On the call site side, it seems cleaner to catch an IAE and log the fact
that some programming error took place. If I catch NPE, I cannot say that
there is a bug or that someone is passing a known bad value, all I can say
is that a null object is where the program does not allow it.

We'll have to agree to disagree.

Gary


On Wed, Apr 23, 2014 at 10:49 AM, Matt Sicker <[email protected]> wrote:

> I used to think like that, but that was due to the fact that not a single
> NPE in the JDK includes an error message. Thus, NPE gets the reputation of
> being a completely useless exception to find in your stack track. In my
> opinion, IAE makes sense when the argument doesn't follow the rules other
> than nullity. For example, I changed the isInstanceOf method to not throw
> an exception if the marker doesn't exist, but if it were kept as an
> exception, I'd use an IAE since it indicates the given marker is not a
> valid argument.
>
> This is pretty consistent with the JDK with the added bonus of useful
> exception messages.
>
>
> On 23 April 2014 08:25, Gary Gregory <[email protected]> wrote:
>
>> I like IAE better than NPE because it shows log4j knows the argument is
>> not allowed, null or otherwise, where NPE says (to me) 'some object was
>> null and I blew up'. IAE is more of a 'this is not allowed by design,
>> definitively not a bug in log4j'.
>>
>> Gary
>>
>> ---------- Forwarded message ----------
>> From: <[email protected]>
>> Date: Wed, Apr 23, 2014 at 10:20 AM
>> Subject: svn commit: r1589425 -
>> /logging/log4j/log4j2/trunk/log4j-api/src/main/java/org/apache/logging/log4j/MarkerManager.java
>> To: [email protected]
>>
>>
>> Author: mattsicker
>> Date: Wed Apr 23 14:20:56 2014
>> New Revision: 1589425
>>
>> URL: http://svn.apache.org/r1589425
>> Log:
>> Update null handling.
>>
>>   - Don't allow null Marker names. They won't work anyway due to the use
>> of ConcurrentHashMap.
>>   - Update methods to throw NullPointerException instead of
>> IllegalArgumentException to match updated javadocs.
>>   - Simplify equals and hashCode based on guaranteed non-nullity of
>> marker name.
>>
>> Modified:
>>
>> logging/log4j/log4j2/trunk/log4j-api/src/main/java/org/apache/logging/log4j/MarkerManager.java
>>
>> Modified:
>> logging/log4j/log4j2/trunk/log4j-api/src/main/java/org/apache/logging/log4j/MarkerManager.java
>> URL:
>> http://svn.apache.org/viewvc/logging/log4j/log4j2/trunk/log4j-api/src/main/java/org/apache/logging/log4j/MarkerManager.java?rev=1589425&r1=1589424&r2=1589425&view=diff
>>
>> ==============================================================================
>> ---
>> logging/log4j/log4j2/trunk/log4j-api/src/main/java/org/apache/logging/log4j/MarkerManager.java
>> (original)
>> +++
>> logging/log4j/log4j2/trunk/log4j-api/src/main/java/org/apache/logging/log4j/MarkerManager.java
>> Wed Apr 23 14:20:56 2014
>> @@ -36,6 +36,7 @@ public final class MarkerManager {
>>       * Retrieve a Marker or create a Marker that has no parent.
>>       * @param name The name of the Marker.
>>       * @return The Marker with the specified name.
>> +     * @throws NullPointerException if the argument is {@code null}
>>       */
>>      public static Marker getMarker(final String name) {
>>          markerMap.putIfAbsent(name, new Log4jMarker(name));
>> @@ -48,6 +49,7 @@ public final class MarkerManager {
>>       * @param parent The name of the parent Marker.
>>       * @return The Marker with the specified name.
>>       * @throws IllegalArgumentException if the parent Marker does not
>> exist.
>> +     * @throws NullPointerException if the argument is {@code null}
>>       * @deprecated Use the Marker add or set methods to add parent
>> Markers. Will be removed by final GA release.
>>       */
>>      @Deprecated
>> @@ -66,6 +68,7 @@ public final class MarkerManager {
>>       * @param name The name of the Marker.
>>       * @param parent The parent Marker.
>>       * @return The Marker with the specified name.
>> +     * @throws NullPointerException if the argument is {@code null}
>>       * @deprecated Use the Marker add or set methods to add parent
>> Markers. Will be removed by final GA release.
>>       */
>>      @Deprecated
>> @@ -83,15 +86,27 @@ public final class MarkerManager {
>>          private final String name;
>>          private volatile Marker[] parents;
>>
>> +        /**
>> +         * Constructs a new Marker.
>> +         * @param name the name of the Marker.
>> +         * @throws NullPointerException if the argument is {@code null}
>> +         */
>>          public Log4jMarker(final String name) {
>> +            if (name == null) {
>> +                // we can't store null references in a ConcurrentHashMap
>> as it is, not to mention that a null Marker
>> +                // name seems rather pointless. To get an "anonymous"
>> Marker, just use an empty string.
>> +                throw new NullPointerException("Marker name cannot be
