Re: Escaping in strings according to WOML
Re: Escaping in strings according to WOML
- Subject: Re: Escaping in strings according to WOML
- From: "Mr. Pierre Frisch" <email@hidden>
- Date: Tue, 30 Jun 2009 10:25:07 -0700
Mike,
Have you done any performance testing on this? I would have question
about invoking OGNL for each binding. I am not saying this is bad just
that the core WO is made to be very efficient and that pose some
constrains on what we can do. I know that RoR has some very nice
feature but it is also very slow. We have to find a happy medium, the
Association factories are quite efficient and the overhead is very
minimal but there is a bit more code to write, although we could
probably template a lot of it.
As for the comments in bindings I agree. I have not had time to look
into it, would you like to?
Pierre
--
Pierre Frisch
email@hidden
On Jun 30, 2009, at 10:13, Mike Schrag wrote:
This requires an association factory registered per function,
though, right? Part of the niceness of helpers is how they are
registered and resolved ...
For instance, I can just create PersonHelper, and somePerson|
displayName will automatically look for PersonHelper and call
displayName on it, just based on the type of the bound object. It
will also walk the inheritance tree, so you can have
EOGenericRecordHelper and it will look up the inheritance tree to
find the class or interface helper that defines the function you're
referencing (so you can do "$anyEO|id" and it will call
EOGenericRecordHelper.id(eo) ). Also, because it's implemented
behind the scenes as an OGNL expression, you can pass arbitrary
params "for free" -- so I can do "$paragraph|highlight(word)" and it
will call StringHelper.highlight(String paragraph, String word). On
top of that, if your binding evaluates to null (say "somePerson" was
a null in the first example) it actually computes the Class name of
the bindings (we have an NSKVC impl that operates purely on Classes,
not on the objects themselves) and figures out what type the object
WOULD have been (obviously limited by what you declare it to be) and
calls the appropriate helper passing a null, so you can do helpers
with null values. So in the first example somePerson was null, it
would figure out that somePerson was a Person class from the type
declaration of the binding (this works on keypaths too) and would
dispatch to PersonHelper.displayName(null) so you can display "No
name available" (or whatever).
My impression, though, is that it's quite a bit more
programmatically expensive to register functions in the 5.4 parser
and it wouldn't have the fancy scoping or resolution. I think maybe
you'd have to do something like "helper: person|displayName", maybe,
and parse it with a fancier association factory?
In the Wonder one, the way we do it is that as it parses, it
actually turns into an OGNLAssociation behind the scenes, and we
just jam the parameters onto the end of the registry lookup. For
example, that "$paragraph|highlight(word)" becomes:
~
@ognl
.helperfunction
.WOHelperFunctionRegistry
@registry()._helperInstanceForFrameworkNamed(#this, "paragraph",
"highlight", "app").highlight(paragraph, word)
it's sort of magical, but actually works :) gotta love ognl ...
Oh, the other thing that would be nice in the 5.4 parser is support
for "// VALID" There are cases where the WOLips validator can't
provably validate an expression (not often, but happens). The
Wonder parser has an extension to WOD and inline bindings that
allows you to do:
Test : WOString {
value = someKVCExpressionThatCantBeStaticallyResolved; // VALID
}
and
<wo:str value="$someKVCExpressionThatCantBeStaticallyResolved //
VALID"/>
so the WOLips validator can hide warnings on them.
ms
On Jun 30, 2009, at 12:36 PM, Mr. Pierre Frisch wrote:
Mike there is only one space difference:
<wo:string value="[person|displayName]" />
or
<wo:string value="[displayName: person]" />
Cheers
Pierre
--
Pierre Frisch
email@hidden
On Jun 29, 2009, at 15:46, Mike Schrag wrote:
I suppose we just need to integrate them as association and add
them to the association factory. This is trivial to do.
yeah, i'm pretty sure 5.4's parser can do it with a custom
factory, but I'm concerned about how much typing it will require
to actually use it ... helpers are cool because it's so easy to
drop into a binding (change from "$person" to "$person|
displayName"). if you have to switch to a different association
namespace to get it on a binding, i think it might suck more, but
I'll withhold judgment.
ms
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Webobjects-dev mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden
Attachment:
smime.p7s
Description: S/MIME cryptographic signature
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Webobjects-dev mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden