Thread: [Springframework-developer] Creating Application Context getBean 'closures'

Brought to you by: colins, jhoeller, johnsonr, robharrop, trevormarshall

springframework-developer

[Springframework-developer] Creating Application Context getBean 'closures'

From: Colin S. <col...@ex...> - 2003-11-18 17:08:56

I think I may be missing something, but I think it's desireable to be 
able to create in an easy fashion, 'closures' which encapsulate getting 
a bean from a bean factory.

Say you have an interface
  interface MyInteface {   ... whatever }

and you have a factory Interface
  interface MyInterfaceFactory {
    MyInterface getInstance();
  }

And you have a user of the factory, who you'd rather have no knowledge 
of Spring, which is why he's using the factory instead of calling 
getBean himself:
class User {
  MyInterfaceFactory _myfac;
  public void setMyInterfaceFactory(MyInterfaceFactory myfac) { _myfac = 
myfac; }
  public void someMethod() {
    // need a new instance of MyInterface to work with
    MyInterface myint = _myfac.getInstance();
      ...
  }
}

Now I have a bean factory
<beans>
  <bean id="myinterface" singleton="false" 
class="com.whatever.MyInterfaceImpl">
  </bean>
  <bean id="user" class="com.whatever.User">
    <property name="myInterfaceFactory"><ref bean="xxxxxxxx"/></property>
  </bean>
</beans>

now, as per the above, context.getBean("myinterface") is already a 
factory for objects implementing MyInterface. But I don't want the User 
object to know anything about contexts. And I'd rather not create an 
actual object that implements MyInterfaceFactory. It seems like a waste, 
since all I am doing here is trying to create a level of indirection, 
and I already have a factory inside the context itself, and I may want 
to use this approach in 30 different places, just to add a level of 
indirection in creating new objects.

So what I think is needed is some variation of ProxyFactoryBean (but a 
separate class), which given a target bean (which is itself a factory), 
and a factory interface having a method with no args which returns a 
certain type, creates on the fly a new class implementing the factory 
interface, which will just use the target factory bean to actually 
supply the instance. So the bean def above would become:
<beans>
  <bean id="myinterface" singleton="false" 
class="com.whatever.MyInterfaceImpl">
  </bean>
  <bean id="myinterface-factory" class="org.springframework.whatever.XXXX">
    <property name="targetBean"><ref bean="xxxxxxxx"/></property>
    <property 
name="interface"><value>x.y.z.AFactoryInterface</value></property>
  </bean>
  <bean id="user" class="com.whatever.User">
    <property name="myInterfaceFactory"><ref 
bean="myinterface-factory"/></property>
  </bean>
</beans>

Am I missing an existing way to do this? Is this worth adding to spring 
as a convenience built-in, along the lines of TransactionProxyFactoryBean?

[Springframework-developer] Re: Creating Application Context getBean 'closures'

From: Chris N. <ch...@si...> - 2003-11-18 17:40:48

Colin Sampaleanu wrote:

> I think I may be missing something, but I think it's desireable to be
> able to create in an easy fashion, 'closures' which encapsulate getting
> a bean from a bean factory.
> 
> Say you have an interface
>   interface MyInteface {   ... whatever }
> 
> and you have a factory Interface
>   interface MyInterfaceFactory {
>     MyInterface getInstance();
>   }

You may want to look at:
http://cglib.sourceforge.net/apidocs/net/sf/cglib/reflect/MethodDelegate.html

It can create an object that implements an interface but delegates the
implementation to a virtual or static method (with matching signature) of
your choice. Essentially it is a implementation of C# delegates for Java.

Chris

Re: [Springframework-developer] Re: Creating Application Context getBean 'closures'

From: Colin S. <col...@ex...> - 2003-11-18 18:11:04

Chris Nokleberg wrote:

>Colin Sampaleanu wrote:
>  
>
>>I think I may be missing something, but I think it's desireable to be
>>able to create in an easy fashion, 'closures' which encapsulate getting
>>a bean from a bean factory.
>>
>>Say you have an interface
>>  interface MyInteface {   ... whatever }
>>
>>and you have a factory Interface
>>  interface MyInterfaceFactory {
>>    MyInterface getInstance();
>>  }
>>    
>>
>
>You may want to look at:
>http://cglib.sourceforge.net/apidocs/net/sf/cglib/reflect/MethodDelegate.html
>
>It can create an object that implements an interface but delegates the
>implementation to a virtual or static method (with matching signature) of
>your choice. Essentially it is a implementation of C# delegates for Java.
>  
>
Thanks Chris. I don't think though this provides any fundamental 
advantage over just using AdvisedSupport in Spring as used by 
ProxyFactoryBean. The fact that it only delegates to a method with a 
matching signature is a problem too (unless I'm missing something). The 
code would have to delegate to a method with an Object return value, 
with the proxy doing something like a ClassCastException if the return 
value (at runtime) didn't match.

[Springframework-developer] Re: Creating Application Context getBean 'closures'

From: roger h. <apo...@sn...> - 2003-11-21 01:50:01

Hi Colin

Your thoughts on implementation hiding have got me thinking...

But before I test anything could you clarify the sort of usage you
have in mind - specifically, would it make a difference whether the
implementation hiding factories were themselves 'typed' or 'generic' ie. do
you want / need the type of the factory to vary with the type that it
generates, or would a single generic factory be adequate ?

In other words, does one or other of the following make a difference ?

1.  Typed interface factories:

interface MyFirstInterfaceFactory {
    MyFirstInterface getInstance()
}

interface MySecondInterfaceFactory {
    MySecondInterface getInstance()
}


2.  Generic interface factory:

interface GenericFactory {
    Object getInstance()
    Class getClass()
}

where different instances of GenericFactory will return from getClass() the
relevant type, MyFirstInterface, MySecondInterface etc..

I'll try & test something tomorrow

Roger



"Colin Sampaleanu" <col...@ex...> wrote in message
news:3FB...@ex......
> I think I may be missing something, but I think it's desireable to be
> able to create in an easy fashion, 'closures' which encapsulate getting
> a bean from a bean factory.
>
> Say you have an interface
>   interface MyInteface {   ... whatever }
>
> and you have a factory Interface
>   interface MyInterfaceFactory {
>     MyInterface getInstance();
>   }
>
> And you have a user of the factory, who you'd rather have no knowledge
> of Spring, which is why he's using the factory instead of calling
> getBean himself:
> class User {
>   MyInterfaceFactory _myfac;
>   public void setMyInterfaceFactory(MyInterfaceFactory myfac) { _myfac =
> myfac; }
>   public void someMethod() {
>     // need a new instance of MyInterface to work with
>     MyInterface myint = _myfac.getInstance();
>       ...
>   }
> }
>
> Now I have a bean factory
> <beans>
>   <bean id="myinterface" singleton="false"
> class="com.whatever.MyInterfaceImpl">
>   </bean>
>   <bean id="user" class="com.whatever.User">
>     <property name="myInterfaceFactory"><ref bean="xxxxxxxx"/></property>
>   </bean>
> </beans>
>
> now, as per the above, context.getBean("myinterface") is already a
> factory for objects implementing MyInterface. But I don't want the User
> object to know anything about contexts. And I'd rather not create an
> actual object that implements MyInterfaceFactory. It seems like a waste,
> since all I am doing here is trying to create a level of indirection,
> and I already have a factory inside the context itself, and I may want
> to use this approach in 30 different places, just to add a level of
> indirection in creating new objects.
>
> So what I think is needed is some variation of ProxyFactoryBean (but a
> separate class), which given a target bean (which is itself a factory),
> and a factory interface having a method with no args which returns a
> certain type, creates on the fly a new class implementing the factory
> interface, which will just use the target factory bean to actually
> supply the instance. So the bean def above would become:
> <beans>
>   <bean id="myinterface" singleton="false"
> class="com.whatever.MyInterfaceImpl">
>   </bean>
>   <bean id="myinterface-factory"
class="org.springframework.whatever.XXXX">
>     <property name="targetBean"><ref bean="xxxxxxxx"/></property>
>     <property
> name="interface"><value>x.y.z.AFactoryInterface</value></property>
>   </bean>
>   <bean id="user" class="com.whatever.User">
>     <property name="myInterfaceFactory"><ref
> bean="myinterface-factory"/></property>
>   </bean>
> </beans>
>
> Am I missing an existing way to do this? Is this worth adding to spring
> as a convenience built-in, along the lines of TransactionProxyFactoryBean?
>
>
>
>
> -------------------------------------------------------
> This SF. Net email is sponsored by: GoToMyPC
> GoToMyPC is the fast, easy and secure way to access your computer from
> any Web browser or wireless device. Click here to Try it Free!
> https://www.gotomypc.com/tr/OSDN/AW/Q4_2003/t/g22lp?Target=mm/g22lp.tmpl

Re: [Springframework-developer] Re: Creating Application Context getBean 'closures'

From: Colin S. <col...@ex...> - 2003-11-21 02:51:16

I was thinking of the first, with the idea that there is no Spring class 
dependency, and you achieve total inversion of control.

Your second option would work of course. It does achieve the IOC aspect, 
but doesn't remove the Spring  (if the GenericFactory interface comes 
from Spring), and you would have to cast in the using bean instead of in 
the proxy. The IOC and hiding of other beans in the BF are probably the 
most important things anyway.

(Of course, both of these do have a negative aspect, compared to just 
using BeanFactory directly; with BeanFActory the using bean can declare 
itself BeanFactoryAware and get the reference with no extra configuration).


roger holbrook wrote:

>Hi Colin
>
>Your thoughts on implementation hiding have got me thinking...
>
>But before I test anything could you clarify the sort of usage you
>have in mind - specifically, would it make a difference whether the
>implementation hiding factories were themselves 'typed' or 'generic' ie. do
>you want / need the type of the factory to vary with the type that it
>generates, or would a single generic factory be adequate ?
>
>In other words, does one or other of the following make a difference ?
>
>1.  Typed interface factories:
>
>interface MyFirstInterfaceFactory {
>    MyFirstInterface getInstance()
>}
>
>interface MySecondInterfaceFactory {
>    MySecondInterface getInstance()
>}
>
>
>2.  Generic interface factory:
>
>interface GenericFactory {
>    Object getInstance()
>    Class getClass()
>}
>
>where different instances of GenericFactory will return from getClass() the
>relevant type, MyFirstInterface, MySecondInterface etc..
>
>I'll try & test something tomorrow
>
>Roger
>
>
>
>"Colin Sampaleanu" <col...@ex...> wrote in message
>news:3FB...@ex......
>  
>
>>I think I may be missing something, but I think it's desireable to be
>>able to create in an easy fashion, 'closures' which encapsulate getting
>>a bean from a bean factory.
>>
>>Say you have an interface
>>  interface MyInteface {   ... whatever }
>>
>>and you have a factory Interface
>>  interface MyInterfaceFactory {
>>    MyInterface getInstance();
>>  }
>>
>>And you have a user of the factory, who you'd rather have no knowledge
>>of Spring, which is why he's using the factory instead of calling
>>getBean himself:
>>class User {
>>  MyInterfaceFactory _myfac;
>>  public void setMyInterfaceFactory(MyInterfaceFactory myfac) { _myfac =
>>myfac; }
>>  public void someMethod() {
>>    // need a new instance of MyInterface to work with
>>    MyInterface myint = _myfac.getInstance();
>>      ...
>>  }
>>}
>>
>>Now I have a bean factory
>><beans>
>>  <bean id="myinterface" singleton="false"
>>class="com.whatever.MyInterfaceImpl">
>>  </bean>
>>  <bean id="user" class="com.whatever.User">
>>    <property name="myInterfaceFactory"><ref bean="xxxxxxxx"/></property>
>>  </bean>
>></beans>
>>
>>now, as per the above, context.getBean("myinterface") is already a
>>factory for objects implementing MyInterface. But I don't want the User
>>object to know anything about contexts. And I'd rather not create an
>>actual object that implements MyInterfaceFactory. It seems like a waste,
>>since all I am doing here is trying to create a level of indirection,
>>and I already have a factory inside the context itself, and I may want
>>to use this approach in 30 different places, just to add a level of
>>indirection in creating new objects.
>>
>>So what I think is needed is some variation of ProxyFactoryBean (but a
>>separate class), which given a target bean (which is itself a factory),
>>and a factory interface having a method with no args which returns a
>>certain type, creates on the fly a new class implementing the factory
>>interface, which will just use the target factory bean to actually
>>supply the instance. So the bean def above would become:
>><beans>
>>  <bean id="myinterface" singleton="false"
>>class="com.whatever.MyInterfaceImpl">
>>  </bean>
>>  <bean id="myinterface-factory"
>>    
>>
>class="org.springframework.whatever.XXXX">
>  
>
>>    <property name="targetBean"><ref bean="xxxxxxxx"/></property>
>>    <property
>>name="interface"><value>x.y.z.AFactoryInterface</value></property>
>>  </bean>
>>  <bean id="user" class="com.whatever.User">
>>    <property name="myInterfaceFactory"><ref
>>bean="myinterface-factory"/></property>
>>  </bean>
>></beans>
>>
>>Am I missing an existing way to do this? Is this worth adding to spring
>>as a convenience built-in, along the lines of TransactionProxyFactoryBean?
>>
>>    
>>

[Springframework-developer] Re: Re: Creating Application Context getBean 'closures'

From: roger h. <apo...@sn...> - 2003-12-08 15:15:09

Colin,

A long delayed response, but just in case this topic may still be of
interest to you...

I played around with a couple of ways of doing this, & have attached a
straight forward implementation of the initial scheme that you first
outlined, ie: given the following:

public interface ITestBean_Factory1 {
    ITestBean getInstance();
}

<bean id="testBean"
  class="org.springframework.beans.TestBean">
 <property ...  >
</bean>

<bean id="testTypedFactoryBean1"
  class="org.springframework.beans.factory.support.TypedFactoryBean">
  <property name="interface">
    <value>
      org.springframework.beans.factory.support.ITestBean_Factory1
    </value>
  </property>
 <property name="target">
   <value>
     testBean
   </value>
  </property>
</bean>

then a call to beanFactory.getBean("testTypedFactoryBean1") will return
dynamic proxy instance that implements the given typed factory interface.


FYI, the following is a short description of the 'other' approach I tried:

My first thought was that your initial scheme could be simplified by using
a factory-interface attribute on the target bean definition, & doing away
with the need to have a separate typed factory bean, ie:

<bean id="testBean"
  class="org.springframework.beans.TestBean">
factory-interface"org.springframework.beans.factory.support.ITestBean_Factor
y1"
  <property... >
</bean>

With a few tweaks in core classes, this all works fine - the only problem is
that it just doesn't feel clean enough, ie: it requires the following things
to change:

- new factory-interface bean definition attribute
- new behaviour for beanFactory.getBean("&beanName"), since the typed
factory dynamic proxy instance returned, will no longer reliably be an
instance of FactoryBean
- small, but none the less, additional complexity to AbstractBeanFactory,
one more level of indirection to contend with

On reflection, in this instance, end of idea ;)

I would be interested to know how you ended up actually dealing with the
issue in practice ?

Roger



"Colin Sampaleanu" <col...@ex...> wrote in message
news:3FB...@ex......
> I was thinking of the first, with the idea that there is no Spring class
> dependency, and you achieve total inversion of control.
>
> Your second option would work of course. It does achieve the IOC aspect,
> but doesn't remove the Spring  (if the GenericFactory interface comes
> from Spring), and you would have to cast in the using bean instead of in
> the proxy. The IOC and hiding of other beans in the BF are probably the
> most important things anyway.
>
> (Of course, both of these do have a negative aspect, compared to just
> using BeanFactory directly; with BeanFActory the using bean can declare
> itself BeanFactoryAware and get the reference with no extra
configuration).
>
>
> roger holbrook wrote:
>
> >Hi Colin
> >
> >Your thoughts on implementation hiding have got me thinking...
> >
> >But before I test anything could you clarify the sort of usage you
> >have in mind - specifically, would it make a difference whether the
> >implementation hiding factories were themselves 'typed' or 'generic' ie.
do
> >you want / need the type of the factory to vary with the type that it
> >generates, or would a single generic factory be adequate ?
> >
> >In other words, does one or other of the following make a difference ?
> >
> >1.  Typed interface factories:
> >
> >interface MyFirstInterfaceFactory {
> >    MyFirstInterface getInstance()
> >}
> >
> >interface MySecondInterfaceFactory {
> >    MySecondInterface getInstance()
> >}
> >
> >
> >2.  Generic interface factory:
> >
> >interface GenericFactory {
> >    Object getInstance()
> >    Class getClass()
> >}
> >
> >where different instances of GenericFactory will return from getClass()
the
> >relevant type, MyFirstInterface, MySecondInterface etc..
> >
> >I'll try & test something tomorrow
> >
> >Roger
> >
> >
> >
> >"Colin Sampaleanu" <col...@ex...> wrote in message
> >news:3FB...@ex......
> >
> >
> >>I think I may be missing something, but I think it's desireable to be
> >>able to create in an easy fashion, 'closures' which encapsulate getting
> >>a bean from a bean factory.
> >>
> >>Say you have an interface
> >>  interface MyInteface {   ... whatever }
> >>
> >>and you have a factory Interface
> >>  interface MyInterfaceFactory {
> >>    MyInterface getInstance();
> >>  }
> >>
> >>And you have a user of the factory, who you'd rather have no knowledge
> >>of Spring, which is why he's using the factory instead of calling
> >>getBean himself:
> >>class User {
> >>  MyInterfaceFactory _myfac;
> >>  public void setMyInterfaceFactory(MyInterfaceFactory myfac) { _myfac =
> >>myfac; }
> >>  public void someMethod() {
> >>    // need a new instance of MyInterface to work with
> >>    MyInterface myint = _myfac.getInstance();
> >>      ...
> >>  }
> >>}
> >>
> >>Now I have a bean factory
> >><beans>
> >>  <bean id="myinterface" singleton="false"
> >>class="com.whatever.MyInterfaceImpl">
> >>  </bean>
> >>  <bean id="user" class="com.whatever.User">
> >>    <property name="myInterfaceFactory"><ref
bean="xxxxxxxx"/></property>
> >>  </bean>
> >></beans>
> >>
> >>now, as per the above, context.getBean("myinterface") is already a
> >>factory for objects implementing MyInterface. But I don't want the User
> >>object to know anything about contexts. And I'd rather not create an
> >>actual object that implements MyInterfaceFactory. It seems like a waste,
> >>since all I am doing here is trying to create a level of indirection,
> >>and I already have a factory inside the context itself, and I may want
> >>to use this approach in 30 different places, just to add a level of
> >>indirection in creating new objects.
> >>
> >>So what I think is needed is some variation of ProxyFactoryBean (but a
> >>separate class), which given a target bean (which is itself a factory),
> >>and a factory interface having a method with no args which returns a
> >>certain type, creates on the fly a new class implementing the factory
> >>interface, which will just use the target factory bean to actually
> >>supply the instance. So the bean def above would become:
> >><beans>
> >>  <bean id="myinterface" singleton="false"
> >>class="com.whatever.MyInterfaceImpl">
> >>  </bean>
> >>  <bean id="myinterface-factory"
> >>
> >>
> >class="org.springframework.whatever.XXXX">
> >
> >
> >>    <property name="targetBean"><ref bean="xxxxxxxx"/></property>
> >>    <property
> >>name="interface"><value>x.y.z.AFactoryInterface</value></property>
> >>  </bean>
> >>  <bean id="user" class="com.whatever.User">
> >>    <property name="myInterfaceFactory"><ref
> >>bean="myinterface-factory"/></property>
> >>  </bean>
> >></beans>
> >>
> >>Am I missing an existing way to do this? Is this worth adding to spring
> >>as a convenience built-in, along the lines of
TransactionProxyFactoryBean?
> >>
> >>
> >>
>
>
>
>
>
> -------------------------------------------------------
> This SF.net email is sponsored by: SF.net Giveback Program.
> Does SourceForge.net help you be more productive?  Does it
> help you create better code?  SHARE THE LOVE, and help us help
> YOU!  Click Here: http://sourceforge.net/donate/














