This article show how to make code more clear, easier to write and more extensible in a generic way. Use of Option type in C# in combination of Func delegates
Introduction
Problem to solve: If you look at a lot of code in C#, the code must be read from top to bottom, following all the If
, Else
, Try
and other branches to really understand what is happening. Furthermore, a lot of this code is basically technically, such as checking on null
-values. I wondered if you could write code that reads from left to right, and is in principle not-technical. It reads like a series of operations that makes sense as such. A ‘similar’ concept is known in F# as pipelining. In F#, it looks like:
Let squared =
Numbers
|> List.filter isOdd
|> List map square;;
The object Numbers are pipelined through the functions that follows. In C#, I wanted to do something like this:
[SomeObject].AndThen(isValid)
. AndThen(Save)
. AndThenIfElse(Do_this_If_ThisIsTrue, Do_this_IfOtherwise)
Etc. In this case, the SomeObject
is passed to function IsValid
and then to Save
and then to the Do_this_If_ThisIsTrue
, which function can return 'false
', in which case SomeObject
is passed to Do_this_ifOtherWise
. The beauty of this chaining is that we can add any number of AndThen
, AndThenIfElse
as we want.
You could do this years ago using delegates. Delegates are eventually not generic. Now we can use these, but we also want to handle Error handling and null
-alike code. Tomas Petricek and John Skeet Real-World Functional Programming comes with the solution. The secret is in essentially two things: the Option type and the using of essentially only one fixed interface: (func
of Option
of T
, Option
of T
), meaning you pass in a function that receives an Option
and returns one. This interface can be chained or pipelined as much as you want.
The Option
Type is a well-known Type in F#. An Option
can be None
or Some
. Some wraps an actual business object. Continuations is the concept in which you pass the result of a function to another function.
What if you combine those two concepts? There is a lot of theory involved, but I won't bother you with it. If you wrap an object in an Option
and then use Continuation
, you can keep on chaining as shown in the above AndThen
code. The Option
object hides the actual type of the object inside it.
Inside the Option
can be anything. Using the Option
-concept, we can let behave a strict typed language in a generic way.
For continuation, you can use Lambdas, but then the typing won't let you implement these in a generic way.
This article is about implementing Continuation in a generic way.
The Option
type is an abstract
class with two concrete implementations: None
and Some
. None
ends the chaining, Some
encapsulates the [SomeObject]
. A function must first get the encapsulated object, do some processing on it and return either a None
or a Some
with the encapsulated object. The Option Some
and None
classes are generic!
To this Option
type, I added some functions, such as the above AndThenIfElse
. If you look at the parameters, you see that you pass the following two parameters: Func
of Option
of T
, Option
of T tryFirst
, Func
of Option
of T
, Option
of T tryNext
, meaning you pass in two functions.
I added:
AndThen
(continuation). If the previous function returns a Some
, then continue with the following function AndThenCheck
. You pass in as the first parameter a function that takes an Option
and returns a Boolean
. If this Boolean
is true
, the second parameter is executed, which is a continuation. If not true, return None. AndThenCaseCheck
. You pass in a list of AndThenCheck
. If a check returns true
, the accompanying continuation is executed. If none of the checks returns true
, a None
is returned. In the accompanying example, I use lambdas to compose this list. Funcs can be replaced by lambdas and in this case, I thought it was a good idea. AndThenIfElse
. You pass in two functions. If the first return None
, the second continuation is executed.
The other interesting thing to note is that the above pipeline of continuations is not executed until a call to Exec()
. The reason to do this, is because Exceptions are handled in the Exec()
call only, and the continuations stop with the first None
returns. The pipeline is in fact an arraylist of MultiCastDelgates
that can be serialized. The option/Continuation concept can be used in other scenarios as well such as postponing execution by saving until it’s triggered in time.
I added a concrete example of how to use this in the real world. The code is not doing much, but gives you a good idea how you can use it. Looking at the code below (ClientController
), you see the real benefit:
someClientData.AndThen(isValid)
.AndThen(ClientBusiness.Save)
.AndThenIfElse(ClientBusiness.CheckOnDustin, ClientBusiness.CheckOnNotDustin)
.AndThenCheck(ClientBusiness.ANameLengthCheck, ClientBusiness.GetLastOneInLocality)
.AndThenCaseCheck(ClientBusiness.SomeCaseListOnProfession())
.Exec();
Does the code read from left to right and does it read as a series of chained operations? Does it 'hide' technical handlings in favor of business handlings? What do you think? Let me know.
History
- 23rd August, 2011: Initial version