Related
I'm looking at the article C# - Data Transfer Object on serializable DTOs.
The article includes this piece of code:
public static string SerializeDTO(DTO dto) {
try {
XmlSerializer xmlSer = new XmlSerializer(dto.GetType());
StringWriter sWriter = new StringWriter();
xmlSer.Serialize(sWriter, dto);
return sWriter.ToString();
}
catch(Exception ex) {
throw ex;
}
}
The rest of the article looks sane and reasonable (to a noob), but that try-catch-throw throws a WtfException... Isn't this exactly equivalent to not handling exceptions at all?
Ergo:
public static string SerializeDTO(DTO dto) {
XmlSerializer xmlSer = new XmlSerializer(dto.GetType());
StringWriter sWriter = new StringWriter();
xmlSer.Serialize(sWriter, dto);
return sWriter.ToString();
}
Or am I missing something fundamental about error handling in C#? It's pretty much the same as Java (minus checked exceptions), isn't it? ... That is, they both refined C++.
The Stack Overflow question The difference between re-throwing parameter-less catch and not doing anything? seems to support my contention that try-catch-throw is-a no-op.
EDIT:
Just to summarise for anyone who finds this thread in future...
DO NOT
try {
// Do stuff that might throw an exception
}
catch (Exception e) {
throw e; // This destroys the strack trace information!
}
The stack trace information can be crucial to identifying the root cause of the problem!
DO
try {
// Do stuff that might throw an exception
}
catch (SqlException e) {
// Log it
if (e.ErrorCode != NO_ROW_ERROR) { // filter out NoDataFound.
// Do special cleanup, like maybe closing the "dirty" database connection.
throw; // This preserves the stack trace
}
}
catch (IOException e) {
// Log it
throw;
}
catch (Exception e) {
// Log it
throw new DAOException("Excrement occurred", e); // wrapped & chained exceptions (just like java).
}
finally {
// Normal clean goes here (like closing open files).
}
Catch the more specific exceptions before the less specific ones (just like Java).
References:
MSDN - Exception Handling
MSDN - try-catch (C# Reference)
First, the way that the code in the article does it is evil. throw ex will reset the call stack in the exception to the point where this throw statement is losing the information about where the exception actually was created.
Second, if you just catch and re-throw like that, I see no added value. The code example above would be just as good (or, given the throw ex bit, even better) without the try-catch.
However, there are cases where you might want to catch and rethrow an exception. Logging could be one of them:
try
{
// code that may throw exceptions
}
catch(Exception ex)
{
// add error logging here
throw;
}
Don't do this,
try
{
...
}
catch(Exception ex)
{
throw ex;
}
You'll lose the stack trace information...
Either do,
try { ... }
catch { throw; }
OR
try { ... }
catch (Exception ex)
{
throw new Exception("My Custom Error Message", ex);
}
One of the reason you might want to rethrow is if you're handling different exceptions, for
e.g.
try
{
...
}
catch(SQLException sex)
{
//Do Custom Logging
//Don't throw exception - swallow it here
}
catch(OtherException oex)
{
//Do something else
throw new WrappedException("Other Exception occured");
}
catch
{
System.Diagnostics.Debug.WriteLine("Eeep! an error, not to worry, will be handled higher up the call stack");
throw; //Chuck everything else back up the stack
}
C# (before C# 6) doesn't support CIL "filtered exceptions", which VB does, so in C# 1-5 one reason for re-throwing an exception is that you don't have enough information at the time of catch() to determine whether you wanted to actually catch the exception.
For example, in VB you can do
Try
..
Catch Ex As MyException When Ex.ErrorCode = 123
..
End Try
...which would not handle MyExceptions with different ErrorCode values. In C# prior to v6, you would have to catch and re-throw the MyException if the ErrorCode was not 123:
try
{
...
}
catch(MyException ex)
{
if (ex.ErrorCode != 123) throw;
...
}
Since C# 6.0 you can filter just like with VB:
try
{
// Do stuff
}
catch (Exception e) when (e.ErrorCode == 123456) // filter
{
// Handle, other exceptions will be left alone and bubble up
}
My main reason for having code like:
try
{
//Some code
}
catch (Exception e)
{
throw;
}
is so I can have a breakpoint in the catch, that has an instantiated exception object. I do this a lot while developing/debugging. Of course, the compiler gives me a warning on all the unused e's, and ideally they should be removed before a release build.