begin 666 TypeFactory.zip
M4$L#!!0````(`-MTA"\1&L1BE (``#P'``!S````1&5V+TED96$S,#(O<')O
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M`WTY<+5C?UAH0ET>[]!E?O7C^D/28N0)M+4M-\58'YH]ELT9\H[KZX*HCI.4
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M!PKR?R-_4+8E'?@6NI5(,XA#(0GR;&A&`)TZ<E<VO=H3<]^<O\O%/U!+`P04
M````" `*7(0OU4]^,.,#``"$"P``<0```$1E=B])9&5A,S R+W!R;VIE8W1S
M+U-&7T1E8V5M8F5R,#$O<W!R:6YG+W-R8R]O<F<O<W!R:6YG9G)A;65W;W)K
M+V)E86YS+V9A8W1O<GDO<W5P<&]R="].86UE9$)E86Y&86-T;W)Y0F5A;BYJ
M879AE551;]PV#'Z^`/D/+ H,=G!P@+VFMW8M.J! L0%+_H!LRV<U/LF3Y+L>
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MD4^*-+\,<J"[IW._N"N-Z3@5]X@`G?1&!P/6TH;$A+1[.U0>=;1:!2OHI+!/
M@HR6>FM,F+2].I(Z"-(+BSV\*&T[%UIE#BA(EF=!XXQ=[P5Q[Z5U6U0`.K<X
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M,5P*R#25'H!4`]IX>-;FQ$6+^GB9W:LJ@P"G(BF[+%^O:U3+(O$O[M*1%_KY
MRB3%5C]1B0I+X56:<BZZDSB/0\V1'"\)OG\_%XIJ0M@RC% :=KWZE7%;J_[U
M6?C\SR ZY<_X'./#>%EHM$?3@:4MO-N-VJ;7]9LXBJ(3>E^$<M]*@G+9\CQ?
M9C_F'>VC5 T]##%]I"+CWQARFHNIJ@:]PN3/[*C8F1T^$O')7%UT4#?EKQ#P
M%VCCP**,USQCEIL(D*W;,!0#!\H9_\ULS161`@*:#O-7BGUS-:=+A'B^#I)*
M(Y 3<L,__P-02P,$% ````@`.'F$+Z L'8;3`P``;PH``&D```!$978O261E
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M[!Q*`FDH7&Z0H'*'*+E[E=-<MMM$56Z[J>-$QK\3*N "@+/FZ.Q]NM%T@,(B
M847L"2T1<"<%%WZ0XS_=ECZ^"PY:(35'H=&!`](_QF^LZ6.0/ W9.)-PVZ4-
M%_)F?K<3"W]W4N%<W)37$%GNQ4CD`1+)9J2G))\I/64"I4GV=SBSNXF9!$>4
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M!!0````(`*Z:AB\SY% _4P<``!X8``!M````1&5V+TED96$S,#(O<')O:F5C
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MK$T92"<$3@<UKVV1`V&9J\J$NBJ1R8/6P@17DDA;^J#+# *W@.\VN@HVJPM=
MJ?RNU&N;J4WE;N\:I8T.S!Q,C>N[C<FC%2 JF I@X*Q0+S>OVF2K7$$:0NO 
M%#T(3ND,757:*_0?M6T,V& \R0=[`!X0ZU5I3&YRD@AL-B+L#='--7R(-D:'
M4H6:X3&F:V4R8V\@J[5_P<;5!5OWLK"OGI,,<#>BL.@[M)<RZ&II`ID^5=N5
MS59JK>\@O/2G=/A5-TXLZKA5'T&:+0)5F]TYH%8$BBF%L[G:VK!2-@S@FC(R
M8)U(7L38FS#M8XF0; `LD(>IXGH8=A)C2NF#!#$C3#<E*% EI5K@P/_JMN;&
M5 1(85K$\B&N*@,?YJTMN84XA>)N:-'..Y8D<"<9K@2NN72JD05YM@"\J%[*
M7K:+L+#E%PTI1&[*?FGU`34*4+G:9)# 7-"=.7A69]@[$;&=*396&@1M4VC*
MS(N28^UMJ*E>.41<I"I 57&U& 5N7D1=R41!_%:.@Q8[\A#3=>TY[;RWRU+/
M(02+RJU)FI0L,.P6[%4]]^:O&LH<X2,[W?P+H+GS^@OJ6-1EAK9[K%Z-LFVF
MBQ%4@\*>7S1%W3:@7H]II=L2D/ ;D&P06[#R(^D'`$!!$;/A"2%*'B'>.ZD@
M)=1>+1R/+P>_*F8D>XCYM3</F56REL3G[Q$ALZ;?!QXH(D[,5'9;DM=_\*-"
MK_B9&GD6!_VFGA<0,*[]@3WF-I@R]V/\N]GA)0Q3"2O-6O4/3KR#X^-.J^TE
M+ 67^^V6.SRDT(W-<2YI5M]H`5F PXT.L/($6GE8#NK%TCZ1ZM[&A/S434@:
MIL(!JHE]6CA7.:>;TF4E$;P^;)#6X//!-S8$1_T!5B&X>@\"*5.]IJX>OT!W
M]M$3^?'.B!MG<YP-UT22=*!!.6S3`33/1 @_/55E713JZU=!FA:F7(85N %?
M_QP9#\*J<EN8VEMH=C<PO'+I=KO@)$=L`,U*,.=H<H+<W_ 7HIQV@Z5.A=Z3
MAX'6U.]%VR$IG%L+?D";K+%+0Y?)*H/1"Q35)O+^/Y%O^^X`_('>R$? >'6&
M)?3!A7,'FW:+!TZK<:-AS,5]8H$<^^+9,C0A[1@X4?=K;V/>=4N$O?/%CT5^
MY]M(\,G 9KKM*$_9\A3J'PWHV@G#JJKA]_4*8IFG65U54*[\OV220@#/'-#?
M!I+QWFFX<)()*Z5D`QO.>QJ3O@F3>]).C@L>/W*?PRE%&SG1=E9G'H \95ED
M[9L=6[8]%+$OGTVU2^;.>MS.[V[FSN4\DYF)7OA!1DHSP+6A%=B%U]:CW5.D
M8#X6W?P,N7C94I0GC7TXP$9&R$]$!$4ZB-$97Q3RSV1?@8@8)3)>`HZV2GI(
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MEZH7*/&\%['13+KP^Q"6+W,'9H^,AWZ\8;2U^N#7OU!+`P04````" #;=(0O
MGY/OD0$!``"G`@``>0```$1E=B])9&5A,S R+W!R;VIE8W1S+U-&7T1E8V5M
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MP(8J?JLLH:=%[K:55A*DQMBE1=^BZ]>7#M.>R=0!;D<"7VDFV8"Z!QDO.94`
M;6A>DBO+8M;DQ^D<!YL='P!-8)5FCY4FX)_4293.X+B=`%!+`P04````" `9
MIX4OAL&229H'``"=& ``;@```$1E=B])9&5A,S R+W!R;VIE8W1S+U-&7T1E
M8V5M8F5R,#$O<W!R:6YG+W-R8R]O<F<O<W!R:6YG9G)A;65W;W)K+V)E86YS
M+V9A8W1O<GDO<W5P<&]R="]4>7!E9$9A8W1O<GE0<F]X>2YJ879AM5A;;]M&
M%GYV`/^'62U@2*E*`_L8UT5=Q]UXX3AIY6 ?@B 8D4-I:HK#S@RMJ$7^>\]E
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M:0JT@T.&/Z>AG%[N2KG1J2 #Q<!KD1LKTD(ZIYP`J87:J-)C\6%UQ3D1NO3*
M0O*XL,1WU?>A7J$X28*P*E7Z`01)^"]24T+%UL@LI%W5*'@.Y['0E\JE5E=(
M8I:_@@-SDIG1\1+-6)NM\$8L:UUDP%Q)ZW5:%]**+'A6D6=LP;X';.&X%P);
MS2J9PP':E<JB0%4KY2E8TYG8:CBRRM>V!(*@D&2R4@T>RC)5H$KZ3KFCX%GU
M6ZT<MFN^KQO4E>!-X0P)0^KVF>O;>W(@7OTD+.JE0W6E+W8H?$?>.&%*$MX/
M56LUN;=43=XRL=Q%]CR8>P4QV%#MLR3(])KKACT`3*JD!P#*,'"F!(C)7[!A
M=<&6?5?H[[]EUI!A*CE4`09#"%H5\QZ]&OC-9=(8SV)Z7AXHK"[JD0+(IBK4
M2GHE(OCJ0H/",8P&!+,:9CYM_0J1_\&IOP"L22^W\9>O%18Y/'[\M0HH/OL'
M7RMTP6=BY Q%APE6U<L"JI=;/"9E4(OZ;HAO?R 8'H%A#YCK7)>R$)!C`!&Q
M\#1E_WUU]_'Z=G%W<7MY]?'UU=VK-R\_WEZ\OA+G8@)8<!U*8G+6DQ1D\&00
M5S^_N[A9!.ZS@V2HZLV/_[FZO.M(@?CT5/Q7B5(I1+C L2Q,>H]=Q0TGH,G2
