disposeGracefully method

Future<void> disposeGracefully({ Duration timeout = const Duration(seconds: 5), })

Gracefully disposes the worker: sends workerDisposeOp to the isolate as a request and awaits its acknowledgement — giving the isolate a chance to free native resources — before force-killing it via dispose.

dispose alone kills the isolate synchronously with Isolate.kill(priority: immediate), which races past any queued dispose message, so the isolate can die before releasing its native TFLite interpreters (~10-26MB leaked per detector on Android; under sequential create/dispose load the low-memory killer reaps the process). Routing teardown through this method first lets the isolate ack and clean up.

Best-effort: if the worker is not ready, workerDisposeOp is null, or the ack does not arrive within timeout, it falls through to dispose. The isolate's dispose handler must reply {'id': id, 'result': ...} for the ack to resolve.

Implementation

Future<void> disposeGracefully({
  Duration timeout = const Duration(seconds: 5),
}) async {
  final String? op = workerDisposeOp;
  if (_initialized && op != null) {
    try {
      await sendRequestUnchecked<dynamic>(
        op,
        const <String, dynamic>{},
      ).timeout(timeout);
    } catch (_) {
      // Fall through to the synchronous force-kill below.
    }
  }
  await dispose();
}