Eric Dobbs wrote:

> On Thursday, April 25, 2002, at 05:27  PM, Sam Joseph wrote:
>
>> Please find patch attached.  Sorry about the corruption - my mailer 
>> seems to be weirding out.
>
> Sam, I've attached a revised patch.
>
> things I changed:
>
> * removed the new methods that parse strings into
>   Criterions (prefer to keep this patch confined to
>   fixing the precedence) 

Fair enough, they were just there for you to see them

> * renamed types to conjunctions, including the getter. 

ok

> * changed clauses and conjunctions from Vectors to
>   Lists with ArrayList implementations 

works for me

> * made getClauses and getConjunctions private

ditto

> * removed the previous and and or fields and their
>   getters (they had previously been commented out)

right

> * split a couple more long lines

always helpful

> * removed x_ from a few variable names. 

Yes, I got a mail from Jon Scott Stevens about that.  He helpfully 
pointed me at the coding conventions for turbine, which I hadn't seen 
yet.  Will follow in future.

you may want to remove o_ and p_ from some variable names as well, 
although I think they're swish :-)

> I stopped short of committing this patch for the
> following reasons:
>
> I haven't analyzed the recursion in the appendTo
> methods to satisfy myself that they work correctly. And
> I also haven't analyzed the equals() and hashcode()
> methods either.  They all look reasonable enough, but I
> just want to be careful.

Sure.  I have some test code for that.  I extended the CriteriaTest 
object in the src/test/org/apaceh/torque/util/ directory
Would you like a copy?  Unfortunately it makes use of those parsing 
methods you removed.

> More importantly...
>
> Removing the getAnd and getOr methods is a change to
> the public API.  They were public but disfunctional (in
> that they didn't allow a coherent precedence to be
> set).  They probably should have been private.  It
> seems fairly unlikely that there would be much code out
> there using those methods. 

Sure

> Any objections to just removing them?
> Or do you prefer that they be deprecated? 

Deprecation would be safer no?  I mean they could even be set up to 
return the first ANDed or ORed clause if one was present?

> I know that Torque is in use in a few projects even
> though it hasn't been released.  Because it hasn't been
> released I think we can remove these methods without
> violating the letter of the deprecation policy.  But it
> is a violation of the spirit.  Because these two methods
> would be mostly useless outside of Criteria, I would
> not object to bending the spirit of the deprecation
> policy in these specific cases.  What say you?

I say fine, but I gues you're polling everyone rather than just me.

CHEERS> SAM


--
To unsubscribe, e-mail:   <mailto:[EMAIL PROTECTED]>
For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>

Reply via email to