Functional Programming With C# Using Railway-Oriented Programming

[@MilanJovanovicTech]

Want to master Clean Architecture? Go here: bit.ly/3PupkOJ Want to unlock Modular Monoliths? Go here: bit.ly/3SXlzSt

[@ajirahman5289]

Already subscribed...

[@MilanJovanovicTech]

@@ajirahman5289 Because you are awesome 🔥

[@soverain]

Functional programming, yes! Very excited to see more on this topic!

[@MilanJovanovicTech]

I'm loving the community response, definitely going to make more videos 😁

[@aToa5241]

Fantastic video in such an interesting topic. Looking forward to your advanced videos on this - there is not a lot of content out there (even paid content) on functional programming in C#, and I think the benefits of using it are huge. There is the amazing book Functional Programming in C# by Enrico Buonanno (now with a 2nd edition), and a nice blog from Mark Seemann. However, at the end of the day, C# is not a functional-first programming language, and I have realised that either I move to another language, or I acept those imperative if statements. Your video shows a practical and clean way to turn an imperative method to a declarative one. Railway-oriented programming looks very interesting and practical. Thank you for sharing, Milan - another excellent video!

[@AboutCleanCode]

I know what you mean - on the other hand more and more functional concepts get introduced in C# (e.g. pattern matching, record types). In the mean time you may check out this video kzbin.info/www/bejne/f6C4pKx7rryKZs0 which closely related to what Milan just presented

[@aToa5241]

​@@AboutCleanCode Nice video, thank you for sharing. Looking forward to seeing your videos on Uncle's Bob Clean Architecture as well.

[@MilanJovanovicTech]

Thank you so much, I'm glad you enjoyed the video! 😁 I think we need to preach functional programming more as a community, because it will improve our OO design even.

[@devjonie]

Wow! Interesting concept. Thanks Milan

[@MilanJovanovicTech]

It's even more interesting when you add side effects (database calls) and branching on the Result value 😁

[@JochenZeischka]

Hi Milan, thanks for your efforts on this. It would have been nice however to give credit to Scott Wlaschin (F# for fun and profit). He invented the ROP term for this kind of functional composition. Also, it would have been better to take a different example than validation. Chaining functions which can fail would have been a good example. Validation however, requires a different composition, because you don't want to stop validating after the first failure, but instead return info on all failures. As you indicate, it indeed complicates the consuming code and this should not be underestimated. You did not touch the topic of composing the results, which is pretty important. Especially if some results arrive asynchronously ;-) But my main remark is that Scott really should have been given credit in this video. Thanks.

[@MilanJovanovicTech]

Although I have watched Scott's video, I wasn't at all aware he invented the term, so it was an honest miss on my end. As far as coming up with examples, I definitely want to make more videos on ROP so those will surely be covered at some point.

[@reinhardlackner1925]

​@@MilanJovanovicTech watch some more of Scott's videos or maybe Simon Painter's, then you would have known Scott coined the term. Anyways, If you want to go down the rabbit hole of functional C#, watch out for Simon's NDC talks and Enrico Buonnano's book Funktional Programming in C# (Simon's book is still in the making) Ah, and I nearly forgot to mention Paul Louth's library Language.Ext which were the goto functional library (if you didn't want to write everyting yourself) before Microsoft started to introduce more and more functional stuff into C#

[@yohm31]

@MilanJovanovicTech indeed, Language.Ext library offers all the Monad you are dreaming of :)

[@christianmarpert3844]

Hey Milan, fantastic introduction to FP. I started watching your DDD serious and then stumbled upon this one. To be honest, I am not yet deeply into C#, am rather programming in Dart for mobile. However, I use FP there to stream line exception handling. There is a well documented package called fpdart, accompanied by some blogs on functional programming. Looks like the Result type in your project corrresponds to the Either type in fpdart, where Left is by convention the Failure, and Right is the Success, whereby only one state is possible. Method chaining can also be done easily. Really enjoy your videos, as you explain the topic in a great way!

[@MilanJovanovicTech]

Yeah, the Result class is very similar to the Either monad!

[@christianmarpert3844]

monad? Scare us all! A monad is a monoid in the category of endofunctors! I got scared by that definition 😁, however its usecases made me calm down 😎

[@janovrom]

I was trying to wrap my head around monads and some actually useful example 👍 Thank you.

[@MilanJovanovicTech]

Thanks :) Here are a few more videos in the series: - kzbin.info/www/bejne/sKbcY51uq92aj8U - kzbin.info/www/bejne/d4ekY6CBfJmKhcU - kzbin.info/www/bejne/rHjOmKaaqtyhmZY

[@kevinaubuchon5979]

Good video. These functional programming concepts are difficult to get other team members to embrace though. Some appreciate them, others simply don't understand.

[@MilanJovanovicTech]

I know, I've had to fight with this myself. Some people will just never embrace FP and that's fine. But what you can do is slowly introduce some functional concepts to your code base that are helpful in general.

[@tinytownsoftware3837]

This is what I am doing as an architect. Simply introducing Maybe/Option and Result concepts into the codebase are huge improvements and they are not that hard to use.

