Adam,
--- Adam J Chesney <[EMAIL PROTECTED]> wrote:
> Dimitri,
>
<snip>
> >
> > > I think it might be a show stopper for me. I am almost certainly
> trying
> to
> > > use JXPath in a way that was not intended. I am trying to build a
> 2 way
> > data
> > > binding mapping structure which I think JXPath is excellent for,
> but I
> > would
> > > also like to add some easy micro-scripting, which is maybe not
> possible
> > with
> > > JXPath.
> > Right, I think you might be stretching JXPath's capabilities. The
> > getValue() method is supposed to be a query, that is a request
> without
> side
> > effects.
>
> Thought so. However, it would make it a very cool tool indeed if it
> could be
> used like that, no? It seems that JXPath is very close to allowing me
> to do
> this.
>
> >
> > > One other thing, can you explain to me why the returned
> collections
> should
> > > be immutable? Is it necessary to enforce that behaviour?
> > I make the collection immutable simply to get it to throw an
> exception. If
> > the collection is mutable, operations changing it are successful,
> but
> since
> > the changes are made to some temporary collection, not the original
> one
> you
> > want to change.
>
> I understand that if you are performing a query or a slice or
> something
> (within an XPath) that you must have to create a temporary collection
> to
> hold the result, however I still don't see why, if you are
> referencing a
> whole collection (i.e. $myList), a copy is made before operating on
> it. I
> still don't see the fundemental difference between this:
>
> getValue("map:put($myMap, 'key', 'value')") returns: null
>
> which calls 'put' on my original Map:
>
> and this:
>
> getValue("list:add($myList, 'hello')") returns: true
>
> which calls 'add' on a temporary copy of my original list,
>
> Why do we create a temporary structure just because the target of
> the
> ExtensionFunction is a Collection? Is it just because of the
> implementation
> details, or is it by design?
>
> Sorry for being a pain! :)
No, no, this is a legitimate complaint.
Collections have a special status in XPath. XPath works with objects
of four kinds: string, boolean, number and node set (aka collection).
This why they are treated differently.
Like I said earlier, we could make an exception for a variable, but
other cases are more complicated. Consider for example this:
"add(//foo, 'hello')"
Let say jxpath finds a bean that has a property called "foo" and it
happens to be a collection. Intuitively, it should then call the "add"
method on that collection. But unfortunately it cannot, because there
might be another bean somewhere else in three that also has a property
called "foo". It needs to combine all values from the first list with
all values from the second list and continue search. Only when the
whole tree has been examined, can it call "add" on the result.
> BTW - I have hacked a patch that allows XPaths like this:
>
> context.createPathAndSetValue("$myList[-1]", "'hello'");
>
> which now adds to the List. I had to do this because when I tried
> this:
>
> context.createPathAndSetValue("$myList[last()+1]", "'hello'");
>
> I got an exception:
>
> org.apache.commons.jxpath.JXPathException: Cannot set value for
> xpath:
> $myList[last()+1]
>
> at
>
org.apache.commons.jxpath.ri.JXPathContextReferenceImpl.setValue(JXPathConte
> xtReferenceImpl.java:412)
>
> I would have expected that my Factory would have been invoked in
> order to
> expand the List, but it never got called.
Thanks for bringing this up. I should make the exception message more
concrete. The message should say: "createPathAndSetValue can only be
called with a 'simple' path, that is a path using only the child and
attribute axes and no context-dependent predicates"
Since last() is a context dependent predicate, the invocation does not
work.
Have you considered using Jexl instead of JXPath? JXPath is intended
primarily for mixed models containing both beans and XML. After all
XPath itself is only defined for XML. Jexl does not have to deal with
the burden of non-java bean models like DOM and JDOM, so its handling
of beans is more intuitive.
>
> Cheers,
>
> Adam.
Regards,
- Dmitri
>
>
>
>
>
>
>
> >
> > > Thanks for your help,
> >
> > >
> > > Adam.
> >
> > I hope this does not make jxpath completely unusable for you.
> >
> > - Dmitri
> > >
> > >
> > >
> > >
> > > ----- Original Message -----
> > > From: "Dmitri Plotnikov" <[EMAIL PROTECTED]>
> > > To: "Jakarta Commons Developers List"
> <[EMAIL PROTECTED]>
> > > Sent: Thursday, February 06, 2003 9:47 PM
> > > Subject: Re: [JXPath] Problem adding objects to an ArrayList
> Variable
> > >
> > >
> > > > Adam,
> > > >
> > > > JXPath is in fact incapable of doing what you want it to do.
> Before a
> > > > collection is passed to an extension function, it undergoes
> some
> > > > conversions, which make it unmodifiable. JXPath should of
> course
> throw
> > > > an exception in this case. I have added that exception and
> will check
> > > > that new code in tonight.
> > > >
> > > > The rationale for this behavior is that from JXPath's
> perspective an
> > > > extension function cannot take a collection by reference.
> Arguments of
> > > > an extension function are either primitive values or "node
> sets",
> which
> > > > are really results of search. JXPath finds all nodes matching
> the
> > > > path, loads them into a list and then passes the list to the
> function.
> > > > Consider for example this:
> > > >
> > > > context.getValue("size(//foo[bar = 3])")
> > > >
> > > > From this example it should be clear that the argument of the
> function
> > > > is in fact an unmodifiable collection - results of search.
> > > >
> > > > We could of course do something special for variables, but then
> the
> > > > extension function's code would have to be aware of the
> difference
> > > > between variables and general node sets. I think we should
> avoid this
> > > > implicit difference.
> > > >
> > > > What you could do to add a node to the collection is call
> > > > createPathAndSetValue() with a path like "$myList[24]".
> > > >
> > > > Thanks for bringing this issue up - I should document it.
> > > >
> > > > - Dmitri
> > > >
>
=== message truncated ===
__________________________________________________
Do you Yahoo!?
Yahoo! Mail Plus - Powerful. Affordable. Sign up now.
http://mailplus.yahoo.com
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