M>R!0S6ATP!AHP<>C(P\%11^.>L: %V^H[Q(*7P(>L3'3"???9 Z:M^(2G[[_
M(/[HD7^>G9'(@>$@-BJ#H6SXR'+VQ0>)G_%7BM@AIK=F4:=KYFPGOU"SX Y$
MYU*6I?$0BJI29</^N9>/`[T`0RC7JY Z+J-!`4T/\/J\^Q*,01Q+6"9$("8X
M:RVBR7L$?7'=U"45)$#CL#R3!F:3P%*F9H-%&5"7\6TC=PC3?;15FJ I1E:6
M@5 83&P0-P!N"WG-0%I2"Q^)8/ /,%WE)DP)9.!/8/MVK2%3)!FGY%:B49GJ
ML063D:C]"%,1:=G+K %J$MOCA?W M6.'(]3N# Z&/JPGY#1Y&\4F"/%K:[8`
M#?A'+@NV$H)#`P[1":&\ZQS<3T#93N"V3+PEC#[2[T&K\B3VM"L9A@TNZ&8R
MAF_DR[QI\##)`BG4.SHV$P/[J)AT+J:]7DW"1&0ILZ@!HB5K;\O8&Z)">Z>*
MG/C"XL(/FDK#",[9K*X15>$@]&!.H(40O%2@#O&1.A<*5+L+Y_2J1/M_LF9S
MH&L8"EK;#]H0NNC_8THDD!FXSB.$`BYX$ 1\L74'A?TM9@<-U\V^$AD:1A5L
M<0IPI]-Q"P?3@-&780</6^<OY!LFJ<?1'??YX":A"[2LQ_>E@>5ZCZ5"T=X8
M<\^+E(CF*)0J8PAV:=EB$C%A`-''IA$.C>:9.#DAAJ,1=X=Q'O&MC2P:VD$=
M`48JG1=+R>.W`X5ND03<!+=4N!GP50/Z$ZZQM6K,HL(:C,W_M<:PR$3X1_5&
MB$*3=71Z3B?MR*3M(8##^XGXIOG\C9A\>$'NF;+8";J#AV>.@+&7L[D`5O9M
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M=-[6BA'Q:'@Q]GGU+\'"]T0U3H /`[9F,@;NZ0@!"1E9?CC9VM%?QFTB'=N]
MAE=UO09U+()FS]YS%A7F(6K[QQ0X'O6$0S6,Z6.^Z?6(8X]&;7!+;),3=5XH
MA,B'3D./J;.2VP$7E.L<UR6<HOC&<LZ)U3S*2&*372I(K!,@.%0GAXU#KG#4
M]2/]^A-02P,$% ````@`PFV(+TG P<"4"@``6BP``' ```!$978O261E83,P
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MU%1 #=%8@0>*'5J ]ZSX+U):SO\`FAG)AMH67(.X`;<:>0&+I2A%A8(CG34=
M<37F<W)%3+C,;X!5SR"ALF>A:C&'5^+/&G12[&^M53C9'AP+`0Q\U=\-T[HP
M7%\7^9M7)$3?'Q@8HI569-Z*UG^8$$BZ7>7IRM>]I_<V+XK6O6L*?+!D42@F
MR[@,9L=C)RV$E!*>,N2'UJ$B!R85XW!=PJY5C=P8KY8URJ]F3-)-DAB$VPB(
M=J&2!K(\QJ4P3@7FE-D<(8LKRI&\FC$E4>"\<G<IP!5JQW<$;I"JQ!?5DJ6-
M1M'=&]=4@F>AJ]\J(=A/;E_KY(O&+T,WVK7[`Y<?1N]MLH*5_,L/87L)MS/,
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M=HO_H;))7X)8`H"TO1"]RZK=BC%BV2;:J$P,F3?FZ?% \DQ,% ^TKMLX5&U?
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MDRB0^=(`'A-A([#R1V+'IH^1B(40AC(<H%M!;,DRXXY5<4R8XR98WEIWC/4F
M+$<K[RQC&)HA(+,<[Q"D8(L"%CQH0-?.'=11A36$[;-;`\A/;;OB,AM0S8VK
M3-8M;#*XPIO<'VZ'0IRX=]L;Z\<6LBY=,V;P8=*K[P#:$<Q.AB;OJ9$./M:Y
M\2YH-EB$$=BH-YD>C1.9.FBZ%JM"VR Z@".MNDVX&8-MVE"#17Y6#?)QUV:P
M`(!(Y^1_BW _`G _C&^/`F^MN"T`_1_=[D(W6FH#]BV=QO7=WY?%'--1-&--
MEI6+33-+F/BDJ&P*><;X7-Z(V$Z8:00!^\G?]O?!@:8;7D",+:\RX3+(?JZQ
MRVB.&]F:[[QC3*X(J)(AZ5H,\&1L/$_A82/*4'ICTH@B#Q?01G. G_>)\J'A
M;19(A!4-R<AQ?CDSH5@T]>R_4A_-4&+/$LT4.5@,ALZV@NLNS"*76__>L_AV
M([A3/:$6`U4I._&+E* R#&1Y00`#XM1KXT(8P91F&PGH`W>(2V\3?SI-*!C#
M(:"7P9'FOTN3Y,H?-%SY!PTHS4DV;UMFSYT5^?2>6&98CY7[BY9+(Y7JR3SK
MNM;69Y1:?-,53_6CI(5AVC]N'NA9NH<=7;$]*5O;_G"!:G@.< M"L(]49GDG
MICW<ZX20(?6-`0%TUDCT`02*,VI#JYD2N\<\^+FUW_NVG_8LB$3%>F2M]?R0
MO1Q^`O'2;N+P[O#Y#&.@YW_#II=5+VG!8<-F,CD$[UQ^NJ:'OX!;Z)IMKE=,
M\PK/&,@=/J^.C9&A:Y.=!03,U8W%P+M;P5;\1O1 Q6M//&!YU2ZDYV=Y<[P!
MH[5&9O@\+Y.6;-@A1AK7NS_T)'S8CLQ:SF_DPN/;1SI]]V:13K_^BW@06HP_
MOJ$$-T^=N6'?H#_V<L%8S@MI'T]O^>[ %D*,2S<K!L6PBS*C#SBBV.7.V' +
M#;(3)?X(@*F?#7>,`TF2.$;F50NVCCS), =+4+PJD1FRZ5">SLCOI,)OO[._
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M`,;?O\_H70+EMQ$`^YX&G4=UM*@[+& _X9G!6XUH399I5:$58]+1%EAJKG!1
MZLYP;3SB9H64&P:!:WC[-J:Y4)29R/I28&-L*_XD4+#9T!/";H8K80K2+ >Z
MGT_@Z[7'V!8YN/SR93/VH)54PK-LVA+^EO_NS46@02;+YU@IS-@0VHPDQ1?<
MVK>S[,(QJP6;M=;R&_LFDNF__P!02P,$% ````@`PFV(+Z3$C/J'" ``TCD`
M`',```!$978O261E83,P,B]P<F]J96-T<R]31E]$96-E;6)E<C Q+W-P<FEN
M9R]T97-T+V]R9R]S<')I;F=F<F%M97=O<FLO8F5A;G,O9F%C=&]R>2]S=7!P
M;W)T+U1Y<&5D1F%C=&]R>51E<W13=6ET92YJ879A[5I=;]LX%GUV@/P'-AAD
MY2*5:TF+`1H4F,QTNLCNMEVTF9UY&\@2%;.51:U(^0.#_O>]I"B+DJFOU,&D
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MWZYB^[=5+ @9U-NGJ283F)GM;C"G];Y7X9HP'$J%S9X6`?2#G_,ES=![&J)_
MTF7":%)<7^.,`3/HN^OP!:J/5VCO0TXXMH7&+]9H;L^_1\[SY^YL/I_-_X[F
MS@O/>^%\CSXN*8YCB+^?MRGZ3L65C,T`!;'/&-)9W.,BO.4X">&N"@KTAQCS
MZ<E$]5U3$D(\,ZYZ7B<0W$ POF9O"&/ @36%$,_HAJ']) N0R60V0SX2,X+4
MA#PC96?D<YZ110X#6.6,HQ!')(&K"=HW$0A:0(M4\Q+=8OZ3<,::VO#U/68T
MSP)\Q8HVUAG7?!33?S:]%#AU4M%V$0%6@C<-MBW"BO8<&DD7)N\6'W$`,P[M
MH9<P*MI;9U$;&\KB)/));!7?/XM_@<^#);(JBO!VJFP`3^\^R6_@&>2"FRS'
MEH6WP 7C?A)@2*(]:6**SL\E`JK^`,'F%)B1<V03R,[;=Y%UUA6=9U/TY"5Z
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MA<]XL7TED72]2F.\P@F'"?\E83XG+"(XW-^?WREH`C]1>W)$0:U9%2Q"S2BF
M0EU,"'F'Q$D:*4>!0[38E5(&F8<XQK<^!)\:MZ#MHHB7#J5[:+,DP/(>EB%U
MEE&&H;K/T%M1`XEFRC9ZI2QJ!A^(F@^])'P1>5+1!_=Z--T[[=^(?)W')E]M
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M8%B_]&5QX R.`Z<M#LP0G>([5/ 8D/8X&(&R#\G'6U+IQJL%]49\R/.B/WE!
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MBWD]/8&/_P-02P,$% ````@`XIJ&+VS3>J@=`0``,@(``&X```!$978O261E