They are nice during debugging though.
A valid reason for rethrowing exceptions can be that you want to add information to the exception, or perhaps wrap the original exception in one of your own making:
public static string SerializeDTO(DTO dto) {
try {
XmlSerializer xmlSer = new XmlSerializer(dto.GetType());
StringWriter sWriter = new StringWriter();
xmlSer.Serialize(sWriter, dto);
return sWriter.ToString();
}
catch(Exception ex) {
string message =
String.Format("Something went wrong serializing DTO {0}", DTO);
throw new MyLibraryException(message, ex);
}
}
Isn't this exactly equivalent to not
handling exceptions at all?
Not exactly, it isn't the same. It resets the exception's stacktrace.
Though I agree that this probably is a mistake, and thus an example of bad code.
You don't want to throw ex - as this will lose the call stack. See Exception Handling (MSDN).
And yes, the try...catch is doing nothing useful (apart from lose the call stack - so it's actually worse - unless for some reason you didn't want to expose this information).
This can be useful when your programming functions for a library or dll.
This rethrow structure can be used to purposefully reset the call stack so that instead of seeing the exception thrown from an individual function inside the function, you get the exception from the function itself.
I think this is just used so that the thrown exceptions are cleaner and don't go into the "roots" of the library.
A point that people haven't mentioned is that while .NET languages don't really make a proper distinction, the question of whether one should take action when an exception occurs, and whether one will resolve it, are actually distinct questions. There are many cases where one should take action based upon exceptions one has no hope of resolving, and there are some cases where all that is necessary to "resolve" an exception is to unwind the stack to a certain point--no further action required.
Because of the common wisdom that one should only "catch" things one can "handle", a lot of code which should take action when exceptions occur, doesn't. For example, a lot of code will acquire a lock, put the guarded object "temporarily" into a state which violates its invariants, then put it object into a legitimate state, and then release the lock back before anyone else can see the object. If an exception occurs while the object is in a dangerously-invalid state, common practice is to release the lock with the object still in that state. A much better pattern would be to have an exception that occurs while the object is in a "dangerous" condition expressly invalidate the lock so any future attempt to acquire it will immediately fail. Consistent use of such a pattern would greatly improve the safety of so-called "Pokemon" exception handling, which IMHO gets a bad reputation primarily because of code which allows exceptions to percolate up without taking appropriate action first.
In most .NET languages, the only way for code to take action based upon an exception is to catch it (even though it knows it's not going to resolve the exception), perform the action in question and then re-throw). Another possible approach if code doesn't care about what exception is thrown is to use an ok flag with a try/finally block; set the ok flag to false before the block, and to true before the block exits, and before any return that's within the block. Then, within finally, assume that if ok isn't set, an exception must have occurred. Such an approach is semantically better than a catch/throw, but is ugly and is less maintainable than it should be.
While many of the other answers provide good examples of why you might want to catch an rethrow an exception, no one seems to have mentioned a 'finally' scenario.
An example of this is where you have a method in which you set the cursor (for example to a wait cursor), the method has several exit points (e.g. if () return;) and you want to ensure the cursor is reset at the end of the method.
To do this you can wrap all of the code in a try/catch/finally. In the finally set the cursor back to the right cursor. So that you don't bury any valid exceptions, rethrow it in the catch.
try
{
Cursor.Current = Cursors.WaitCursor;
// Test something
if (testResult) return;
// Do something else
}
catch
{
throw;
}
finally
{
Cursor.Current = Cursors.Default;
}
One possible reason to catch-throw is to disable any exception filters deeper up the stack from filtering down (random old link). But of course, if that was the intention, there would be a comment there saying so.
It depends what you are doing in the catch block, and if you are wanting to pass the error on to the calling code or not.
You might say Catch io.FileNotFoundExeption ex and then use an alternative file path or some such, but still throw the error on.
Also doing Throw instead of Throw Ex allows you to keep the full stack trace. Throw ex restarts the stack trace from the throw statement (I hope that makes sense).
In the example in the code you have posted there is, in fact, no point in catching the exception as there is nothing done on the catch it is just re-thown, in fact it does more harm than good as the call stack is lost.
You would, however catch an exception to do some logic (for example closing sql connection of file lock, or just some logging) in the event of an exception the throw it back to the calling code to deal with. This would be more common in a business layer than front end code as you may want the coder implementing your business layer to handle the exception.
