This title begs for more explanation.
Basically, I'm rolling an API that wraps a web service in a nice heirarchical model. This model exposes things in the form:
var obj = MyRemoteResource.GetForId(1234, SourceEnum.ThatSource);
ApiConsumerMethod(obj.SomeProperty); //SomeProperty is lazily loaded, and often exposes such lazily loaded properties itself
... etc ...
Where many different RemoteResources* (each with many properties exist). There's really aggresive cacheing going on, and request throttling to prevent inadvertantly DOS'ing the servers (and getting the IP of the caller banned).
I've got all this code working, but I'm not doing much in the way of error handling at the moment. Basically, if the consumer of an API provides an invalid ID, the web server is down, a connection times out, or any other of a plethora of request layer errors occur an exception just percolates up when a property is accessed.
I consider this far less than ideal.
So my question is, how should I wrap these errors up so that it is convenient for a user of this API to manage them?
Some routes I have considered:
Just wrap all exceptions in some API defined ones, and document them as thrown.
Expose a static ErrorHandler class that allows a user to register notification callbacks for specific errors; falling back to the above behavior when no registration has been made for specific errors.**
Null properties on error, and set a LastErrorCode.
Each of these approachs have strengths and weaknesses. I'd appreciate opinons on them, as well as alternatives I haven't thought of.
If it impacts the discussion at all, the platform class throwing these exceptions is WebClient. Furthermore, usage of WebClient is sufficiently abstract that it could easily be replaced with some other download scheme if needed.
*Which is to say, many different classes
**This would be... wierd. But it maps to the global nature of the failure. This is my least favorite idea thus far.
I wouldn't implement fancy error technologies (like events and stuff like this). It's not easy to judge where and how to use exceptions, but this is no reason to implements other stuff.
When you request an object by an id which doesn't exist, what do you have to tell the caller about this? If you just return null, the client knows that it doesn't exist, there is nothing more to say.
Exceptions force the caller to care about it. So they should only be used where the caller is expected to do something special. Exception can provide the information why something didn't work. But if it is an "error" the user could also ignore, an exception is not the best choice.
There are two common alternatives to exceptions.
Use return values which provide information about the result of an action. For instance, logon could return a LogonResult instead of throwing an exception.
Write two methods, one throwing an exception, and one (Try...) returning a boolean. The caller decides if it wants to ignore the "error" or not.
Personally I believe this is entirely dependent on the context of what your API end-user is trying to do. If this is the case, you should let them decide how to handle erors.
Recently when working with CSV files (very error prone) I used an API that allowed you to define the exception behaviour. You could either ignore the error, replace by some default value/action, or pass them off to an event handler.
I really liked this approach of letting the end user decide how to handle internal exceptions, because unless your API is very specific its difficult to preempt all scenarios.
I prefer to have an exception thrown at me when something bad happens. Exceptions are (for me, anyways) the common way to tell that something went wrong, and i find it more intuitive to catch, than to check a return value for null and check another property to see what went wrong..
Related
I was creating a common interface to verify operationality of data providers when a question came to my mind.
Essentially the method i'm looking at has a signature like Task<bool> TryConnectAsync();
While that method clearly indicates whether or not a provider will be able to work, it provides no generic way to point out what to do, if it cannot work (false).
Looking at for example the AspNetCore.Identity package, that issue was resolved by using the return type IdentityResult for any method which can succeed/fail on top of providing additional information on how to resolve issue or point out what went wrong.
I do not want to use exceptions in this case, since a connection failing is not a scenario which is entirely unexpected and would make the method handling wonky since the method is either true, or it throws an exception which delivers details on what went wrong.
I also do not want to dump that information just into logs - i want a library consumer to be able to decide whether to display that information or not.
So my question is:
Is an informational class which provides success/failure like AspNetCore.Identity the standard practice, or what other ways of handling this are there?
For google: How to handle success or failure of a method and provide additional information.
Not sure if the question is right here tho. Might be better suited for SE.
I've noticed that most exception messages don't include instance-specific details like the value that caused the exception. They generally only tell you the "category" of the error.
For example, when attempting to serialize an object with a 3rd. party library, I got a MissingMethodException with message:
"No parameterless constructor defined for this object."
In many cases this is enough, but often (typically during development) a message like
"No parameterless constructor defined for this object of type 'Foo'."
can save a lot of time by directing you straight to the cause of the error.
InvalidArgumentException is another example: it usually tells you the name of the argument but not its value.
This seems to be the case for most framework-raised exceptions, but also for 3rd party libraries.
Is this done on purpose?
Is there a security implication in exposing an internal state like the "faulty" value of a variable?
Two reasons I can think of:
Firstly, maybe the parameter that threw the exception was a value that was a processed form of the one that was passed to the public interface. The value may not make sense without the expense of catching to rethrow a different exception that is going to be the same in most regards anyway.
Secondly, and more importantly, is that there can indeed be a security risk, that can be very hard to second-guess (if I'm writing a general-purpose container, I don't know what contexts it will be used in). We don't want "Credit-Card: 5555444455554444" appearing in an error message if we can help it.