M83,P,B]P<F]J96-T<R]31E]$96-E;6)E<C Q+W-P<FEN9R]T97-T+V]R9R]S
M<')I;F=F<F%M97=O<FLO8F5A;G,O9F%C=&]R>2]S=7!P;W)T+TEN=F%L:61?
M1F%C=&]R>3$N:F%V89613VO#, S%SPOD.^BX]9"P<R_;&(7 8(?V/AQ'2=4_
MLI'EEC+VW6=G7MFQPP<;\7Y/>G*[6-05+&"S15A[(9Y@%'/$LY,]4("!@@KU
M47& R ,*:%(^>V.W.(-K-^K9",(;6>2 3:ZV=5572;,W$X*3J0FS]=6YZ=%P
M:$9CU<FE"=%[)[K,%!WS\P9H]7._I.+R!JS;8- BSJ<MP8L-=*PHR1LAAI25
M&*SC762KY!C.I%LPL+EX' KQBL$*^?2<?=3!A(QB%$&S+%D$-6PQ@!OAVG[>
MSTP\!?S'<IJ_O<N.?>P/9%.CW]$[/ID##1]%]@CPF;/>O?<[M)H&U*X,=?^P
M!&C;Q)X2`(%2#00U"L_CY[]GIV#XFB,[?=75-U!+`P04````" #39(@O/^)!
MY10!```R`@``< ```$1E=B])9&5A,S R+W!R;VIE8W1S+U-&7T1E8V5M8F5R
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M% ````@`_&2(+S-?`$L5`0``,@(``' ```!$978O261E83,P,B]P<F]J96-T
M<R]31E]$96-E;6)E<C Q+W-P<FEN9R]T97-T+V]R9R]S<')I;F=F<F%M97=O
M<FLO8F5A;G,O9F%C=&]R>2]S=7!P;W)T+TE497-T0F5A;E]&86-T;W)Y,BYJ
M879AE5%-3\,P##U3J?_!1]BAD[CN`@B0)G%BNZ,L=;MLU(EL1]V$^.\D)50[
M#N5@RWH?]LMRL:@K6,!VC[ )[*B'CLV H^<C.('6B;+;1<46(K7(H GY&(S=
MXT3<^$Y'PPAOSB()-GFZK*NZ2IBCZ1$\]XU,TK-RLT-#TG3&JN=S(S$$S[K*
M+#?D]@K2ZV]]2L/5%;3U%D4+.+]E.;S(P)H4.6DC1$FW.@+KZ1#)JO,$H],]
M&-B>`[:%\8QBV8743CKJH4="-HJ@&98D1 U9%/ =S/93/A/C0? ?X327WB7C
M$'>?SB:CO]5GEX\"O <\*5(K<!$7?.4`;F9P6ES79=G;N_3![$>!]TCJ!GPY
M60PY@RFW[[KZ`5!+`P04````" #\9(@OA<O =2$!``!G`@``< ```$1E=B])
M9&5A,S R+W!R;VIE8W1S+U-&7T1E8V5M8F5R,#$O<W!R:6YG+W1E<W0O;W)G
M+W-P<FEN9V9R86UE=V]R:R]B96%N<R]F86-T;W)Y+W-U<'!O<G0O251E<W1"
M96%N7T9A8W1O<GDS+FIA=F&54K%.PS 0G8F4?[@1*N0.C%TH0DB1V-H=.?8E
M-6UMZ^Y"J!#_CF-,Z$:9[GQZ[][SLY>+15W!`K8[A$TDYWOH2!]Q#+0'QV =
M"[EV$+0P>(L$DI#KJ,T.,W$3.ADU(3P[@YY13=-E7=55PNQUCQ"H5YQ7SYM5
MB]JSZK210"?%0XR!9#6QW'%J+R ]?=>'-%Q=0&NVR/(W6(>H?D]K^^88;396
M5\N255&&Q@M2LH,P)! X#R;XU\$;<<'#Z&0'&K:GB+8P'I$-N9C:O$<"].B1
MM"#(!$LK6+0WR! ZF!WG2#/CGO$?>:IS[?(L<6@/SB2A'^NSRDL!W@&^"WK+
M<);P+90HX&-*XFIFI1M(4UQ?WZ3/06'D])M2T>T!<W*?=?4%4$L#!!0````(
M`,)MB"]"U830&@$``%H"``!P````1&5V+TED96$S,#(O<')O:F5C=',O4T9?
M1&5C96UB97(P,2]S<')I;F<O=&5S="]O<F<O<W!R:6YG9G)A;65W;W)K+V)E
M86YS+V9A8W1O<GDO<W5P<&]R="])5&5S=$)E86Y?1F%C=&]R>30N:F%V8951
ML4[#,!"=B91_N!$RN M;%XH`*1);LR/'OJ2FU+;N+J05XM]Q3$C9*)//I_?N
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MH\$X!<_+^BR++U!+`P04````" !JG(8OQ)]H'<,"```<&@``:0```$1E=B])
M9&5A,S R+W!R;VIE8W1S+U-&7T1E8V5M8F5R,#$O<W!R:6YG+W1E<W0O;W)G
M+W-P<FEN9V9R86UE=V]R:R]B96%N<R]F86-T;W)Y+W-U<'!O<G0O='EP961&
M86-T;W)Y+GAM;.6946_:,!#'GSMIW\&U]IJXT#Y,4Z J!:I(&T-MT+2GRI #
MHB5V9#NE?/LZH80V#2LD`QK&0S$.=_?W[\X^HUJ7CX&/'D!(C[,&KIEG& $;
M<==CDP8>.%WC*[YL?OYDG;9_7CN_^QTT!,HDZ@]:W^UKA+!!R%W_UN[=$-)V
MVJC5N>H1TNEAA*=*A=\(F<UFI@R%]C<6-( 9%W],+B;$52Y9S!N)2U-/8!U)
MQTH^)\.39(P\MX&'G/MZ[(!4+?V.]<.3D4^E;&#M[DV(A<]6QJ@96UFAX"$(
M-4=,?[N!X[^X:3U0/X*FC6B *,H86F3QU")+V]B31>(H6:%J*X7;2Y->$/J 
MRDB[WY4VUQN/00!3952.Z4AQ,;>9`J''8,L?GI1:WV:JG\U-&84A%\ITYB&X
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M!!0````(`->:AB\DZ?<R'@$``#H"``!N````1&5V+TED96$S,#(O<')O:F5C
M=',O4T9?1&5C96UB97(P,2]S<')I;F<O=&5S="]O<F<O<W!R:6YG9G)A;65W
M;W)K+V)E86YS+V9A8W1O<GDO<W5P<&]R="]);G9A;&ED7T9A8W1O<GDP+FIA
M=F&54<%.PS ,/5.I_^#CMD/+?1= "*D2M^V.LM1MPU8GBEVF"?'O."%,'(=R
M2.2\]_S\W&XV=04;V$\(NQ =C3!$,^/9QR,XAMZQ1'=8!'M8J,<(HLC'8.R$
MF;CS@YQ-1'AU%HFQ2=6VKNI*,4<S(O@X-IREK\K-`0UQ,Q@K/EX:7D+P4;:)
MY>;TO('T\G,_:7%[`ZW;(TL!I].6P8L,="0851MA89W5$5A/[PM9<9[@[&0"
M`_M+P+XPGI%M=$&?64<\C$@8C2!(@JD$BR&+#'Z :_N<3V8\,/XCG.9O[Y)Q
M6 XG9[71K_6./LS)]6\%=@_PF6:]:UL%Y2]@IY;2$G5ENE]*MJ4K5E=KF%$F
MWROGZC@!=C[5U>5JG?/[JJMO4$L!`A0`% ````@`VW2$+Q$:Q&*4`@``/ <`
M`',``````````0`@`+:!`````$1E=B])9&5A,S R+W!R;VIE8W1S+U-&7T1E
M8V5M8F5R,#$O<W!R:6YG+W-R8R]O<F<O<W!R:6YG9G)A;65W;W)K+V)E86YS
M+V9A8W1O<GDO<W5P<&]R="]4>7!E9$9A8W1O<GE$97-C<FEP=&]R+FIA=F%0
M2P$"% `4````" `*7(0OU4]^,.,#``"$"P``<0`````````!`" `MH$E`P``
M1&5V+TED96$S,#(O<')O:F5C=',O4T9?1&5C96UB97(P,2]S<')I;F<O<W)C
M+V]R9R]S<')I;F=F<F%M97=O<FLO8F5A;G,O9F%C=&]R>2]S=7!P;W)T+TYA
M;65D0F5A;D9A8W1O<GE"96%N+FIA=F%02P$"% `4````" `X>80OH"P=AM,#
M``!O"@``:0`````````!`" `MH&7!P``1&5V+TED96$S,#(O<')O:F5C=',O
M4T9?1&5C96UB97(P,2]S<')I;F<O<W)C+V]R9R]S<')I;F=F<F%M97=O<FLO
M8F5A;G,O9F%C=&]R>2]S=7!P;W)T+U1Y<&5D1F%C=&]R>2YJ879A4$L!`A0`
M% ````@`KIJ&+S/D4#]3!P``'A@``&T``````````0`@`+:!\0L``$1E=B])
M9&5A,S R+W!R;VIE8W1S+U-&7T1E8V5M8F5R,#$O<W!R:6YG+W-R8R]O<F<O
M<W!R:6YG9G)A;65W;W)K+V)E86YS+V9A8W1O<GDO<W5P<&]R="]4>7!E9$9A
M8W1O<GE"96%N+FIA=F%02P$"% `4````" #;=(0O...
 
