policy interception
Interception class拦截器还可以用Policy 以一种更高效的方式进行加载使用。拦截和被拦截对象的关系常见的是一对多,因为被拦截的主要是对象的方法所以有时数量巨大。Policy具有定义匹配模式的功能,所以可以简化这一匹配工作,不用一一添加被拦截对象。匹配的模式有
• Assembly name
• Namespace
• Type
• Tag attribute
• Custom attribute
• Member name
• Method signature
• Parameter type
• Property
• Return type
有时候在Interception class中也可进行一些过滤,这部分工作也可以移转到这里进行。我猜想这里的效率更高一些。不过两处目的不同,不可一概而论。
照例,显示被拦截对象
从代码中可以看到使用了PolicyInjectionBehavior指出Policy行为,打了一个行为桩。后续往容器中添加定义的CallHandler。模式匹配的定义也在这里。简单期间我选用了MemberNameMatchingRule,选中Msg方法。这里的方面名也可以加入*号作为通配符。
运行的结果和之前用InterceptionBehavior的方法的结果是一样的。比较感兴趣的是Order参数的行为,因为Fsharp在这里不是直接引用接口中的Order属性。这一层多余的层次会不会影响到顺序的表现呢?
我添加了一个CallHandler并切断了包括前一个接口实现时,往后续Pipe传递的通路。
随后在调用时增加一块,切换 order顺序
通过实现
可以简化Policy的注册流程,之前如果说是大面积轰炸的话,现在则是精确战术打击。
使用时不用注册policy了。
以上
Interception class拦截器还可以用Policy 以一种更高效的方式进行加载使用。拦截和被拦截对象的关系常见的是一对多,因为被拦截的主要是对象的方法所以有时数量巨大。Policy具有定义匹配模式的功能,所以可以简化这一匹配工作,不用一一添加被拦截对象。匹配的模式有
• Assembly name
• Namespace
• Type
• Tag attribute
• Custom attribute
• Member name
• Method signature
• Parameter type
• Property
• Return type
有时候在Interception class中也可进行一些过滤,这部分工作也可以移转到这里进行。我猜想这里的效率更高一些。不过两处目的不同,不可一概而论。
照例,显示被拦截对象
open System
open System.Collections
open Microsoft.Practices.Unity
open Microsoft.Practices.Unity.InterceptionExtension
type ITenantStore =
abstract member Msg : unit->unit
type TenantStore() as x=
//do printfn "new TenantStore %A" (x.GetHashCode())
abstract Msg : unit -> unit
default x.Msg() =
printfn "Hello, it's TenantStore"
abstract Dispose : unit -> unit
default x.Dispose() =
printfn "TenantStore hase been cleaned"
interface ITenantStore with
member x.Msg() =
x.Dispose() //测试Virtual Method中拦截内部方法的结果
x.Msg()
interface IDisposable with
member x.Dispose() =
x.Dispose()
声明实现ICallHandler接口的类。此处有一个地方需要注意。还是Fsharp语言的特性,无法隐式对接口进行转换。Fsharp的接口都是显示声明实现。如果不在类中暴露Order属性,在注册CallHandler的时候会提示缺少Order属性。好在ICallHandler接口比IInterceptionBehavior功能少,附加的点也就这一个。
type LoggingCallHandler() =
let WriteLog message =
printfn "From the logging interceptor: %A" message
member val Order=0 with get,set
interface ICallHandler with
member x.Invoke((input:IMethodInvocation), (getNext:GetNextHandlerDelegate)) =
String.Format("Invoke method {0}:{2} at {1}", input.MethodBase, DateTime.Now.ToLongTimeString(), input.MethodBase.DeclaringType) |> WriteLog
let result = getNext.Invoke().Invoke(input, getNext)
match result.Exception with
| null ->
String.Format("Method {0}:{3} returned {1} at {2}", input.MethodBase, result.ReturnValue, DateTime.Now.ToLongTimeString(), input.MethodBase.DeclaringType) |> WriteLog
| _ ->
String.Format("Method {0} threw exception {1} at {2}", input.MethodBase, result.Exception.Message, DateTime.Now.ToLongTimeString()) |> WriteLog
result
member x.Order with get()=x.Order and
set(v) = x.Order <- v
下面是具体的注册调用
using(new UnityContainer())(fun ctner->
ctner.AddNewExtension<Interception>() |> ignore
ctner.RegisterType<ITenantStore, TenantStore>(new InterceptionBehavior<PolicyInjectionBehavior>(),
new Interceptor<VirtualMethodInterceptor>()) |> ignore
let first = new InjectionProperty("Order", 1)
ctner.Configure<Interception>().AddPolicy("logging")
.AddMatchingRule<MemberNameMatchingRule>(new InjectionConstructor([|"Msg"|], true))
.AddCallHandler<LoggingCallHandler>(new ContainerControlledLifetimeManager(), new InjectionConstructor(),first) |> ignore
showregistrations ctner
let t = ctner.Resolve<ITenantStore>()
t.Msg()
)
从代码中可以看到使用了PolicyInjectionBehavior指出Policy行为,打了一个行为桩。后续往容器中添加定义的CallHandler。模式匹配的定义也在这里。简单期间我选用了MemberNameMatchingRule,选中Msg方法。这里的方面名也可以加入*号作为通配符。
运行的结果和之前用InterceptionBehavior的方法的结果是一样的。比较感兴趣的是Order参数的行为,因为Fsharp在这里不是直接引用接口中的Order属性。这一层多余的层次会不会影响到顺序的表现呢?