To re-iterate though the There is NO point in catching the exception in the example you posted. DON'T do it like that!
Sorry, but many examples as "improved design" still smell horribly or can be extremely misleading. Having try { } catch { log; throw } is just utterly pointless. Exception logging should be done in central place inside the application. exceptions bubble up the stacktrace anyway, why not log them somewhere up and close to the borders of the system?
Caution should be used when you serialize your context (i.e. DTO in one given example) just into the log message. It can easily contain sensitive information one might not want to reach the hands of all the people who can access the log files. And if you don't add any new information to the exception, I really don't see the point of exception wrapping. Good old Java has some point for that, it requires caller to know what kind of exceptions one should expect then calling the code. Since you don't have this in .NET, wrapping doesn't do any good on at least 80% of the cases I've seen.
In addition to what the others have said, see my answer to a related question which shows that catching and rethrowing is not a no-op (it's in VB, but some of the code could be C# invoked from VB).
Most of answers talking about scenario catch-log-rethrow.
Instead of writing it in your code consider to use AOP, in particular Postsharp.Diagnostic.Toolkit with OnExceptionOptions IncludeParameterValue and
IncludeThisArgument
Rethrowing exceptions via throw is useful when you don't have a particular code to handle current exceptions, or in cases when you have a logic to handle specific error cases but want to skip all others.
Example:
string numberText = "";
try
{
Console.Write("Enter an integer: ");
numberText = Console.ReadLine();
var result = int.Parse(numberText);
Console.WriteLine("You entered {0}", result);
}
catch (FormatException)
{
if (numberText.ToLowerInvariant() == "nothing")
{
Console.WriteLine("Please, please don't be lazy and enter a valid number next time.");
}
else
{
throw;
}
}
finally
{
Console.WriteLine("Freed some resources.");
}
Console.ReadKey();
However, there is also another way of doing this, using conditional clauses in catch blocks:
string numberText = "";
try
{
Console.Write("Enter an integer: ");
numberText = Console.ReadLine();
var result = int.Parse(numberText);
Console.WriteLine("You entered {0}", result);
}
catch (FormatException) when (numberText.ToLowerInvariant() == "nothing")
{
Console.WriteLine("Please, please don't be lazy and enter a valid number next time.");
}
finally
{
Console.WriteLine("Freed some resources.");
}
Console.ReadKey();
This mechanism is more efficient than re-throwing an exception because
of the .NET runtime doesn’t have to rebuild the exception object
before re-throwing it.
Does using keyword catch or handle exceptions in C# when connecting to a database? Or should I use try catch block in all of my database methods inside the using? Does using try catch create unnecessary code?
using (var db = new ApplicationContext())
{
try {
/* Query something */
} catch(Exception e) {
logger.Debug(e);
}
}
Does using keyword catch or handle exceptions in C# when connecting to a database?
A using is logically equivalent to a try-finally, so yes, it handles exceptions, but it does not stop the exception. A finally propagates the exception.
should I use try catch block in all of my database methods inside the using?
No. The try-catch should go outside the using. That way it will protect the resource creation.
Does using try catch create unnecessary code?
I have no idea what this question means.
Some questions you did not ask:
Should I catch all exceptions for logging and then fail to re-throw them?
No. Only catch and eat exceptions that you know how to handle. If you want to log exceptions then re throw them when you're done logging; other code might want to handle them.
What is the correct way to write this code?
Separate your concerns. You have three concerns:
Dispose the resource
Log all exceptions
Handle expected exogenous exceptions
Each of those should be handled by a separate statement:
try // handle exogenous exceptions
{
try // log all exceptions
{
using(var foo = new Foo()) // dispose the resource
{
foo.Bar();
}
}
catch(Exception x)
{
// All exceptions are logged and re-thrown.
Log(x);
throw;
}
}
catch(FooException x)
{
// FooException is caught and handled
}
If your goal is to only log unhandled exceptions then invert the nesting of the two handlers, or use another mechanism such as the appdomain's unhandled exception event handler.
The using block does not "handle" exceptions for you, it only ensures that the Dispose() method gets called on the IDisposable object (in this case, your db instance), even in the event of an exception. So yes, you need to add try-catch blocks where needed.
That said, in general, you only want to catch exceptions where you can actually do something meaningful with them. If you only need to log the exceptions, consider doing your exception handling in a single location higher in the call stack so that you don't have to litter your code with try-catch blocks all over the place.
You can read about the using Statement here to see what it actually does, and how it gets translated.
EDIT:
If, for whatever reason, you choose to keep your try-catch block where it is, at the very least, make sure to rethrow the exception instead of swallowing it, which would be like sweeping the mess under the rug and pretending everything is fine. Also, make sure to rethrow it without losing your precious stack trace. Like this:
using (var db = new ApplicationContext())
{
try {
/* Query something */
} catch(Exception e) {
logger.Debug(e);
throw; // rethrows the exception without losing the stack trace.
}
}
EDIT 2: Very nice blog entry by Eric Lippert about exception handling.
Your using will converts to bottom code by the C# compiler, by 8.13 The using statement:
{
var db = new ApplicationContext();
try
{
try {
/* Query something */
} catch(Exception e) {
logger.Debug(e);
}
}
finally
{
// Check for a null resource.
if (db!= null)
// Call the object's Dispose method.
((IDisposable)myRes).Dispose();
}
}
So, my opinion, for your situation I think better without using statement, it will be a little bit clear and will have less steps:
var db = new ApplicationContext();
try
{
/* Query something */
}
catch(Exception e)
{
logger.Debug(e);
}
finally
{
if (db!= null)
{
((IDisposable)myRes).Dispose();
}
}
Because using it's just a syntax sugar.
P.S: The performance cost of try statement is very small and you can left your code as it is.
The using keyword is used to that after the block finishes, the object inside the using parameter, in this case, the context, is disposed of properly, and so yes, you still need to use try-catch keywords for handling exceptions.
Since using does not handle any exception, you may want to swap.
using is syntactic sugar for try-finally so you know that, even if exception is not handled on constructor, connection will be either closed or not established at all.
try {
using (var db = new ApplicationContext())
{
/* Query something */
}
} catch(Exception e) {
logger.Debug(e);
}
Apparently, some exceptions may just get lost while using nested using statement. Consider this simple console app:
using System;
namespace ConsoleApplication
{
public class Throwing: IDisposable
{
int n;
public Throwing(int n)
{
this.n = n;
}
public void Dispose()
{
var e = new ApplicationException(String.Format("Throwing({0})", this.n));
Console.WriteLine("Throw: {0}", e.Message);
throw e;
}
}
class Program
{
static void DoWork()
{
// ...
using (var a = new Throwing(1))
{
// ...
using (var b = new Throwing(2))
{
// ...
using (var c = new Throwing(3))
{
// ...
}
}
}
}
static void Main(string[] args)
{
AppDomain.CurrentDomain.UnhandledException += (sender, e) =>
{
// this doesn't get called
Console.WriteLine("UnhandledException:", e.ExceptionObject.ToString());
};
try
{
DoWork();
}
catch (Exception e)
{
// this handles Throwing(1) only
Console.WriteLine("Handle: {0}", e.Message);
}
Console.ReadLine();
}
}
}
Each instance of Throwing throws when it gets disposed of. AppDomain.CurrentDomain.UnhandledException never gets called.
The output:
Throw: Throwing(3)
Throw: Throwing(2)
Throw: Throwing(1)
Handle: Throwing(1)
I prefer to at least be able to log the missing Throwing(2) and Throwing(3). How do I do this, without resorting to a separate try/catch for each using (which would kinda kill the convenience of using)?
In real life, those objects are often instances of classes over which I have no control. They may or may not be throwing, but in case they do, I'd like to have an option to observe such exceptions.
This question came along while I was looking at reducing the level of nested using. There's a neat answer suggesting aggregating exceptions. It's interesting how this is different from the standard behavior of nested using statements.
[EDITED] This question appears to be closely related:
Should you implement IDisposable.Dispose() so that it never throws?
There's a code analyzer warning for this. CA1065, "Do not raise exceptions in unexpected locations". The Dispose() method is on that list. Also a strong warning in the Framework Design Guide, chapter 9.4.1:
AVOID throwing an exception from within Dispose(bool) except under critical situations where the containing process has been corrupted (leaks, inconsistent shared state, etc.).
This goes wrong because the using statement calls Dispose() inside a finally block. An exception raised in a finally block can have an unpleasant side-effect, it replaces an active exception if the finally block was called while the stack is being unwound because of an exception. Exactly what you see happening here.
