Runtime concurrency and state

MontyRuntime runs Python on sessions from a Monty pool. Choose shared mode when later code must read state created by earlier code. Choose fresh-session mode when calls require isolated globals or independent scheduling.

Shared mode requires caller serialization

The runtime retains one session in shared mode. The application must serialize requests when several callers share that runtime. A small queue can preserve request order without moving that policy into the runtime:

import 'dart:async';

import 'package:dart_monty/dart_monty.dart';

class RuntimeQueue {
  RuntimeQueue(this.runtime);

  final MontyRuntime runtime;
  Future<void> _tail = Future<void>.value();

  Future<MontyValue> run(String code) {
    final previous = _tail;
    final done = Completer<void>();
    _tail = done.future;
    return () async {
      try {
        await previous.catchError((Object _) {});
        return await runtime.execute(code).result;
      } finally {
        done.complete();
      }
    }();
  }
}

Fresh mode trades state for concurrency

Fresh mode isolates Python globals while pool capacity permits independent calls. Pool capacity must also cover nested work started by host functions or sandbox children.

MontyRuntime and ExecutionHandle define lifecycle contracts. See Extension system for attachment and child-execution ownership.