我添加了一个CallHandler并切断了包括前一个接口实现时,往后续Pipe传递的通路。
type LoggingCallHandler() =
let WriteLog message =
printfn "From the logging interceptor: %A" message
member val Order=0 with get,set
interface ICallHandler with
member x.Invoke((input:IMethodInvocation), (getNext:GetNextHandlerDelegate)) =
String.Format("Invoke method {0}:{2} at {1}", input.MethodBase, DateTime.Now.ToLongTimeString(), input.MethodBase.DeclaringType) |> WriteLog
let result = input.CreateMethodReturn(0) //break the pipe line
//let result = getNext.Invoke().Invoke(input, getNext)
match result.Exception with
| null ->
String.Format("Method {0}:{3} returned {1} at {2}", input.MethodBase, result.ReturnValue, DateTime.Now.ToLongTimeString(), input.MethodBase.DeclaringType) |> WriteLog
| _ ->
String.Format("Method {0} threw exception {1} at {2}", input.MethodBase, result.Exception.Message, DateTime.Now.ToLongTimeString()) |> WriteLog
result
member x.Order with get()=x.Order and
set(v) = x.Order <- v
type LoggingCallHandler2() =
let WriteLog message =
printfn "From the logging interceptor: %A" message
member val Order=0 with get,set
interface ICallHandler with
member x.Invoke((input:IMethodInvocation), (getNext:GetNextHandlerDelegate)) =
String.Format("Invoke method2 {0}:{2} at {1}", input.MethodBase, DateTime.Now.ToLongTimeString(), input.MethodBase.DeclaringType) |> WriteLog
//let result = getNext.Invoke().Invoke(input, getNext)
let result = input.CreateMethodReturn(0) //break the pipe line
match result.Exception with
| null ->
String.Format("Method2 {0}:{3} returned {1} at {2}", input.MethodBase, result.ReturnValue, DateTime.Now.ToLongTimeString(), input.MethodBase.DeclaringType) |> WriteLog
| _ ->
String.Format("Method2 {0} threw exception {1} at {2}", input.MethodBase, result.Exception.Message, DateTime.Now.ToLongTimeString()) |> WriteLog
result
member x.Order with get()=x.Order and
set(v) = x.Order <- v
随后在调用时增加一块,切换 order顺序
using(new UnityContainer())(fun ctner->
ctner.AddNewExtension<Interception>() |> ignore
ctner.RegisterType<ITenantStore, TenantStore>(new InterceptionBehavior<PolicyInjectionBehavior>(),
new Interceptor<VirtualMethodInterceptor>()) |> ignore
let first = new InjectionProperty("Order", 1)
let second = new InjectionProperty("Order", 2)
ctner.Configure<Interception>().AddPolicy("logging")
.AddMatchingRule<MemberNameMatchingRule>(new InjectionConstructor([|"Msg"|], true))
.AddCallHandler<LoggingCallHandler>(new ContainerControlledLifetimeManager(), new InjectionConstructor(),second) |> ignore
ctner.Configure<Interception>().AddPolicy("logging2")
.AddMatchingRule<MemberNameMatchingRule>(new InjectionConstructor([|"Msg"|], true))
.AddCallHandler<LoggingCallHandler2>(new ContainerControlledLifetimeManager(), new InjectionConstructor(),first) |> ignore
showregistrations ctner
let t = ctner.Resolve<ITenantStore>()
t.Msg()
)
结果与设想一致。
通过实现
type LoggingCallHandlerAttribute(od:int) =
inherit HandlerAttribute()
let order = od
override x.CreateHandler(container:IUnityContainer) =
let result = new LoggingCallHandler()
result.Order <- order
result :> ICallHandler
可以简化Policy的注册流程,之前如果说是大面积轰炸的话,现在则是精确战术打击。
type TenantStore()=
//do printfn "new TenantStore %A" (x.GetHashCode())
abstract Msg : unit -> unit
default x.Msg() =
printfn "Hello, it's TenantStore"
abstract Dispose : unit -> unit
default x.Dispose() =
printfn "TenantStore hase been cleaned"
interface ITenantStore with
[<LoggingCallHandler(1)>]
member x.Msg() =
x.Msg()
interface IDisposable with
member x.Dispose() =
x.Dispose()
member x.NewMethod() =
printfn "just test from TenantStore"
使用时不用注册policy了。
using(new UnityContainer())(fun ctner->
ctner.AddNewExtension<Interception>() |> ignore
ctner.RegisterType<ITenantStore, TenantStore>(new InterceptionBehavior<PolicyInjectionBehavior>(),
new Interceptor<InterfaceInterceptor>())|> ignore
showregistrations ctner
let t = ctner.Resolve<ITenantStore>()
t.Msg()
)
以上