Pub Version Dart CI

A Dart package for easier and more comprehensible flow of error handling and logging.

Motivation

I made this package because I found it hard to handle exceptions:

  • An app stops on an exception if it is not caught.
  • It is sometimes unclear if an exception has already been caught somewhere.
  • It is not preferable to catch an exception in a different layer that should be agnostic about things specific to the original layer (e.g. a DB error / a network error).
  • However, it is difficult to return an error value (instead of the exception itself to avoid the above issue) together with the result of some processing from a method to its caller located in another layer.

Solutions:

  • Result in package:async
  • package:errflow (this package)

These look very different, but roughly speaking, they are similar in that they provide a way to pass a result and/or an error from a method to the caller; the former holds either of the values, and the latter allows the caller to evaluate both values.

A big difference is that this package also provides handlers and a logger to enable errors to be handled more easily in a unified manner. It makes it possible that error handling is centralized.

Usage

Initialisation and clean-up

Instantiate ErrFlow, with the default value representing that there is no error. The value is used as the initial value in the notifier of each scope().

When the ErrFlow object is no longer needed, call dispose() to ensure that the resources held in the object will not be used any more.

enum CustomError {
  none,
  foo,
  bar,
}

...

final errFlow = ErrFlow<CustomError>(CustomError.none);

...

errFlow.dispose();

If you prefer using Exception and its subtypes instead of a custom error type, specify Exception as the error type, and pass null or none to the constructor to use null as the default value.

final errFlow = ErrFlow<Exception>();

Setting/logging an error

The notifier passed to the callback function of scope() has the set() method used for updating the error value and calling the logger, and the log() method for only triggering the logger.

  1. Use scope() to pass an ErrNotifier to a function that can cause an error.
  2. Call set() on the notifier when some error happens in the function.
  3. The listener is notified of the error and stores it as the last error (lastError) so that it can be checked later inside the function.
  4. The listener also calls the logger to log a set of information about the error if it is provided via set() or log().
  5. One of the error handlers is called after the callback completes if a condition to call it is met.

Error handlers are explained in detail later in this document.

final result = await errFlow.scope<bool>(
  (notifier) => yourFunc(notifier),
  ...,
);
Future<bool> yourFunc(ErrNotifier notifier) async {
  try {
    await errorProneProcess();
  } catch (e, s) {
    // This updates the last error value and also triggers the logger.
    notifier.set(CustomError.foo, e, s, 'additional info');
  }

  // If necessary, you can use hasError to check if some error was set.
  if (notifier.hasError) {
    ...
    return false;
  }

  return true;
}

You can also use only the first argument of set() to not trigger the logger:

notifier.set(CustomError.foo);

or use log() for only logging:

notifier.log(e, s, 'additional info');

Isn't it inconvenient to have to pass a notifier?

It is not impossible to remove the hassle to have to pass over an object of ErrNotifier, but I choose not to do so because method signatures with a parameter of type ErrNotifier help you spot which methods require error handling.

Handling an error

scope() executes a function, and handles an error that has occurred inside there at the point when the function finishes according to the conditions specified by errorIf and criticalIf. Use both or either of them to set the conditions of whether to treat the function result as a non-critical/critical error. The condition of criticalIf is evaluated prior to that of errorIf.

If either of the conditions is met, the relevant handler, onError or onCriticalError, is called. Do some error handling in these handlers, like showing different messages depending on the severity of the error.

final result = await errFlow.scope<bool>(
  (notifier) => yourMethod(notifier),

  // Use both or only either of errorIf and criticalIf.
  errorIf: (result, error) => error == CustomError.foo,
  criticalIf: (result, error) => error == CustomError.bar,

  // There is a way to avoid writing onError and onCriticalError
  // every time, which is explained later.
  onError: (result, error) => _onError(result, error),
  onCriticalError: (result, error) => _onCriticalError(result, error),
);

The handler functions receive the function result and the error value, which means you can combine them to tweak the conditions for triggering the handlers.

e.g. To trigger the onError handler if any error was set:

errorIf: (result, error) => error != errFlow.defaultError

e.g. To trigger the onError handler when the process fails for reasons other than a connection error:

errorIf: (result, error) => !result && error != CustomError.connection

Handling an error manually

combiningScope() is useful when you want to manually check and handle an error, leaving only its logging to ErrFlow.

