Avoiding declaration of variables was actually one of the goals of the
simple-methods, just like it is for most "scripting" languages (as
opposed to compiled languages).
I have no problem with the proposed attribute/etc, but it would be
nice to find a less cumbersome alternative... :)
-David
On Apr 14, 2009, at 3:19 PM, Adrian Crum wrote:
David,
Thanks for the reply!
I don't like the prefixing either, that's why I suggested the new
attribute values. To me, that makes the OFBiz scripting more like
other script engines - you can't use a variable unless it is declared.
One clarification that needs to be made though, in UEL the [] and .
operators don't determine whether the referenced object is a List or
a Map. UEL expressions can use ${object.property} or $
{object[property]} - there are no rules on the object's data type.
The object data type could be a List, a Map, or an arbitrary Object.
The property could be a List index, a Map key, or an Object method
name. In UEL, the object's data type determines how the property is
used.
I'll continue to tinker with the auto-vivify code and see if I can
come up with a better solution. In the meantime, I'd still like to
get comments on the new attribute values.
-Adrian
David E Jones wrote:
Thanks for writing about this Adrian. I played with this a bit
yesterday after seeing your messages and the issue about the income
statement.
The main point that I think is important is that it is a WAY better
approach to use the natural data instead of trying to prefix it, as
that is pretty error prone since ALL code (even outside of simple-
methods for some code) has to be changed.
So anyway, trying to use a UEL syntax that supports this with
things like: theMap["${foo.bar}"] does work well for reading no
matter what "foo.bar" evaluates to (even a number), but like you
said for the auto-vivify o the Map the square-brace ([]) syntax
means a list and so a List is created instead of a Map.
The funny thing about the above is that even if you put a number in
quotes it evaluates to a number and not a String, and that causes
the auto-vivify to create a list instead of a Map for the square
braces ([]). If we could do something about that it would be nice.
Another thing I'm considering is: let's just have the *Entity
Engine* prefix all generated keys so things in the database and
other places have a character first letter. That's easy to do with
a single attribute on the delegator element in the entityengine.xml
file. Along with that we'd have to change the current demo data to
do this as well.
It would still be nice to do something more about the Map auto-
vivify issue as it is much cleaner to treat things like data
instead of like a variable name. What I propose above is just a
work-around, and not the best option IMO... requiring identifiers
to not be a number is a bit of a bad thing...
-David
On Apr 14, 2009, at 12:38 PM, Adrian Crum wrote:
As many of you know, the introduction of the Uniform Expression
Language into the framework caused some backward compatibility
problems - specifically with IDs being used as Map keys. I tried
to accommodate that with a couple of bits of code, but there are
still some issues.
There have been suggestions on the mailing list for workarounds,
but not all of them work, nor do they work all the time. I don't
want to hack up the UEL library further, because its my viewpoint
that OFBiz should conform to existing standards, not twist
existing standards to conform to OFBiz.
The crux of the main problem is the implementation of auto-vivify
that I added to the UEL. Using the following code as an example:
<set field="someVar[someEntity.someId]" value="Hello World!"/>
If someVar was not previously defined, the auto-vivify code tries
to guess what data type someVar is supposed to be. If
someEntity.someId evaluates to an integer, the code assumes it is
an index and creates a List, otherwise it creates a Map. This
works fine in most cases.
The problem comes when someEntity.someId evaluates to an integer
(most IDs are integers) and the developer intended someVar to be a
Map, not a List. There needs to be a way to create and empty Map
in those cases.
I believe the best solution is to modify the <set> element to
allow the creation of an empty Map or List. This will remove the
ambiguity and solve the problem - without further modifying UEL.
I modified my local copy to allow the creation of an empty Map:
<set field="someVar" type="NewMap"/>
<set field="someVar[someEntity.someId]" value="Hello World!"/>
and it works great.
So, the bottom line is, I would like to add "NewMap" and "NewList"
to the <set> element type attribute. What do you think?
-Adrian