beforeInsert event problem

[Marc Palmer Local]

Hi All,

I tried to use beforeInsert on a domain class for the first time  
today, following an example like the one in the docs:

class Person {
    Date dateCreated
    def beforeInsert = {
        dateCreated = new Date()
    }
}

... my assignment is different of course but the same principle. All  
fine except... this is called after validate() so this is not useful  
at all if you have any "required" constraints on the properties.

You could work around it by assigning some non-null value at init time  
to the property so it survives validation.

However this implies that validation is either being performed twice -  
before the calls to events and again after, or only once before  
calling events, in which case events can bypass constraints and hence  
introduce bugs.

Anybody got any views on this?

I would have thought that the events have to occur before validation,  
always.

Marc


---------------------------------------------------------------------
To unsubscribe from this list, please visit:

    http://xircles.codehaus.org/manage_email

[Marc Palmer Local]

On 24 Apr 2008, at 13:26, Marc Palmer wrote:

> Hi All,
>
> I tried to use beforeInsert on a domain class for the first time  
> today, following an example like the one in the docs:
>
> class Person {
>   Date dateCreated
>   def beforeInsert = {
>       dateCreated = new Date()
>   }
> }

Actually this gets worse. I am either cursed, stupid, or doing  
something wrong.

I have found:

1) validation is performed before events
2) changes to properties made in the event have no effect

Here's my domain class:

import java.rmi.server.UID
import java.security.*
import java.sql.Timestamp

class PendingEmailConfirmation {
        String confirmationToken
        String userToken
        Date timestamp = new Date()

     // Init the unique token
     def beforeInsert = {
         println "invoking beforeInsert - conf token is  
[$confirmationToken]"
         makeToken()
         println "after beforeInsert - conf token is  
[$confirmationToken]"
     }

     void makeToken()
     {
         def uid = new UID().toString()
         confirmationToken = uid.encodeAsSHA1Hex()
     }

        static constraints = {
                confirmationToken(size:1..100)
                userToken(size:1..100)
        }
}

This code above in a unit test that instantiates and then saves the  
object fails to save because "confirmationToken" is null because  
validation has kicked in, the event never happens and my printlns do  
not occur.

If I change the code so that confirmationToken is init'd with "?":


class PendingEmailConfirmation {
        String confirmationToken = "?"
...

Then the output from the test is:

invoking beforeInsert - conf token is [?]
after beforeInsert - conf token is [?]

So tell me... how is this possible? I think I need to raise a jira or  
two don't I?

My gut tells me I'm missing a stupid typo. I hope so.

Marc




---------------------------------------------------------------------
To unsubscribe from this list, please visit:

    http://xircles.codehaus.org/manage_email

[Peter Ledbrook-2]

>     // Init the unique token
>     def beforeInsert = {
>         println "invoking beforeInsert - conf token is [$confirmationToken]"
>         makeToken()
>         println "after beforeInsert - conf token is [$confirmationToken]"
>     }

What happens if you replace the "makeToken()" line with
"confirmationToken = 'something'"?

>  Then the output from the test is:
>
>  invoking beforeInsert - conf token is [?]
>  after beforeInsert - conf token is [?]

But what's in the database?

>  So tell me... how is this possible? I think I need to raise a jira or two
> don't I?
>
>  My gut tells me I'm missing a stupid typo. I hope so.

It doesn't look like it, but I wonder what's happening when the
closure offloads to the method. If the closure's delegate is
overridden and the resolve strategy changed, strange things might
happen. But that shouldn't be happening.

Cheers,

Peter

--
Software Engineer
G2One, Inc.
http://www.g2one.com/

---------------------------------------------------------------------
To unsubscribe from this list, please visit:

    http://xircles.codehaus.org/manage_email

[Marc Palmer Local]

On 24 Apr 2008, at 14:38, Peter Ledbrook wrote:

>>    // Init the unique token
>>    def beforeInsert = {
>>        println "invoking beforeInsert - conf token is  
>> [$confirmationToken]"
>>        makeToken()
>>        println "after beforeInsert - conf token is  
>> [$confirmationToken]"
>>    }
>
> What happens if you replace the "makeToken()" line with
> "confirmationToken = 'something'"?
>
>> Then the output from the test is:
>>
>> invoking beforeInsert - conf token is [?]
>> after beforeInsert - conf token is [?]
>
> But what's in the database?

Not sure, running in memory and haven't had time to explore. However  
the property was definitely being left as ? because the data was then  
used to send an email and the token was ? ... which started the whole  
sorry tale of how beforeInsert is b0rked in 1.0.2 :)

I had to give up completely on using an event, it just isn't working.  
Code has to go live today...

Marc

---------------------------------------------------------------------
To unsubscribe from this list, please visit:

    http://xircles.codehaus.org/manage_email