Repro code:
class Program {
static void Main(string[] args) {
try {
try {
throw new Exception("You won't see this");
}
finally {
throw new Exception("You'll see this");
}
}
catch (Exception ex) {
Console.WriteLine(ex.Message);
}
Console.ReadLine();
}
}
What you are noticing is a fundamental problem in the design of Dispose and using, for which no nice solution as yet exists. IMHO the best design would be to have a version of Dispose which receives as an argument any exception which may be pending (or null, if none is pending), and can either log or encapsulate that exception if it needs to throw one of its own. Otherwise, if you have control of both the code which could cause an exception within the using as well as within the Dispose, you may be able to use some sort of outside data channel to let the Dispose know about the inner exception, but that's rather hokey.
It's too bad there's no proper language support for code associated with a finally block (either explicitly, or implicitly via using) to know whether the associated try completed properly and if not, what went wrong. The notion that Dispose should silently fail is IMHO very dangerous and wrongheaded. If an object encapsulates a file which is open for writing, and Dispose closes the file (a common pattern) and the data cannot be written, having the Dispose call return normally would lead the calling code to believe the data was written correctly, potentially allowing it to overwrite the only good backup. Further, if files are supposed to be closed explicitly and calling Dispose without closing a file should be considered an error, that would imply that Dispose should throw an exception if the guarded block would otherwise complete normally, but if the guarded block fails to call Close because an exception occurred first, having Dispose throw an exception would be very unhelpful.
If performance isn't critical, you could write a wrapper method in VB.NET which would accept two delegates (of types Action and an Action<Exception>), call the first within a try block, and then call the second in a finally block with the exception that occurred in the try block (if any). If the wrapper method was written in VB.NET, it could discover and report the exception that occurred without having to catch and rethrow it. Other patterns would be possible as well. Most usages of the wrapper would involve closures, which are icky, but the wrapper could at least achieve proper semantics.
An alternative wrapper design which would avoid closures, but would require that clients use it correctly and would provide little protection against incorrect usage would have a usage batter like:
var dispRes = new DisposeResult();
...
try
{
.. the following could be in some nested routine which took dispRes as a parameter
using (dispWrap = new DisposeWrap(dispRes, ... other disposable resources)
{
...
}
}
catch (...)
{
}
finally
{
}
if (dispRes.Exception != null)
... handle cleanup failures here
The problem with this approach is that there's no way to ensure that anyone will ever evaluate dispRes.Exception. One could use a finalizer to log cases where dispRes gets abandoned without ever having been examined, but there would be no way to distinguish cases where that occurred because an exception kicked code out beyond the if test, or because the programmer simply forgot the check.
PS--Another case where Dispose really should know whether exceptions occur is when IDisposable objects are used to wrap locks or other scopes where an object's invariants may temporarily be invalidated but are expected to be restored before code leaves the scope. If an exception occurs, code should often have no expectation of resolving the exception, but should nonetheless take action based upon it, leaving the lock neither held nor released but rather invalidated, so that any present or future attempt to acquire it will throw an exception. If there are no future attempts to acquire the lock or other resource, the fact that it is invalid should not disrupt system operation. If the resource is critically necessary to some part of the program, invalidating it will cause that part of the program to die while minimizing the damage it does to anything else. The only way I know to really implement this case with nice semantics is to use icky closures. Otherwise, the only alternative is to require explicit invalidate/validate calls and hope that any return statements within the part of the code where the resource is invalid are preceded by calls to validate.
Maybe some helper function that let you write code similar to using:
void UsingAndLog<T>(Func<T> creator, Action<T> action) where T:IDisposabe
{
T item = creator();
try
{
action(item);
}
finally
{
try { item.Dispose();}
catch(Exception ex)
{
// Log/pick which one to throw.
}
}
}
UsingAndLog(() => new FileStream(...), item =>
{
//code that you'd write inside using
item.Write(...);
});
Note that I'd probably not go this route and just let exceptions from Dispose to overwrite my exceptions from code inside normal using. If library throws from Dispose against strong recommendations not to do so there is a very good chance that it is not the only issue and usefulness of such library need to be reconsidered.
I don't have a problem; I'm just curious. Imagine the following scenario:
foreach (var foo in list)
{
try
{
//Some code
}
catch (Exception)
{
//Some more code
}
finally
{
continue;
}
}
This won't compile, as it raises compiler error CS0157:
Control cannot leave the body of a finally clause
Why?
finally blocks run whether an exception is thrown or not. If an exception is thrown, what the heck would continue do? You cannot continue execution of the loop, because an uncaught exception will transfer control to another function.
