You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@mahout.apache.org by Sean Owen <sr...@gmail.com> on 2009/06/17 17:37:44 UTC
Re: svn commit: r785542 - in /lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix:
CardinalityException.java IndexException.java UnboundLabelException.java
Yeah, Benson noted the same about Eclipse -- we had a thread about
this a couple weeks ago. I personally think setting serialVersionUID
is a bad idea, and at least, is exceptional behavior. Unfortunately,
Eclipse defaults to warning about this.
The logic is this: when the JVM deserializes an object it needs to
check for compatibility between the serialized representation and its
current version of the class. It does so once, based on the
non-transient instance fields in the class. If it finds what it thinks
is a mismatch, it will throw an exception.
That is, by default, the deserialization is conservative: it will
never allow deserialization where an incompatibility exists, but will
sometimes disallow a serialization where, actually, there isn't a
compatibility problem.
That is, maybe you added a new instance field, but, you still think
it's valid to deserialize old versions of the object without that
field (maybe you are fine with leaving the new instance field as
null). To overcome this, you can keep track of what's compatible
yourself by setting serialVersionUID. As long as the old/new IDs
match, Java will try to proceed with the deserialization.
There is a use case for this, but, it's quite rare. If you never use
incompatbile versions of your classes in one context, it's not an
issue -- and indeed, that's not exactly a great idea. There is no
performance implication.
The problem is, by setting this field, you're leaving yourself quite
open to the opposite problem: adding a change that really isn't
backwards-compatible, and forgetting to update the ID. This is a big
issue.
Hence, I think it is almost never right to set serialVersionUID, and
certainly not a default. What do you all think of the logic? I think
Benson was on board with it. Now as to how to shut up Eclipse...
On Wed, Jun 17, 2009 at 4:20 PM, Jeff Eastman<jd...@windwardsolutions.com> wrote:
> Hi Sean,
>
> What's the logic for removing these? They now give me a warning in Eclipse.
> I'm willing to turn that off but I'd like to understand why.
>
> Jeff
>
>
> srowen@apache.org wrote:
>>
>> Author: srowen
>> Date: Wed Jun 17 10:25:43 2009
>> New Revision: 785542
>>
>> URL: http://svn.apache.org/viewvc?rev=785542&view=rev
>> Log:
>> Remove serialVersionUID per our standards (right?)
>>
>> Modified:
>>
>> lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/CardinalityException.java
>>
>> lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/IndexException.java
>>
>> lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/UnboundLabelException.java
>>
>> Modified:
>> lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/CardinalityException.java
>> URL:
>> http://svn.apache.org/viewvc/lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/CardinalityException.java?rev=785542&r1=785541&r2=785542&view=diff
>>
>> ==============================================================================
>> ---
>> lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/CardinalityException.java
>> (original)
>> +++
>> lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/CardinalityException.java
>> Wed Jun 17 10:25:43 2009
>> @@ -22,9 +22,4 @@
>> */
>> public class CardinalityException extends RuntimeException {
>> - /**
>> - * - */
>> - private static final long serialVersionUID = 1L;
>> -
>> }
>>
>> Modified:
>> lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/IndexException.java
>> URL:
>> http://svn.apache.org/viewvc/lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/IndexException.java?rev=785542&r1=785541&r2=785542&view=diff
>>
>> ==============================================================================
>> ---
>> lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/IndexException.java
>> (original)
>> +++
>> lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/IndexException.java
>> Wed Jun 17 10:25:43 2009
>> @@ -22,9 +22,4 @@
>> */
>> public class IndexException extends RuntimeException {
>> - /**
>> - * - */
>> - private static final long serialVersionUID = 1L;
>> -
>> }
>>
>> Modified:
>> lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/UnboundLabelException.java
>> URL:
>> http://svn.apache.org/viewvc/lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/UnboundLabelException.java?rev=785542&r1=785541&r2=785542&view=diff
>>
>> ==============================================================================
>> ---
>> lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/UnboundLabelException.java
>> (original)
>> +++
>> lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/UnboundLabelException.java
>> Wed Jun 17 10:25:43 2009
>> @@ -22,9 +22,4 @@
>> */
>> public class UnboundLabelException extends RuntimeException {
>> - /**
>> - * - */
>> - private static final long serialVersionUID = 1L;
>> -
>> }
>>
>>
>>
>>
>>
>
>
Re: svn commit: r785542 - in /lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix:
CardinalityException.java IndexException.java UnboundLabelException.java
Posted by Benson Margulies <bi...@gmail.com>.
At CXF, we have maven scripting to automatically set up all the
eclipse preferences, and that turns out to include spiking this
warning. I suppose I could set up a patch ... However, it involves
setting up a more complex maven project structure that I'm not sure is
desired here.
I wrote up a blog on it once:
http://dssheep.blogspot.com/2008/12/maven-eclipse-checkstyle-pmd-oy-vey.html
On Wed, Jun 17, 2009 at 12:02 PM, Jeff
Eastman<jd...@windwardsolutions.com> wrote:
> Thanks Sean I do recall that thread now. I've turned off the warning in
> Eclipse (Preferences/Java/Compiler/Errors-Warnings). Kind of moot on an
> exception with no state anyway.