It basically works like loggingScope() (described later), but returns CombinedResult that has both a function result and an error. Using the combined result, it is possible to handle the error after the scope finishes.

final result = await errFlow.combiningScope(
  (notifier) => yourMethod(notifier),
);

if (result.hasError) {
  switch (result.error!) {
    case AppError.init:
      print('[ERROR] Initialization failed.');
  }
}

CombinedResult is similar to Result of package:async, but with some clear differences:

  • It has both a value and an error, whereas Result has only either of them.
  • It always has a value regardless of an error.
  • The default error value is set to error if there was no error.
    • Having a non-null value in error does not always mean an error has occurred.

Ignoring errors

If a method, in which set() can be used, is called from some different places in your code, you may want to show an error message at some of them but not at the other places. It is possible with the use of loggingScope() and ignorableScope(), allowing you to only log errors without handling them, or ignore them completely.

loggingScope()

notifier passed from loggingScope() is an object of LoggingErrNotifier. Calling set() on that object only updates the value of lastError and triggers the logger (and added listener functions), without triggering the error handlers.

final result = await errFlow.loggingScope<bool>(
  (LoggingErrNotifier notifier) => yourMethod(notifier),
);

bool yourMethod(ErrNotifier notifier) {
  try {
    return ...;
  } catch (e, s) {
    notifier.set(CustomError.foo, e, s);  // Only updates lastError and logs the error.
    return false;
  }
}

ignorableScope()

notifier passed from ignorableScope() is an object of IgnorableErrNotifier. Calling set() and log() on that object does not trigger the error handlers nor the logger. set() only updates the value of lastError.

final result = await errFlow.ignorableScope<bool>(
  (IgnorableErrNotifier notifier) => yourMethod(notifier),
);

bool yourMethod(ErrNotifier notifier) {
  try {
    return ...;
  } catch (e, s) {
    notifier.set(CustomError.foo, e, s);  // Only updates lastError.
    return false;
  }
}

Default error handlers

You may want to consistently use a particular handler for non-critical errors, and the same or another one for critical errors. In such a case, errorHandler and criticalErrorHandler will come in handy. You can specify in advance how errors should be handled, and omit onError and onCriticalError in each scope().

void _errorHandler<T>(T result, CustomError? error) {
  switch (error) {
    case CustomError.foo:
      // Handle the foo error (e.g. showing the error details)
      break;
    default:
      // Handle other errors
      break;
  }
}

...

errFlow
  ..errorHandler = _errorhandler
  ..criticalErrorHandler = _errorHandler;

final result = await errFlow.scope<bool>(
  (notifier) => yourMethod(notifier),
  errorIf: (result, error) => !result,
);

Logger

To use the default logger, which simply prints information to the console, call useDefaultLogger() before the first logging.

errFlow.useDefaultLogger();

If it lacks functionality you need, set your own logger.

// The return type can be Future or non-Future.
// Note: Even if a Future is returned, set() and log() won't await it.
void _logger(Object e, StackTrace? s, {Object? reason}) {
  // Logging operations
}

...

errFlow.logger = _logger;

In flutter, the recordError() method of the firebase_crashlytics package can be assigned to the logger as is.

import 'package:firebase_crashlytics/firebase_crashlytics.dart';

...

errFlow.logger = FirebaseCrashlytics.recordError;

Make sure to set the default or a custom logger, otherwise an assertion error will occur in the debug mode.

Adding/removing a listener

This is usually unnecessary, but you can add a custom listener for your special needs.

void _listener({CustomError? error, Object? exception, StackTrace? stack, Object? context}) {
  // Some processing
}

...

errFlow.addListener(_listener);

...

errFlow.removeListener(_listener);

Libraries

errflow