[WIP] ILogger integration

[m0sa]

initial stab at ILogger integration, seems like we're getting duplicate entries now (host as well as middleware logging the same exception)

[m0sa]

Alternatively, we could attach something to Exception.Data, to prevent the same exception from being logged twice.

A reasonable case of this happening would be if an exception gets logged to an ILogger, is then rethrown, is ultimately captured by the middleware, and is logged again...

[AmyJeanes]

I ran into exactly that while doing my own kinda version of this and doing that worked great:

/// <summary>
/// This is used internally to stop an exception from being logged twice
/// <para />
/// This is usually needed when code closer to the source logs extra information e.g. ids
/// and then rethrows to trigger the error middleware, which ends up logging it again
/// <para />
/// Returns true if the exception has already been logged, or false otherwise
/// </summary>
/// <param name="exception"></param>
public static bool ExceptionLogged(Exception exception)
{
    var key = $"Exceptional.Logged";
    var logged = exception.Data.Contains(key);
    if (!logged)
    {
        exception.Data[key] = true;
    }
    return logged;
}

[NickCraver]

I got this working, tests in, etc. But there's a problem. We've started out with bad assumptions of basic flows here, and that's my fault. This needs a poke at possibly refactoring the queue behavior in an ErrorStore (maybe a timer instead of a thread...but that's not without downsides) and the ILogger only creates the Error and queues, but does not log to the store. Retry background behavior (which already exists, just not 1:1 with what we need here) would be to actually log the errors to the store. This needs to handle fetching behavior as well, as is currently the case only when in retry today but we can simulate that with a length check on the queue.

Going to poke this some more tonight and hopefully have a 2.2 RTM release this week.

Additional thought: this should be optional - we need to make it configurable.

[mmillican]

@NickCraver Any ideas on next steps here? I'd love to help out if you have any other ideas for your last comment.

[mrahhal]

This has been open for a long while. Any plans on ever continuing this? This sounds like a really great thing to have and would allow Exceptional to be the central place for viewing error-ish messages.