Even if no exception is thrown, finally will run when other control transfer statements inside the try/catch block run, like a return, for example, which brings the same problem.
In short, with the semantics of finally it doesn't make sense to allow transferring control from inside a finally block to the outside of it.
Supporting this with some alternative semantics would be more confusing than helpful, since there are simple workarounds that make the intended behaviour way clearer. So you get an error, and are forced to think properly about your problem. It's the general "throw you into the pit of success" idea that goes on in C#.
If you want to ignore exceptions (more often than not is a bad idea) and continue executing the loop, use a catch all block:
foreach ( var in list )
{
try{
//some code
}catch{
continue;
}
}
If you want to continue only when no uncaught exceptions are thrown, just put continue outside the try-block.
Here is a reliable source:
A continue statement cannot exit a finally block (Section 8.10). When
a continue statement occurs within a finally block, the target of the
continue statement must be within the same finally block; otherwise, a
compile-time error occurs.
It is taken from MSDN, 8.9.2 The continue statement.
The documentation say that:
The statements of a finally block are always executed when control leaves a
try statement. This is true whether the control transfer occurs as a
result of normal execution, as a result of executing a break,
continue, goto, or return statement, or as a result of propagating an
exception out of the try statement. If an exception is thrown during
execution of a finally block, the exception is propagated to the next
enclosing try statement. If another exception was in the process of
being propagated, that exception is lost. The process of propagating
an exception is discussed further in the description of the throw statement (Section 8.9.5).
It is from here 8.10 The try statement.
You may think it makes sense, but it doesn't make sense actually.
foreach (var v in List)
{
try
{
//Some code
}
catch (Exception)
{
//Some more code
break; or return;
}
finally
{
continue;
}
}
What do you intend to do a break or a continue when an exception is thrown? The C# compiler team doesn't want to make decision on their own by assuming break or continue. Instead, they decided to complain the developer situation will be ambiguous to transfer control from finally block.
So it is the job of developer to clearly state what he intends to do rather than compiler assuming something else.
I hope you understand why this doesn't compile!
As others have stated, but focused on exceptions, it's really about ambiguous handling of transferring control.
In your mind, you're probably thinking of a scenario like this:
public static object SafeMethod()
{
foreach(var item in list)
{
try
{
try
{
//do something that won't transfer control outside
}
catch
{
//catch everything to not throw exceptions
}
}
finally
{
if (someCondition)
//no exception will be thrown,
//so theoretically this could work
continue;
}
}
return someValue;
}
Theoretically, you can track the control flow and say, yes, this is "ok". No exception is thrown, no control is transferred. But the C# language designers had other issues in mind.
The Thrown Exception
public static void Exception()
{
try
{
foreach(var item in list)
{
try
{
throw new Exception("What now?");
}
finally
{
continue;
}
}
}
catch
{
//do I get hit?
}
}
The Dreaded Goto
public static void Goto()
{
foreach(var item in list)
{
try
{
goto pigsfly;
}
finally
{
continue;
}
}
pigsfly:
}
The Return
public static object ReturnSomething()
{
foreach(var item in list)
{
try
{
return item;
}
finally
{
continue;
}
}
}
The Breakup
public static void Break()
{
foreach(var item in list)
{
try
{
break;
}
finally
{
continue;
}
}
}
So in conclusion, yes, while there is a slight possibility of using a continue in situations where control isn't being transferred, but a good deal (majority?) of cases involve exceptions or return blocks. The language designers felt this would be too ambiguous and (likely) impossible to ensure at compile time that your continue is used only in cases where control flow is not being transferred.
In general continue does not make sense when used in finally block. Take a look at this:
foreach (var item in list)
{
try
{
throw new Exception();
}
finally{
//doesn't make sense as we are after exception
continue;
}
}
"This won't compile and I think it makes complete sense"
Well, I think it doesn't.
When you literally have catch(Exception) then you don't need the finally (and probably not even the continue).
When you have the more realistic catch(SomeException), what should happen when an exception is not caught? Your continue wants to go one way, the exception handling another.
You cannot leave the body of a finally block. This includes break, return and in your case continue keywords.
The finally block can be executed with an exception waiting to be rethrown. It wouldn't really make sense to be able to exit the block (by a continue or anything else) without rethrowing the exception.