[truncated message content]

Re: [Springframework-developer] Re: Re: Creating Application Context getBean 'closures'

From: Colin S. <col...@ex...> - 2003-12-08 16:16:32

Hey Roger,

This is quite interesting. I started implementing something (fairly 
similar to your approach, although more simplistic)  but ran out of 
time, so in the meantime just kept on using BeanFactory.getBean in the 
couple of cases where I would have used this.

I'll take a look at this in more depth on the weekend. This is 
definitely useful to people using Spring. The only real question is if 
it's useful to enough people to warrant including in the core, given the 
size.

Regards.
Colin


roger holbrook wrote:

>Colin,
>
>A long delayed response, but just in case this topic may still be of
>interest to you...
>
>I played around with a couple of ways of doing this, & have attached a
>straight forward implementation of the initial scheme that you first
>outlined, ie: given the following:
>
>public interface ITestBean_Factory1 {
>    ITestBean getInstance();
>}
>
><bean id="testBean"
>  class="org.springframework.beans.TestBean">
> <property ...  >
></bean>
>
><bean id="testTypedFactoryBean1"
>  class="org.springframework.beans.factory.support.TypedFactoryBean">
>  <property name="interface">
>    <value>
>      org.springframework.beans.factory.support.ITestBean_Factory1
>    </value>
>  </property>
> <property name="target">
>   <value>
>     testBean
>   </value>
>  </property>
></bean>
>
>then a call to beanFactory.getBean("testTypedFactoryBean1") will return
>dynamic proxy instance that implements the given typed factory interface.
>
>
>FYI, the following is a short description of the 'other' approach I tried:
>
>My first thought was that your initial scheme could be simplified by using
>a factory-interface attribute on the target bean definition, & doing away
>with the need to have a separate typed factory bean, ie:
>
><bean id="testBean"
>  class="org.springframework.beans.TestBean">
>factory-interface"org.springframework.beans.factory.support.ITestBean_Factor
>y1"
>  <property... >
></bean>
>
>With a few tweaks in core classes, this all works fine - the only problem is
>that it just doesn't feel clean enough, ie: it requires the following things
>to change:
>
>- new factory-interface bean definition attribute
>- new behaviour for beanFactory.getBean("&beanName"), since the typed
>factory dynamic proxy instance returned, will no longer reliably be an
>instance of FactoryBean
>- small, but none the less, additional complexity to AbstractBeanFactory,
>one more level of indirection to contend with
>
>On reflection, in this instance, end of idea ;)
>
>I would be interested to know how you ended up actually dealing with the
>issue in practice ?
>
>Roger
>
>
>
>"Colin Sampaleanu" <col...@ex...> wrote in message
>news:3FB...@ex......
>  
>
>>I was thinking of the first, with the idea that there is no Spring class
>>dependency, and you achieve total inversion of control.
>>
>>Your second option would work of course. It does achieve the IOC aspect,
>>but doesn't remove the Spring  (if the GenericFactory interface comes
>>from Spring), and you would have to cast in the using bean instead of in
>>the proxy. The IOC and hiding of other beans in the BF are probably the
>>most important things anyway.
>>
>>(Of course, both of these do have a negative aspect, compared to just
>>using BeanFactory directly; with BeanFActory the using bean can declare
>>itself BeanFactoryAware and get the reference with no extra
>>    
>>
>configuration).
>  
>
>>roger holbrook wrote:
>>
>>    
>>
>>>Hi Colin
>>>
>>>Your thoughts on implementation hiding have got me thinking...
>>>
>>>But before I test anything could you clarify the sort of usage you
>>>have in mind - specifically, would it make a difference whether the
>>>implementation hiding factories were themselves 'typed' or 'generic' ie.
>>>      
>>>
>do
>  
>
>>>you want / need the type of the factory to vary with the type that it
>>>generates, or would a single generic factory be adequate ?
>>>
>>>In other words, does one or other of the following make a difference ?
>>>
>>>1.  Typed interface factories:
>>>
>>>interface MyFirstInterfaceFactory {
>>>   MyFirstInterface getInstance()
>>>}
>>>
>>>interface MySecondInterfaceFactory {
>>>   MySecondInterface getInstance()
>>>}
>>>
>>>
>>>2.  Generic interface factory:
>>>
>>>interface GenericFactory {
>>>   Object getInstance()
>>>   Class getClass()
>>>}
>>>
>>>where different instances of GenericFactory will return from getClass()
>>>      
>>>
>the
>  
>
>>>relevant type, MyFirstInterface, MySecondInterface etc..
>>>
>>>I'll try & test something tomorrow
>>>
>>>Roger
>>>
>>>
>>>
>>>"Colin Sampaleanu" <col...@ex...> wrote in message
>>>news:3FB...@ex......
>>>
>>>
>>>      
>>>
>>>>I think I may be missing something, but I think it's desireable to be
>>>>able to create in an easy fashion, 'closures' which encapsulate getting
>>>>a bean from a bean factory.
>>>>
>>>>Say you have an interface
>>>> interface MyInteface {   ... whatever }
>>>>
>>>>and you have a factory Interface
>>>> interface MyInterfaceFactory {
>>>>   MyInterface getInstance();
>>>> }
>>>>
>>>>And you have a user of the factory, who you'd rather have no knowledge
>>>>of Spring, which is why he's using the factory instead of calling
>>>>getBean himself:
>>>>class User {
>>>> MyInterfaceFactory _myfac;
>>>> public void setMyInterfaceFactory(MyInterfaceFactory myfac) { _myfac =
>>>>myfac; }
>>>> public void someMethod() {
>>>>   // need a new instance of MyInterface to work with
>>>>   MyInterface myint = _myfac.getInstance();
>>>>     ...
>>>> }
>>>>}
>>>>
>>>>Now I have a bean factory
>>>><beans>
>>>> <bean id="myinterface" singleton="false"
>>>>class="com.whatever.MyInterfaceImpl">
>>>> </bean>
>>>> <bean id="user" class="com.whatever.User">
>>>>   <property name="myInterfaceFactory"><ref
>>>>        
>>>>
>bean="xxxxxxxx"/></property>
>  
>
>>>> </bean>
>>>></beans>
>>>>
>>>>now, as per the above, context.getBean("myinterface") is already a
>>>>factory for objects implementing MyInterface. But I don't want the User
>>>>object to know anything about contexts. And I'd rather not create an
>>>>actual object that implements MyInterfaceFactory. It seems like a waste,
>>>>since all I am doing here is trying to create a level of indirection,
>>>>and I already have a factory inside the context itself, and I may want
>>>>to use this approach in 30 different places, just to add a level of
>>>>indirection in creating new objects.
>>>>
>>>>So what I think is needed is some variation of ProxyFactoryBean (but a
>>>>separate class), which given a target bean (which is itself a factory),
>>>>and a factory interface having a method with no args which returns a
>>>>certain type, creates on the fly a new class implementing the factory
>>>>interface, which will just use the target factory bean to actually
>>>>supply the instance. So the bean def above would become:
>>>><beans>
>>>> <bean id="myinterface" singleton="false"
>>>>class="com.whatever.MyInterfaceImpl">
>>>> </bean>
>>>> <bean id="myinterface-factory"
>>>>
>>>>
>>>>        
>>>>
>>>class="org.springframework.whatever.XXXX">
>>>
>>>
>>>      
>>>
>>>>   <property name="targetBean"><ref bean="xxxxxxxx"/></property>
>>>>   <property
>>>>name="interface"><value>x.y.z.AFactoryInterface</value></property>
>>>> </bean>
>>>> <bean id="user" class="com.whatever.User">
>>>>   <property name="myInterfaceFactory"><ref
>>>>bean="myinterface-factory"/></property>
>>>> </bean>
>>>></beans>
>>>>
>>>>Am I missing an existing way to do this? Is this worth adding to spring
>>>>as a convenience built-in, along the lines of
>>>>        
>>>>
>TransactionProxyFactoryBean?
>  
>
>>>>
>>>>        
>>>>
>>    
>>

[Springframework-developer] Re: Re: Re: Creating Application Context getBean 'closures'

From: roger h. <apo...@sn...> - 2003-12-08 17:43:52

Colin

Yes sorry about the package naming - that's because the 'other' approach I
tried, had to be implemented in spring.springframeork.beans.factory.support.

I very much doubt this code deserves including in the core.  I also don't
know how useful it might be to other people, but it would obviously be very
quick to re'package' if anyone can think of an appropriate home.

