-----Original
Message-----
From:
[EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED]On Behalf Of Drew McAuliffe
Sent: Friday, November 21, 2003
8:05 PM
To:
[EMAIL PROTECTED]
Subject: RE: [OS-webwork] OGNL
madness - evaluated tag attributes
I kind of agree. I think
triple-quoting simple parameters like "name" is kind of silly. In
fact, I've never really liked the triple-quoting much at all (it seems unwieldy
to me), though I understand the need for it. To me, it might be easier to just
ditch the whole triple-quote thing and do the evaluation velocity style, with a
${} combination indicating something that should be evaluated (e.g.,
name="someName" value="${evaluateThis}"). Though of course
I'd expect that to be controversial....
From:
[EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]
On Behalf Of John Patterson
Sent: Friday, November 21, 2003
6:38 AM
To: Webwork
Subject: [OS-webwork] OGNL madness
- evaluated tag attributes
I have just run my app with the
release candidate and am in the process of changing many (but not all) of the
tag attributes to use single quotes for literals. I am thinking that in
many instances, evaluating the attribute tag is not really very helpful.
For example, the Bean tag has an attribute 'name' for the name of the class to
instantiate. This is now treated as an OGNL _expression_ and evaluated to
derive the name of the class to use. OGNL madness!
In 99% of cases it seems that this
attribute does not need to be evaluated. Forcing the developer to
"triple quote" values most of the time will lead to annoying, hard to
spot, errors (and already has for me). What may have been a step forward
for compatibility has IMHO been a large step backwards in usability.
So what I am suggesting is that we
only evaluate attributes where that would be the expected common usage.
We would need to use a standard way of naming literal expressions
that are NOT treated as OGNL expressions. For example, attributes
called 'name' could be treated as literals but passed through the TextUtil
thing (so you can use name="${first}-${second}").
Attributes called "value"
on the other hand are often variant and therefore would make sense to evaluate
them. You would then have to "triple quote" them to use a
literal. For those who think that this would break consistency, consider
that not all attributes are currently evaluated (eg Iterator status, Set
name). So in fact, introducing such a naming standard would increase
consistency (maybe at the expense of backward compatibility)