If you want to continue your loop whatever happens, you do not need the finally statement: Just catch the exception and don't rethrow.
finally runs whether or not an uncaught exception is thrown. Others have already explained why this makes continue illogical, but here is an alternative that follows the spirit of what this code appears to be asking for. Basically, finally { continue; } is saying:
When there are caught exceptions, continue
When there are uncaught exceptions, allow them to be thrown, but still continue
(1) could be satisfied by placing continue at the end of each catch, and (2) could be satisfied by storing uncaught exceptions to be thrown later. You could write it like this:
var exceptions = new List<Exception>();
foreach (var foo in list) {
try {
// some code
} catch (InvalidOperationException ex) {
// handle specific exception
continue;
} catch (Exception ex) {
exceptions.Add(ex);
continue;
}
// some more code
}
if (exceptions.Any()) {
throw new AggregateException(exceptions);
}
Actually, finally would have also executed in the third case, where there were no exceptions thrown at all, caught or uncaught. If that was desired, you could of course just place a single continue after the try-catch block instead of inside each catch.
Technically speaking, it's a limitation of the underlying CIL. From the language spec:
Control transfer is never permitted to enter a catch handler or finally clause except through the exception handling mechanism.
and
Control transfer out of a protected region is only permitted through an exception instruction (leave, end.filter, end.catch, or end.finally)
On the doc page for the br instruction:
Control transfers into and out of try, catch, filter, and finally blocks cannot be performed by this instruction.
This last holds true for all branch instructions, including beq, brfalse, etc.
The designers of the language simply didn't want to (or couldn't) reason about the semantics of a finally block being terminated by a control transfer.
One issue, or perhaps the key issue, is that the finally block gets executed as part of some non-local control transfer (exception processing). The target of that control transfer isn't the enclosing loop; the exception processing aborts the loop and continues unwinding further.
If we have a control transfer out of the finally cleanup block, then the original control transfer is being "hijacked". It gets canceled, and control goes elsewhere.
The semantics can be worked out. Other languages have it.
The designers of C# decided to simply disallow static, "goto-like" control transfers, thereby simplifying things somewhat.
However, even if you do that, it doesn't solve the question of what happens if a dynamic transfer is initiated from a finally: what if the finally block calls a function, and that function throws? The original exception processing is then "hijacked".
If you work out the semantics of this second form of hijacking, there is no reason to banish the first type. They are really the same thing: a control transfer is a control transfer, whether it the same lexical scope or not.
I just had a pretty painful troubleshooting experience in troubleshooting some code that looked like this:
try {
doSomeStuff()
doMore()
} finally {
doSomeOtherStuff()
}
The problem was difficult to troubleshoot because doSomeStuff() threw an exception, which in turn caused doSomeOtherStuff() to also throw an exception. The second exception (thrown by the finally block) was thrown up to my code, but it did not have a handle on the first exception (thrown from doSomeStuff()), which was the real root-cause of the problem.
If the code had said this instead, the problem would have been readily apparent:
try {
doSomeStuff()
doMore()
} catch (Exception e) {
log.error(e);
} finally {
doSomeOtherStuff()
}
So, my question is this:
Is a finally block used without any catch block a well-known java anti-pattern? (It certainly seems to be a not-readily-apparent subclass of the obviously well-known anti-pattern "Don't gobble exceptions!")
In general, no, this is not an anti-pattern. The point of finally blocks is to make sure stuff gets cleaned up whether an exception is thrown or not. The whole point of exception handling is that, if you can't deal with it, you let it bubble up to someone who can, through the relatively clean out-of-band signaling exception handling provides. If you need to make sure stuff gets cleaned up if an exception is thrown, but can't properly handle the exception in the current scope, then this is exactly the correct thing to do. You just might want to be a little more careful about making sure your finally block doesn't throw.
I think the real "anti-pattern" here is doing something in a finally block that can throw, not not having a catch.
Not at all.
What's wrong is the code inside the finally.
Remember that finally will always get executed, and is just risky ( as you have just witnessed ) to put something that may throw an exception there.
There is absolutely nothing wrong a try with a finally and no catch. Consider the following:
InputStream in = null;
try {
in = new FileInputStream("file.txt");
// Do something that causes an IOException to be thrown
} finally {
if (in != null) {
try {
in.close();
} catch (IOException e) {
// Nothing we can do.
}
}
}
If an exception is thrown and this code doesn't know how to deal with it then the exception should bubble up the call stack to the caller. In this case we still want to clean up the stream so I think it makes perfect sense to have a try block without a catch.
