As I am trying to learn more and more about this F# thing and functional programming, in general, I realized that there was a way to cleanup my code from last post to make it more readable. By taking a hint from Scott Wlaschin's post on Railway Oriented Programming, I realized I could pull a lot of the pattern matching out of my code (where it makes sense) and move it into a bind function. Seeing it in my code reminded me a lot of factoring out common terms in math. In this case, we're factoring out the initial check of Some/None.
(As an aside, Scott's website F# for Fun and Profit is a fountain of knowledge and is worth spending a lot of time studying.)
So, we need a function that will do the Some/None checking for us and, since this is the common name everybody uses, we'll call it bind:
Edit: Thanks to Liam McLennan for pointing out that Option.bind can (should) be used instead of writing our own. (Great to have others reviewing my thoughts and it is much appreciated!)
This function takes a function and an Option<'a> as inputs. If the Option<'a> is Some, we take the value of it and bind it to the function f. If the Option<'a> is None, we just return None and our function f is never called.
I also like the infix operator that Scott uses and decided to use the same one for this version of bind. I'm keeping it all in the same module since it's specific to Option<'a> and fairly trivial to write out where it is needed. I'm sure in a larger project it would make sense to wrap all of these common functions into a helper module, but it is small enough.
Edit: And this is how the infix operator can be written using Option.bind:
So we can compose two functions together and into a third. Our new function takes whatever the input is for f1 and binds the output from f1 to f2, thus returning whatever f2 returns. If f1 returns None, then f2 is never called and None is returned from our new function. As such, that means f2 needs to return Option<'a> to keep the signature legal.
But why?
Why would we introduce this to our code? Like when we factor out terms when solving a math problem, we are (usually) reducing the complexity of the problem. If we take one of our original functions, we see that we have to test Option<'a>.isSome and Option<'a>.isNone. Not only do we have to test this in this function, we have to test it in all of them. When we factor this out, we reduce the complexity of our code and reduce the amount of repetitive tests. So, one of our old function:
becomes
In fact, the more functions we have where we're checking for Some/None, the more valuable our bind function becomes.
Cleaned-up code
So here is the same code as the last post but with the new bind function and operator.
Suggestions?
I am always interested in hearing suggestions on how to make my code even better. Have something I could improve? Let me know!
Minor Update (26 Jan 2014)
I realized, after the fact, that I did not use the Simple.Web extension method to write the header. By forgetting this, the header would never actually get written because a NullReferenceException would be thrown. (The context.Response.Headers is initially null.) So, instead of instantiating the dictionary or anything crazy, we just use the extension method which takes care of everything for us. The good news is, this is actually much cleaner than the previous code.
No comments:
Post a Comment