>
> Jeff
>
>
> Sean Owen wrote:
>>
>> Yeah, Benson noted the same about Eclipse -- we had a thread about
>> this a couple weeks ago. I personally think setting serialVersionUID
>> is a bad idea, and at least, is exceptional behavior. Unfortunately,
>> Eclipse defaults to warning about this.
>>
>> The logic is this: when the JVM deserializes an object it needs to
>> check for compatibility between the serialized representation and its
>> current version of the class. It does so once, based on the
>> non-transient instance fields in the class. If it finds what it thinks
>> is a mismatch, it will throw an exception.
>>
>> That is, by default, the deserialization is conservative: it will
>> never allow deserialization where an incompatibility exists, but will
>> sometimes disallow a serialization where, actually, there isn't a
>> compatibility problem.
>>
>> That is, maybe you added a new instance field, but, you still think
>> it's valid to deserialize old versions of the object without that
>> field (maybe you are fine with leaving the new instance field as
>> null). To overcome this, you can keep track of what's compatible
>> yourself by setting serialVersionUID. As long as the old/new IDs
>> match, Java will try to proceed with the deserialization.
>>
>> There is a use case for this, but, it's quite rare. If you never use
>> incompatbile versions of your classes in one context, it's not an
>> issue -- and indeed, that's not exactly a great idea. There is no
>> performance implication.
>>
>> The problem is, by setting this field, you're leaving yourself quite
>> open to the opposite problem: adding a change that really isn't
>> backwards-compatible, and forgetting to update the ID. This is a big
>> issue.
>>
>>
>> Hence, I think it is almost never right to set serialVersionUID, and
>> certainly not a default. What do you all think of the logic? I think
>> Benson was on board with it. Now as to how to shut up Eclipse...
>>
>>
>>
>>
>> On Wed, Jun 17, 2009 at 4:20 PM, Jeff Eastman<jd...@windwardsolutions.com>
>> wrote:
>>
>>>
>>> Hi Sean,
>>>
>>> What's the logic for removing these? They now give me a warning in
>>> Eclipse.
>>> I'm willing to turn that off but I'd like to understand why.
>>>
>>> Jeff
>>>
>>>
>>> srowen@apache.org wrote:
>>>
>>>>
>>>> Author: srowen
>>>> Date: Wed Jun 17 10:25:43 2009
>>>> New Revision: 785542
>>>>
>>>> URL: http://svn.apache.org/viewvc?rev=785542&view=rev
>>>> Log:
>>>> Remove serialVersionUID per our standards (right?)
>>>>
>>>> Modified:
>>>>
>>>>
>>>> lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/CardinalityException.java
>>>>
>>>>
>>>> lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/IndexException.java
>>>>
>>>>
>>>> lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/UnboundLabelException.java
>>>>
>>>> Modified:
>>>>
>>>> lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/CardinalityException.java
>>>> URL:
>>>>
>>>> http://svn.apache.org/viewvc/lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/CardinalityException.java?rev=785542&r1=785541&r2=785542&view=diff
>>>>
>>>>
>>>> ==============================================================================
>>>> ---
>>>>
>>>> lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/CardinalityException.java
>>>> (original)
>>>> +++
>>>>
>>>> lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/CardinalityException.java
>>>> Wed Jun 17 10:25:43 2009
>>>> @@ -22,9 +22,4 @@
>>>> */
>>>> public class CardinalityException extends RuntimeException {
>>>> - /**
>>>> - * - */
>>>> - private static final long serialVersionUID = 1L;
>>>> -
>>>> }
>>>>
>>>> Modified:
>>>>
>>>> lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/IndexException.java
>>>> URL:
>>>>
>>>> http://svn.apache.org/viewvc/lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/IndexException.java?rev=785542&r1=785541&r2=785542&view=diff
>>>>
>>>>
>>>> ==============================================================================
>>>> ---
>>>>
>>>> lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/IndexException.java
>>>> (original)
>>>> +++
>>>>
>>>> lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/IndexException.java
>>>> Wed Jun 17 10:25:43 2009
>>>> @@ -22,9 +22,4 @@
>>>> */
>>>> public class IndexException extends RuntimeException {
>>>> - /**
>>>> - * - */
>>>> - private static final long serialVersionUID = 1L;
>>>> -
>>>> }
>>>>
>>>> Modified:
>>>>
>>>> lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/UnboundLabelException.java
>>>> URL:
>>>>
>>>> http://svn.apache.org/viewvc/lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/UnboundLabelException.java?rev=785542&r1=785541&r2=785542&view=diff
>>>>
>>>>
>>>> ==============================================================================
>>>> ---
>>>>
>>>> lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/UnboundLabelException.java
>>>> (original)
>>>> +++
>>>>
>>>> lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/UnboundLabelException.java
>>>> Wed Jun 17 10:25:43 2009
>>>> @@ -22,9 +22,4 @@
>>>> */
>>>> public class UnboundLabelException extends RuntimeException {
>>>> - /**
>>>> - * - */
>>>> - private static final long serialVersionUID = 1L;
>>>> -
>>>> }
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>
>>>
>>
>>
>>
>
>
Re: svn commit: r785542 - in /lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix:
CardinalityException.java IndexException.java UnboundLabelException.java
Posted by Jeff Eastman <jd...@windwardsolutions.com>.