I think it's far from being an anti-pattern and is something I do very frequently when it's critical do deallocate resources obtained during the method execution.
One thing I do when dealing with file handles (for writing) is flushing the stream before closing it using the IOUtils.closeQuietly method, which doesn't throw exceptions:
OutputStream os = null;
OutputStreamWriter wos = null;
try {
os = new FileOutputStream(...);
wos = new OutputStreamWriter(os);
// Lots of code
wos.flush();
os.flush();
finally {
IOUtils.closeQuietly(wos);
IOUtils.closeQuietly(os);
}
I like doing it that way for the following reasons:
It's not completely safe to ignore an exception when closing a file - if there are bytes that were not written to the file yet, then the file may not be in the state the caller would expect;
So, if an exception is raised during the flush() method, it will be propagated to the caller but I still will make sure all the files are closed. The method IOUtils.closeQuietly(...) is less verbose then the corresponding try ... catch ... ignore me block;
If using multiple output streams the order for the flush() method is important. The streams that were created by passing other streams as constructors should be flushed first. The same thing is valid for the close() method, but the flush() is more clear in my opinion.
I'd say a try block without a catch block is an anti-pattern. Saying "Don't have a finally without a catch" is a subset of "Don't have a try without a catch".
I use try/finally in the following form :
try{
Connection connection = ConnectionManager.openConnection();
try{
//work with the connection;
}finally{
if(connection != null){
connection.close();
}
}
}catch(ConnectionException connectionException){
//handle connection exception;
}
I prefer this to the try/catch/finally (+ nested try/catch in the finally).
I think that it is more concise and I don't duplicate the catch(Exception).
try {
doSomeStuff()
doMore()
} catch (Exception e) {
log.error(e);
} finally {
doSomeOtherStuff()
}
Don't do that either... you just hid more bugs (well not exactly hid them... but made it harder to deal with them. When you catch Exception you are also catching any sort of RuntimeException (like NullPointer and ArrayIndexOutOfBounds).
In general, catch the exceptions you have to catch (checked exceptions) and deal with the others at testing time. RuntimeExceptions are designed to be used for programmer errors - and programmer errors are things that should not happen in a properly debugged program.
In my opinion, it's more the case that finally with a catch indicate some kind of problem. The resource idiom is very simple:
acquire
try {
use
} finally {
release
}
In Java you can have an exception from pretty much anywhere. Often the acquire throws a checked exception, the sensible way to handle that is to put a catch around the how lot. Don't try some hideous null checking.
If you were going to be really anal you should note that there are implied priorities among exceptions. For instance ThreadDeath should clobber all, whether it comes from acquire/use/release. Handling these priorities correctly is unsightly.
Therefore, abstract your resource handling away with the Execute Around idiom.
Try/Finally is a way to properly free resources. The code in the finally block should NEVER throw since it should only act on resources or state that was acquired PRIOR to entry into the try block.
As an aside, I think log4J is almost an anti-pattern.
IF YOU WANT TO INSPECT A RUNNING PROGRAM USE A PROPER INSPECTION TOOL (i.e. a debugger, IDE, or in an extreme sense a byte code weaver but DO NOT PUT LOGGING STATEMENTS IN EVERY FEW LINES!).
In the two examples you present the first one looks correct. The second one includes the logger code and introduces a bug. In the second example you suppress an exception if one is thrown by the first two statements (i.e. you catch it and log it but do not rethrow. This is something I find very common in log4j usage and is a real problem of application design. Basically with your change you make the program fail in an way that would be very hard for the system to handle since you basically march onward as if you never had an exception (sorta like VB basic on error resume next construct).
try-finally may help you to reduce copy-paste code in case a method has multiple return statements. Consider the following example (Android Java):
boolean doSomethingIfTableNotEmpty(SQLiteDatabase db) {
Cursor cursor = db.rawQuery("SELECT * FROM table", null);
if (cursor != null) {
try {
if (cursor.getCount() == 0) {
return false;
}
} finally {
// this will get executed even if return was executed above
cursor.close();
}
}
// database had rows, so do something...
return true;
}
If there was no finally clause, you might have to write cursor.close() twice: just before return false and also after the surrounding if clause.
I think that try with no catch is anti-pattern. Using try/catch to handle exceptional conditions (file IO errors, socket timeout, etc) is not an anti-pattern.
If you're using try/finally for cleanup, consider a using block instead.