Ultimately, just what debug information is most useful will vary according to the error anyway. If the type, method and (when possible) file and line number isn't enough, it's time to write some debug code that traps just what you do want to know, rather than complaining that it isn't already trapped when next time you might want yet different information (field state of instances can be just as likely to be useful as parameters).
InvalidArgumentException and (per #Ian Nelson) "Key not found in dictionary" both share something in common - there's no guarantee that the framework would be able to find a suitable value to show you - if the key/argument is of any user defined type, and ToString() hasn't been overridden, then you would just get the type name - it's not going to add a lot of value.
Exceptions are mostly meant for a program to consume. Most programs wouldn't know what to do with information about the instance.
The Message property is aimed at human consumption. Other than in a debugging scenario, humans won't know what to make of Foo.
Many exception mechanisms try to serve a hodgepodge of orthogonal purposes by passing a single exception-derived object:
Letting the caller know that various specific things have happened, or that various specific problems exist.
Determining when the exceptional condition is "resolved" so that normal program flow can continue.
Providing an indication of what to tell the user of the program
Providing information which could be logged to allow the owners of a system to identify problems, when a secure log is available
Providing information which could be logged to allow the owners of a system to identify problems, but which would not pose a security risk even if secure logging is not available.
Unfortunately, I'm unaware of any exception mechanism in widespread use which can actually accomplish all five of the above, well.
I have a situation where one of our developers wants to put a standard in place where we throw System.ComponentModel.DataAnnotations.ValidationExceptions in all our apps moving forward. An example would be that a user enters bad data into a form and our business logic layer throws a ValidationException which is handled at the calling layer.
However, I'm worried that this exception class is being used out of context and that one day we'll using some Dynamic Data Controls which make use of this exception and it will be difficult to tell the difference between when he's making use of the ValidationException vs times that the Dynamic Controls raise the exception.
We already use a custom exception class called something like "OurCustomException" and I think it would be better to just subclass off of that and create a OurCustomValidationException class. This way exceptions of different types can be clear cut.
Any opinions?
... it will be difficult to tell the
difference between when he's making
use of the ValidationException vs
times that the Dynamic Controls raise
the exception.
I think this is the main point you should be looking at when making that decision.
You seem to imply that the above (not being able to distinguish your own exceptions from "platform" validation exceptions) is a bad thing. That's not necessarily the case. IF you use the ValidationException exclusively to represent validation errors, then all your code can deal correctly with both your own and platform exceptions in the same way. No need to special-case platform exceptions from custom ones.
This is a win in my opinion. If you have both CustomException and ValidationException going back up to your toplevel layer for the same reasons, you'll have to repeat some logic one way or another. That's a bad thing (more maintenance, more chances of bugs creeping in).
So my opinion is that using the platform ValidationException is probably a good way to do it as long as you use it strictly for propagating validation problems.
Also think about the case where you would be giving/selling parts of your code to a third party (say it's real cool and you make a product out of it). It would probably be easier for the third party if your module throws "standard" exceptions (they can integrate it easily) rather than having to special-case all of his interface code for your module. Again, this is only valid if you stick with cases where a standard module would throw ValidationExceptions.
Lets look at it the other way around. You say:
our business logic layer throws a ValidationException
This is why I put strictly and exclusively above. You need to make sure you agree on what a validation error is. Lets look at two hypothetical problems:
"abc" is not a valid number
insufficient funds for this operation
For 1., the problem is simple/expected input validation error.
But in my opinion, 2. is not. It's a business logic problem. You could call it a validation issue (in the transaction, before the debit, you "validate" whether there are sufficient funds available) but I'd say its semantically very different.
I would advise not to put these two types of errors in the same exception "bag" as they have very different meanings, and may (often) lead to different application flow logic. (With the above two examples, 1. should keep the user on the very same form, as would any other "typo" kind of issue, but 2. should probably get him to a page that allows him to refill his account.)
To sum it up: using a standard exception seems like a good idea to me, as long as you stick with its expected semantics.
In Java, you explicitly define what exceptions are thrown using the "throws" keyword. That way, anyone calling your method knows what to catch.
Is there something in C#? If not, how do I know what exceptions to catch, or how do I let others know what exceptions to catch?
Also, if I am defining an interface, is there a way to say "methodX() should throw this exception on error"?
There is nothing equivalent in C#: The Trouble with Checked Exceptions
Other than documentation, there is no way to declare an interface to say "methodX() should throw this exception on error".
C#/.net does not have checked Exceptions, they proved to be less useful in large scale systems than first thought. In a lot of projects the time to maintain the check exception specs was a lot greater than the debugging time saved by having them.
Checked Exceptions seem like a good ideal until you have methods that can take delegates or calls into object you pass in. Take a simple case, the Sort() method on a list can’t know what exceptions it will throw, as it does not know what exceptions the Compar() method on the objects being sorted will throw.