Thanks Sean I do recall that thread now. I've turned off the warning in
Eclipse (Preferences/Java/Compiler/Errors-Warnings). Kind of moot on an
exception with no state anyway.
Jeff
Sean Owen wrote:
> Yeah, Benson noted the same about Eclipse -- we had a thread about
> this a couple weeks ago. I personally think setting serialVersionUID
> is a bad idea, and at least, is exceptional behavior. Unfortunately,
> Eclipse defaults to warning about this.
>
> The logic is this: when the JVM deserializes an object it needs to
> check for compatibility between the serialized representation and its
> current version of the class. It does so once, based on the
> non-transient instance fields in the class. If it finds what it thinks
> is a mismatch, it will throw an exception.
>
> That is, by default, the deserialization is conservative: it will
> never allow deserialization where an incompatibility exists, but will
> sometimes disallow a serialization where, actually, there isn't a
> compatibility problem.
>
> That is, maybe you added a new instance field, but, you still think
> it's valid to deserialize old versions of the object without that
> field (maybe you are fine with leaving the new instance field as
> null). To overcome this, you can keep track of what's compatible
> yourself by setting serialVersionUID. As long as the old/new IDs
> match, Java will try to proceed with the deserialization.
>
> There is a use case for this, but, it's quite rare. If you never use
> incompatbile versions of your classes in one context, it's not an
> issue -- and indeed, that's not exactly a great idea. There is no
> performance implication.
>
> The problem is, by setting this field, you're leaving yourself quite
> open to the opposite problem: adding a change that really isn't
> backwards-compatible, and forgetting to update the ID. This is a big
> issue.
>
>
> Hence, I think it is almost never right to set serialVersionUID, and
> certainly not a default. What do you all think of the logic? I think
> Benson was on board with it. Now as to how to shut up Eclipse...
>
>
>
>
> On Wed, Jun 17, 2009 at 4:20 PM, Jeff Eastman<jd...@windwardsolutions.com> wrote:
>
>> Hi Sean,
>>
>> What's the logic for removing these? They now give me a warning in Eclipse.
>> I'm willing to turn that off but I'd like to understand why.
>>
>> Jeff
>>
>>
>> srowen@apache.org wrote:
>>
>>> Author: srowen
>>> Date: Wed Jun 17 10:25:43 2009
>>> New Revision: 785542
>>>
>>> URL: http://svn.apache.org/viewvc?rev=785542&view=rev
>>> Log:
>>> Remove serialVersionUID per our standards (right?)
>>>
>>> Modified:
>>>
>>> lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/CardinalityException.java
>>>
>>> lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/IndexException.java
>>>
>>> lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/UnboundLabelException.java
>>>
>>> Modified:
>>> lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/CardinalityException.java
>>> URL:
>>> http://svn.apache.org/viewvc/lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/CardinalityException.java?rev=785542&r1=785541&r2=785542&view=diff
>>>
>>> ==============================================================================
>>> ---
>>> lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/CardinalityException.java
>>> (original)
>>> +++
>>> lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/CardinalityException.java
>>> Wed Jun 17 10:25:43 2009
>>> @@ -22,9 +22,4 @@
>>> */
>>> public class CardinalityException extends RuntimeException {
>>> - /**
>>> - * - */
>>> - private static final long serialVersionUID = 1L;
>>> -
>>> }
>>>
>>> Modified:
>>> lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/IndexException.java
>>> URL:
>>> http://svn.apache.org/viewvc/lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/IndexException.java?rev=785542&r1=785541&r2=785542&view=diff
>>>
>>> ==============================================================================
>>> ---
>>> lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/IndexException.java
>>> (original)
>>> +++
>>> lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/IndexException.java
>>> Wed Jun 17 10:25:43 2009
>>> @@ -22,9 +22,4 @@
>>> */
>>> public class IndexException extends RuntimeException {
>>> - /**
>>> - * - */
>>> - private static final long serialVersionUID = 1L;
>>> -
>>> }
>>>
>>> Modified:
>>> lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/UnboundLabelException.java
>>> URL:
>>> http://svn.apache.org/viewvc/lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/UnboundLabelException.java?rev=785542&r1=785541&r2=785542&view=diff
>>>
>>> ==============================================================================
>>> ---
>>> lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/UnboundLabelException.java
>>> (original)
>>> +++
>>> lucene/mahout/trunk/core/src/main/java/org/apache/mahout/matrix/UnboundLabelException.java
>>> Wed Jun 17 10:25:43 2009
>>> @@ -22,9 +22,4 @@
>>> */
>>> public class UnboundLabelException extends RuntimeException {
>>> - /**
>>> - * - */
>>> - private static final long serialVersionUID = 1L;
>>> -
>>> }
>>>
>>>
>>>
>>>
>>>
>>>
>>
>
>
>