diff options
Diffstat (limited to 'website2/templates/features/SneakyThrows.html')
| -rw-r--r-- | website2/templates/features/SneakyThrows.html | 44 |
1 files changed, 44 insertions, 0 deletions
diff --git a/website2/templates/features/SneakyThrows.html b/website2/templates/features/SneakyThrows.html new file mode 100644 index 00000000..5a2d5bbd --- /dev/null +++ b/website2/templates/features/SneakyThrows.html @@ -0,0 +1,44 @@ +<#import "_features.html" as f> + +<@f.scaffold title="@SneakyThrows" logline="To boldly throw checked exceptions where no one has thrown them before!"> + <@f.overview> + <p> + <code>@SneakyThrows</code> can be used to sneakily throw checked exceptions without actually declaring this in your method's <code>throws</code> clause. This somewhat contentious ability should be used carefully, of course. The code generated by lombok will not ignore, wrap, replace, or otherwise modify the thrown checked exception; it simply fakes out the compiler. On the JVM (class file) level, all exceptions, checked or not, can be thrown regardless of the <code>throws</code> clause of your methods, which is why this works. + </p><p> + Common use cases for when you want to opt out of the checked exception mechanism center around 2 situations:<br /> + <ul> + <li> + A needlessly strict interface, such as <code>Runnable</code> - whatever exception propagates out of your <code>run()</code> method, checked or not, it will be passed to the <code>Thread</code>'s unhandled exception handler. Catching a checked exception and wrapping it in some sort of <code>RuntimeException</code> is only obscuring the real cause of the issue. + </li><li> + An 'impossible' exception. For example, <code>new String(someByteArray, "UTF-8");</code> declares that it can throw an <code>UnsupportedEncodingException</code> but according to the JVM specification, UTF-8 <em>must</em> always be available. An <code>UnsupportedEncodingException</code> here is about as likely as a <code>ClassNotFoundError</code> when you use a String object, and you don't catch those either! + </li> + </ul> + </p><p> + Be aware that it is <em>impossible</em> to catch sneakily thrown checked types directly, as javac will not let you write a catch block for an exception type that no method call in the try body declares as thrown. This problem is not relevant in either of the use cases listed above, so let this serve as a warning that you should not use the <code>@SneakyThrows</code> mechanism without some deliberation! + </p><p> + You can pass any number of exceptions to the <code>@SneakyThrows</code> annotation. If you pass no exceptions, you may throw any exception sneakily. + </p> + </@f.overview> + + <@f.snippets name="SneakyThrows" /> + + <@f.confKeys> + <dt> + <code>lombok.sneakyThrows.flagUsage</code> = [<code>warning</code> | <code>error</code>] (default: not set) + </dt><dd> + Lombok will flag any usage of <code>@SneakyThrows</code> as a warning or error if configured. + </dd> + </@f.confKeys> + + <@f.smallPrint> + <p> + Because <code>@SneakyThrows</code> is an implementation detail and not part of your method signature, it is an error if you try to declare a checked exception as sneakily thrown when you don't call any methods that throw this exception. (Doing so is perfectly legal for <code>throws</code> statements to accommodate subclasses). Similarly, <code>@SneakyThrows</code> does not inherit. + </p><p> + For the nay-sayers in the crowd: Out of the box, Eclipse will offer a 'quick-fix' for uncaught exceptions that wraps the offending statement in a try/catch block with just <code>e.printStackTrace()</code> in the catch block. This is so spectacularly non-productive compared to just sneakily throwing the exception onwards, that Roel and Reinier feel more than justified in claiming that the checked exception system is far from perfect, and thus an opt-out mechanism is warranted. + </p><p> + If you put <code>@SneakyThrows</code> on a constructor, any call to a sibling or super constructor is <em>excluded</em> from the <code>@SneakyThrows</code> treatment. This is a java restriction we cannot work around: Calls to sibling/super constructors MUST be the first statement in the constructor; they cannot be placed inside try/catch blocks. + </p><p> + <code>@SneakyThrows</code> on an empty method, or a constructor that is empty or only has a call to a sibling / super constructor results in no try/catch block and a warning. + </p> + </@f.smallPrint> +</@f.scaffold> |