[jira] Commented: (OPENJPA-927) Fix definition of javax.persistence.query.timeout property

["Dianne Richards (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/OPENJPA-927?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12674680#action_12674680 ] 

Dianne Richards commented on OPENJPA-927:
-----------------------------------------

Thanks for checking this out Pinaki.

>1. loadKey is the key with which a property is loaded from configuration artifacts. At this point of execution, no property has been >*actually* loaded, they are merely being declared to exist. Hence we should not be setting load key.

I now understand this and will change it. But, we have to solve a problem if we don't do this. The getProperties() methods return all defined properties, even if they have not been explicitly set. This calls the Configuration .toProperties() method which calls the setValue() method. By default, if we don't do anything, the value javax.persistence.query.timeout will be prefixed with "openjpa.", which is not good. We can get around this doing a setEquivalentKey() instead. But, now I'm wondering what we'll get with the getSupportedProperties() with this change. I'll have to investigat that. But, if that's ok, is this a reasonable alternative? Or, should I figure out something else?

The other option is to define a new openjpa property, such as openjpa.QueryTimeout. I was trying to avoid that option. I'm not sure we want to define a new openjpa property for every new spec property.

>2. configuration declares a Value. But does not assign its value. So setting its value to -1 does not look alright. Setting default >value is OK.

ok - But, I modeled this after the openjpa.LockTimeout definition. So, does this one need to be corrected too?


> Fix definition of javax.persistence.query.timeout property
> ----------------------------------------------------------
>
>                 Key: OPENJPA-927
>                 URL: https://issues.apache.org/jira/browse/OPENJPA-927
>             Project: OpenJPA
>          Issue Type: Bug
>          Components: kernel
>    Affects Versions: 2.0.0
>            Reporter: Dianne Richards
>            Assignee: Dianne Richards
>            Priority: Minor
>             Fix For: 2.0.0
>
>
> This was originally reported by Pinaki in OPENJPA-849. It is being moved to this new JIRA. Here's Pinaki's original comment:
>  
>         queryTimeout.setLoadKey("javax.persistence.query.timeout");
>         queryTimeout.setDefault("-1");
>         queryTimeout.set(-1);
>         queryTimeout.setDynamic(true);
> does not seem kosher for the following reason:
> 1. loadKey is the key with which a property is loaded from configuration artifacts. At this point of execution, no property has been *actually* loaded, they are merely being declared to exist. Hence we should not be setting load key.
> 2. configuration declares a Value. But does not assign its value. So setting its value to -1 does not look alright. Setting default value is OK.
> These issues gain significance in the light of the fact the configuration's hashcode is the key to a factory in JNDI. And computation of hashcode depends on the actual value of the Values.
> As an extreme example, assume two Configuration C1 and C2 nearly identical but differs *only* in their query.timeout value. The requirement is hash code for C1 and C2 must not be equal. And that is what Configuration.hashCode() ensures. But, because we are setting query timeout to -1 (that is not what the user's p.xml sets) and it is marked as dynamic, in both cases Configuration hashcode will treat query.timeout value to be -1 and will end up computing same hashcode for C1 and C2.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.