Errors

Native failures become JavaScript exceptions. A false ground query is an ordinary negative answer; malformed input or a failed solve instead throws.

Diagnostics

Native failures throw DatalogError with the returned status and an immutable diagnostic snapshot. The snapshot includes statusName, precise code and codeName, source, the bigint presence mask, and message/phase/hint strings. Optional sections include location, predicate, capacity, depth, comparison, arity, rule and context. Sections may coexist; absent fields are not meaningful zeros. Errors from operations without a diagnostic do not reuse old messages.

try {
  pg.solve()
} catch (error) {
  if (error instanceof DatalogError) {
    const d = error.diagnostic
    console.error(d.codeName, d.predicate?.name,
      d.capacity?.observed, d.capacity?.limit)
  }
  throw error
}

A SMALL predicate overflow can report 65 / 64 and the predicate together. Read these values from the diagnostic: LARGE has a different bound.

The published declaration also gives Diagnostic optional aggregate and context sections. Test a section's presence before reading its fields; the present mask says which native sections were actually supplied.

Build limits

The published Limits type names the values in this snapshot.

buildLimits() returns the compiled limits: symbols, string pool bytes, predicates, rules, arity, body literals, depth, EDB facts, IDB facts, facts per predicate, maximum string bytes and input text bytes. SMALL and LARGE remain fixed-capacity engine profiles. Emscripten memory growth does not create an elastic engine profile or promise a browser memory ceiling.

const limits = pg.buildLimits()
console.log(limits.maxEdbFacts, limits.inputEdbTextBytes)

Read the active module's limits rather than copying a capacity from an example or from another profile. A failed solve publishes no partial result.