So the spec for the exceptions a method may throw must be able to include information on how exceptions are populated from pass in objects and delegates. No one knows how to do this!
However there are tools that you check if you are catching all exceptions – see Exception Hunter by Red Gate. I personally don’t see much value in these tool, however if you like checked exceptions you may find them useful. ==> Looks like Exception Hunter wasn't too useful, so Redgate discontinued Exception Hunter a long while ago:
This feature is not available in C#. You can make proper XML documentation (3 slashes ///) and state what exceptions are being thrown.
This will be picked up by the IntelliSense mechanism and will be visible for the users of the class/method before they use it.
C# does not support this. (Not that I know anyway).
What you can do is use Xml Comments so that while calling you methods this data will be shown by intellisense.
As far as I'm aware there is no throws declaration in C# you can document your method indicating that it throws an exception but no forced error handling.
C# doesn't support checked exceptions. The language designers consider checked exceptions in the way java uses them a bad idea.
Some workarounds
Let me cite this medium article: It's almost 2020 and yet... Checked exceptions are still a thing
Among the many reasons why it's a bad idea, putting the checked exceptions in the contract (interfaces):
makes it impossible to change the implementation of an interface with a different one which throws different exceptions
exposes implementation details
a change of the checked exceptions of an API interface, makes it necessary to change the whole chain of interfaces in the call stack
For example, imagine that you are implementing a repository based on SQL Server, so you expose all kind of SQL Server specific exceptions. Then you want to move it to MySQL or Cosmos BD. Of course:
the implementation can't be changed to a new one that need to throw different exceptions. Also related to this, if you have different implementations of the storage, you can't just change them by configuration, but you need to have different compilations of the code for each storage backend
this is the explanation for 1: as the interface showed the implementation details (SQL Server exceptions) know you can't just change it
if you need to make the change, prepare to change the interface at all levels, since the API that uses the database up to the lates consumer in the call stack chain.
The articles cited above includes pointers to many explanations to discourage the use of checked exceptions, included this by the creator of C#: The trouble with checked exceptions
How do I implement error handling elegantly? For example, my data access layer can potentially throw 2 types of errors:
1) not authorized access, in which case the page should hide everything and just show the error message
2) errors that inform the user that something like this already exists in the database (say name not unique - for example), and in this case I wouldn't want to hide everything.
EDITED:
As a result of some comments here I devised that I should create derived specialized exception types, such as NotAuthorizedException, DuplicateException, etc etc.... it's all fine and dandy, however I can see 2 problems potentially:
1) Every stored proc has a return field p_error where it contains an error message. Upon getting the data from DB, I need to check this field to see what type of an error has been returned, so I can throw an appropriate exceptions. So, I still need to store my error types/error messages somewhere.....In other words, how do I should the exact message to the user (at certain times I need to) w/o checking the p_error field first. WHich brings me back to error object. Anyone?
2) I can this potentially turning into a nightmare where the number of exceptions equals the number of error message types.
Am I missing something here?
Much thanks to everyone!
You should check out the exception handling block in Enterprise Library. Lots of good tips and codeware surrounding wrapping exceptions and passing them between layers.
Where's your business layer, and why isn't it checking Authorization and integrity? The DAL is too low level to be checking those rules - if you hit a problem there, it's pretty much time to throw an exception. Your business layer or controllers can catch that exception, and display a reasonable message - but it's not something you should regularly be doing.
Create your own exception layer.
DALExceptionManager
DuplicateException
DatabaseException
BLLExceptionManager
NotAuthorizedException
InvalidDateException
In your Presentation Layer, Add this references and create a common exception handler.
In this way you know how to deal with the exception messages.
One option that I was thinking of
using is create an Error class, but
then I would need to pass it from UI
to business layer and the then to data
access layer by reference
I am not sure I understand this. You don't have to pass the error object in every layer. For example, in one of your example, errors that inform the user that something like this already exists in the database (say name not unique - for example) , a sql exception could be thrown by the framework, and you just need to catch the specific exception in your business layer, or UI layer.
Exception handling block by the Enterprise library suggested by other people will allow you define some policy-based exception handling in your web.config file. It could be good place if you want to develop some enterprise application. But for simple application, you may need not go that far.
What happens in the upper layers isn't up to your Data Access Layer. It shouldn't even be aware of what the upper layers are going to do. If you've got a duplicate key error, then it should throw something like a "DuplicateKeyException". If you should hit an authorization error (I presume you mean "exception"), then don't do anything with it - let it bubble back up to the UI layer, which can display an appropriate error page.
Remember that error status values and such things are the reason we invented exceptions.
Enterprise library exception handling block is the bomb as many have pointed out. Using policies you can do things like logging the exception, wrapping it in a different exception, throwing a new exception instead of the original one. Also, if you are looking to perform specific actions based on authentication errors or duplicate record errors etc, you could always create specific derived exception classes and catch those exception types which would precludes the need of passing any objects from the top down. Exceptions should always bubble up not down.