an ExecutionResult.

The report conforms (report.conforms == true) when every output tensor contains only finite values. A sh:Violation-severity ValidationResult is added for each tensor that contains NaN or Inf elements, carrying the tensor name and non-finite count in the message.

pub fn export_as_rdf(&self, result: &ExecutionResult) -> String

Serialise an ExecutionResult as an RDF Turtle string.

The generated document uses a tl: prefix for the configured base_iri and declares standard XSD types. Each execution is identified by a stable IRI derived from the expression_repr hash so that repeated calls for the same expression produce consistent IRIs.

Each output tensor is appended as a blank-node tl:OutputTensor with shape, element count, allFinite, min, and max.