Roger


"Colin Sampaleanu" <col...@ex...> wrote in message
news:3FD...@ex......
> Hey Roger,
>
> This is quite interesting. I started implementing something (fairly
> similar to your approach, although more simplistic)  but ran out of
> time, so in the meantime just kept on using BeanFactory.getBean in the
> couple of cases where I would have used this.
>
> I'll take a look at this in more depth on the weekend. This is
> definitely useful to people using Spring. The only real question is if
> it's useful to enough people to warrant including in the core, given the
> size.
>
> Regards.
> Colin
>
>
> roger holbrook wrote:
>
> >Colin,
> >
> >A long delayed response, but just in case this topic may still be of
> >interest to you...
> >
> >I played around with a couple of ways of doing this, & have attached a
> >straight forward implementation of the initial scheme that you first
> >outlined, ie: given the following:
> >
> >public interface ITestBean_Factory1 {
> >    ITestBean getInstance();
> >}
> >
> ><bean id="testBean"
> >  class="org.springframework.beans.TestBean">
> > <property ...  >
> ></bean>
> >
> ><bean id="testTypedFactoryBean1"
> >  class="org.springframework.beans.factory.support.TypedFactoryBean">
> >  <property name="interface">
> >    <value>
> >      org.springframework.beans.factory.support.ITestBean_Factory1
> >    </value>
> >  </property>
> > <property name="target">
> >   <value>
> >     testBean
> >   </value>
> >  </property>
> ></bean>
> >
> >then a call to beanFactory.getBean("testTypedFactoryBean1") will return
> >dynamic proxy instance that implements the given typed factory interface.
> >
> >
> >FYI, the following is a short description of the 'other' approach I
tried:
> >
> >My first thought was that your initial scheme could be simplified by
using
> >a factory-interface attribute on the target bean definition, & doing away
> >with the need to have a separate typed factory bean, ie:
> >
> ><bean id="testBean"
> >  class="org.springframework.beans.TestBean">
>
>factory-interface"org.springframework.beans.factory.support.ITestBean_Facto
r
> >y1"
> >  <property... >
> ></bean>
> >
> >With a few tweaks in core classes, this all works fine - the only problem
is
> >that it just doesn't feel clean enough, ie: it requires the following
things
> >to change:
> >
> >- new factory-interface bean definition attribute
> >- new behaviour for beanFactory.getBean("&beanName"), since the typed
> >factory dynamic proxy instance returned, will no longer reliably be an
> >instance of FactoryBean
> >- small, but none the less, additional complexity to AbstractBeanFactory,
> >one more level of indirection to contend with
> >
> >On reflection, in this instance, end of idea ;)
> >
> >I would be interested to know how you ended up actually dealing with the
> >issue in practice ?
> >
> >Roger
> >
> >
> >
> >"Colin Sampaleanu" <col...@ex...> wrote in message
> >news:3FB...@ex......
> >
> >
> >>I was thinking of the first, with the idea that there is no Spring class
> >>dependency, and you achieve total inversion of control.
> >>
> >>Your second option would work of course. It does achieve the IOC aspect,
> >>but doesn't remove the Spring  (if the GenericFactory interface comes
> >>from Spring), and you would have to cast in the using bean instead of in
> >>the proxy. The IOC and hiding of other beans in the BF are probably the
> >>most important things anyway.
> >>
> >>(Of course, both of these do have a negative aspect, compared to just
> >>using BeanFactory directly; with BeanFActory the using bean can declare
> >>itself BeanFactoryAware and get the reference with no extra
> >>
> >>
> >configuration).
> >
> >
> >>roger holbrook wrote:
> >>
> >>
> >>
> >>>Hi Colin
> >>>
> >>>Your thoughts on implementation hiding have got me thinking...
> >>>
> >>>But before I test anything could you clarify the sort of usage you
> >>>have in mind - specifically, would it make a difference whether the
> >>>implementation hiding factories were themselves 'typed' or 'generic'
ie.
> >>>
> >>>
> >do
> >
> >
> >>>you want / need the type of the factory to vary with the type that it
> >>>generates, or would a single generic factory be adequate ?
> >>>
> >>>In other words, does one or other of the following make a difference ?
> >>>
> >>>1.  Typed interface factories:
> >>>
> >>>interface MyFirstInterfaceFactory {
> >>>   MyFirstInterface getInstance()
> >>>}
> >>>
> >>>interface MySecondInterfaceFactory {
> >>>   MySecondInterface getInstance()
> >>>}
> >>>
> >>>
> >>>2.  Generic interface factory:
> >>>
> >>>interface GenericFactory {
> >>>   Object getInstance()
> >>>   Class getClass()
> >>>}
> >>>
> >>>where different instances of GenericFactory will return from getClass()
> >>>
> >>>
> >the
> >
> >
> >>>relevant type, MyFirstInterface, MySecondInterface etc..
> >>>
> >>>I'll try & test something tomorrow
> >>>
> >>>Roger
> >>>
> >>>
> >>>
> >>>"Colin Sampaleanu" <col...@ex...> wrote in message
> >>>news:3FB...@ex......
> >>>
> >>>
> >>>
> >>>
> >>>>I think I may be missing something, but I think it's desireable to be
> >>>>able to create in an easy fashion, 'closures' which encapsulate
getting
> >>>>a bean from a bean factory.
> >>>>
> >>>>Say you have an interface
> >>>> interface MyInteface {   ... whatever }
> >>>>
> >>>>and you have a factory Interface
> >>>> interface MyInterfaceFactory {
> >>>>   MyInterface getInstance();
> >>>> }
> >>>>
> >>>>And you have a user of the factory, who you'd rather have no knowledge
> >>>>of Spring, which is why he's using the factory instead of calling
> >>>>getBean himself:
> >>>>class User {
> >>>> MyInterfaceFactory _myfac;
> >>>> public void setMyInterfaceFactory(MyInterfaceFactory myfac) { _myfac
=
> >>>>myfac; }
> >>>> public void someMethod() {
> >>>>   // need a new instance of MyInterface to work with
> >>>>   MyInterface myint = _myfac.getInstance();
> >>>>     ...
> >>>> }
> >>>>}
> >>>>
> >>>>Now I have a bean factory
> >>>><beans>
> >>>> <bean id="myinterface" singleton="false"
> >>>>class="com.whatever.MyInterfaceImpl">
> >>>> </bean>
> >>>> <bean id="user" class="com.whatever.User">
> >>>>   <property name="myInterfaceFactory"><ref
> >>>>
> >>>>
> >bean="xxxxxxxx"/></property>
> >
> >
> >>>> </bean>
> >>>></beans>
> >>>>
> >>>>now, as per the above, context.getBean("myinterface") is already a
> >>>>factory for objects implementing MyInterface. But I don't want the
User
> >>>>object to know anything about contexts. And I'd rather not create an
> >>>>actual object that implements MyInterfaceFactory. It seems like a
waste,
> >>>>since all I am doing here is trying to create a level of indirection,
> >>>>and I already have a factory inside the context itself, and I may want
> >>>>to use this approach in 30 different places, just to add a level of
> >>>>indirection in creating new objects.
> >>>>
> >>>>So what I think is needed is some variation of ProxyFactoryBean (but a
> >>>>separate class), which given a target bean (which is itself a
factory),
> >>>>and a factory interface having a method with no args which returns a
> >>>>certain type, creates on the fly a new class implementing the factory
> >>>>interface, which will just use the target factory bean to actually
> >>>>supply the instance. So the bean def above would become:
> >>>><beans>
> >>>> <bean id="myinterface" singleton="false"
> >>>>class="com.whatever.MyInterfaceImpl">
> >>>> </bean>
> >>>> <bean id="myinterface-factory"
> >>>>
> >>>>
> >>>>
> >>>>
> >>>class="org.springframework.whatever.XXXX">
> >>>
> >>>
> >>>
> >>>
> >>>>   <property name="targetBean"><ref bean="xxxxxxxx"/></property>
> >>>>   <property
> >>>>name="interface"><value>x.y.z.AFactoryInterface</value></property>
> >>>> </bean>
> >>>> <bean id="user" class="com.whatever.User">
> >>>>   <property name="myInterfaceFactory"><ref
> >>>>bean="myinterface-factory"/></property>
> >>>> </bean>
> >>>></beans>
> >>>>
> >>>>Am I missing an existing way to do this? Is this worth adding to
spring
> >>>>as a convenience built-in, along the lines of
> >>>>
> >>>>
> >TransactionProxyFactoryBean?
> >
> >
> >>>>
> >>>>
> >>>>
> >>
> >>
>
>
>
>
> -------------------------------------------------------
> This SF.net email is sponsored by: IBM Linux Tutorials.
> Become an expert in LINUX or just sharpen your skills.  Sign up for IBM's
> Free Linux Tutorials.  Learn everything from the bash shell to sys admin.
> Click now! http://ads.osdn.com/?ad_id=1278&alloc_id=3371&op=click