>> null.");
>> +            }
>>              this.name = name;
>>              this.parents = null;
>>          }
>>
>> +        // TODO: use java.util.concurrent
>> +
>>          @Override
>>          public synchronized Marker add(final Marker parent) {
>>              if (parent == null) {
>> -                throw new IllegalArgumentException("A parent marker must
>> be specified");
>> +                throw new NullPointerException("A parent marker must be
>> specified");
>>              }
>>              // It is not strictly necessary to copy the variable here
>> but it should perform better than
>>              // Accessing a volatile variable multiple times.
>> @@ -100,7 +115,6 @@ public final class MarkerManager {
>>              if (localParents != null && (contains(parent, localParents)
>> || parent.isInstanceOf(this))) {
>>                  return this;
>>              }
>> -            // FIXME: should we really be only increasing the array size
>> by 1 each time? why not something like log(n)?
>>              final int size = localParents == null ? 1 :
>> localParents.length + 1;
>>              final Marker[] markers = new Marker[size];
>>              if (localParents != null) {
>> @@ -113,10 +127,12 @@ public final class MarkerManager {
>>              return this;
>>          }
>>
>> +        // TODO: add(Marker parent, Marker... moreParents)
>> +
>>          @Override
>>          public synchronized boolean remove(final Marker parent) {
>>              if (parent == null) {
>> -                throw new IllegalArgumentException("A parent marker must
>> be specified");
>> +                throw new NullPointerException("A parent marker must be
>> specified");
>>              }
>>              final Marker[] localParents = this.parents;
>>              if (localParents == null) {
>> @@ -184,7 +200,7 @@ public final class MarkerManager {
>>          @Override
>>          public boolean isInstanceOf(final Marker marker) {
>>              if (marker == null) {
>> -                throw new IllegalArgumentException("A marker parameter
>> is required");
>> +                throw new NullPointerException("A marker parameter is
>> required");
>>              }
>>              if (this == marker) {
>>                  return true;
>> @@ -213,8 +229,7 @@ public final class MarkerManager {
>>          @Override
>>          public boolean isInstanceOf(final String markerName) {
>>              if (markerName == null) {
>> -                // FIXME: and normally you'd throw an NPE here (ditto
>> for other cases above)
>> -                throw new IllegalArgumentException("A marker name is
>> required");
>> +                throw new NullPointerException("A marker name is
>> required");
>>              }
>>              if (markerName.equals(this.getName())) {
>>                  return true;
>> @@ -222,8 +237,7 @@ public final class MarkerManager {
>>              // Use a real marker for child comparisons. It is faster
>> than comparing the names.
>>              final Marker marker = markerMap.get(markerName);
>>              if (marker == null) {
>> -                // FIXME: shouldn't this just return false?
>> -                throw new IllegalArgumentException("No marker exists
>> with the name " + markerName);
>> +                return false;
>>              }
>>              final Marker[] localParents = parents;
>>              if (localParents != null) {
>> @@ -292,19 +306,13 @@ public final class MarkerManager {
>>              if (o == null || !(o instanceof Marker)) {
>>                  return false;
>>              }
>> -
>>              final Marker marker = (Marker) o;
>> -
>> -            if (name != null ? !name.equals(marker.getName()) :
>> marker.getName() != null) {
>> -                return false;
>> -            }
>> -
>> -            return true;
>> +            return name.equals(marker.getName());
>>          }
>>
>>          @Override
>>          public int hashCode() {
>> -            return name != null ? name.hashCode() : 0;
>> +            return name.hashCode();
>>          }
>>
>>          @Override
>>
>>
>>
>>
>>
>> --
>> E-Mail: [email protected] | [email protected]
>> Java Persistence with Hibernate, Second 
>> Edition<http://www.manning.com/bauer3/>
>> JUnit in Action, Second Edition <http://www.manning.com/tahchiev/>
>> Spring Batch in Action <http://www.manning.com/templier/>
>> Blog: http://garygregory.wordpress.com
>> Home: http://garygregory.com/
>> Tweet! http://twitter.com/GaryGregory
>>
>
>
>
> --
> Matt Sicker <[email protected]>
>



-- 
E-Mail: [email protected] | [email protected]
Java Persistence with Hibernate, Second Edition<http://www.manning.com/bauer3/>
JUnit in Action, Second Edition <http://www.manning.com/tahchiev/>
Spring Batch in Action <http://www.manning.com/templier/>
Blog: http://garygregory.wordpress.com
Home: http://garygregory.com/
Tweet! http://twitter.com/GaryGregory

Reply via email to