[@marklord7614]

Enrico Buonanno makes the best case for functional programming in C# that I've seen. Hopefully the language will support it in a way that makes code more readable when it comes to handing multiple return types (discriminated unions).

[@MilanJovanovicTech]

I only wish we had real discriminated unions...

[@MaximilienNoal]

@@MilanJovanovicTech There's a nuget package that uses a source generator to add the full functionality of discriminated unions to C#. You just have to write [Union] on top of a record declaration.

[@krccmsitp2884]

Interesting and straightforward as always, keep going! 👍🏼

[@MilanJovanovicTech]

Thanks a lot!

[@tonycaesar7934]

Great video. I like the approach. It looks much cleaner and self-documenting. This is especially true when coupled with the Result and static Error approach. Thanks for sharing!

[@MilanJovanovicTech]

I left a few questions unanswered, such as returning multiple errors or handling async calls. Going to cover that in a future video.

[@manuelknorle1457]

A benchmark on this approach would also be very interesting. Which of the two implementations has the greater impact on performance?

[@MilanJovanovicTech]

Certainly the ROP approach has more allocations, but another benefit is that since ROP is stateless (FP is stateless by design) it's completely parallelizable. In any case, I don't worry to much about this. It's unlikely coding in a functional manner will be your performance bottleneck. That would be on the nanosecond level, which is insignificant compared to a cost of a network round trip.

[@svorskemattias]

You can probably make the result class a struct if it's not already.

[@nicholaspreston9586]

I like your Map function. IMO, it's better to have an even more generic extension method that performs a mapping after a null, rather than having to learn Automapper and have to write and manage a bunch of mappers. Been there, done that, and your Func is much better! src: /// /// Map a Source class to a Target /// public static TResult Map( this TSource source, Func map ) => map(source);

[@MilanJovanovicTech]

Ideally, the Value of a Result will never be null so no issues there

[@branislavpetrovic7486]

Great intro to ROP! Thanks!

[@MilanJovanovicTech]

Thanks a lot, Bane :)

[@branislavpetrovic7486]

@@MilanJovanovicTech Haha, I got a new nickname

[@AboutCleanCode]

That reminds me of the Option module from F# which is just a bit more enhanced as it applies full set of "filter-map-reduce" (LinQ style) functions to the Option type which is modeling existence and absence of values ...

[@MilanJovanovicTech]

All of these methods can be easily coded, and there are libraries out there with rich features for ROP

[@pilotboba]

It's clever code but I'm not sure it's less code always. Depends on where you apply it. Also, as you said, it quite reduces the understanding of the code. You could probably have written the original code with pattern matching, made it a single expression and it would be just as readable and understandable as well as made into an expression. Also, small niggle. Linq is not pronounced 'Lin Queue' with two syllables it is pronounced with one syllable like the word LINK. Like the character's name in Legend of Zelda. :) Keep up the fun videos.

[@MilanJovanovicTech]

I don't think it reduces understanding once you understand the concept - which is simple. I would argue it even improves the ability to reason about a piece of code.

[@DanimoDev]

I actually have a similar solution to this in my own projects, although I didn't know it actually had a name already, I just dubbed it chain validation. I quite like this approach personally as you can just chain call validations and it is neater. I also like the result wrapper and will definitely include this, mine doesn't have this and instead throws an exception when a condition fails (I suppose in this context you would say it gets derailed), but your solution is better.

[@MilanJovanovicTech]

Check out FluentResults for a nice library

[@friedrichhayek4862]

This feels like what you must do in a common Haskell program.

[@MilanJovanovicTech]

Never tried Haskell 🤷‍♂️

[@chrisbaker5284]

Hi Milan, I've watched many of your videos and I do enjoy them. I would like to point out that you said you finished with less lines of code when in fact, you finished using more than 3 times as much code. (The original method had, effectively, 7 lines of code if you removed the blank lines and the superfluous braces). I am not a fan of using functional styles in an OO language (I don't include LINQ as it is effectively a separate language within a language). I admit that your code is more resuable though, but it's arguable as to how much reuse you will achieve. Reuse could be achieved in an imperative style too. I worked for a company where they had taken this style of programming to the nth degree and it was a mess of functions the all did similar things but were meant to be used in differnt situations, everything had similar names wrote bucket loads of code to effectively do null checks and turned the whole thing into a framework. It was difficult to follow and there was no documentation. I know what you may be thinking, that was an example of bad coding style functional or not, well maybe, but these things tend to get abused. IMHO if one wants to use a functional style of programming, then I would recommend F# not force an OO language to be Functional. Just my opinion Milan, please do keep up the great work that you are doing.

[@MilanJovanovicTech]

This is the same argument as trashy OOP code. 😅 It's not the paradigm, it's the engineers, in my opinion. There's nothing inherently wrong with doing FP in C#, except you need to write some helper methods in the background to make it work.

[@chrisbaker5284]

@@MilanJovanovicTech It is true, bad code is bad code. I just think that one could inherently end up with convoluted and more complex code when forcing an OO language to a Functional paradigm. That's all.

[@hugotachoires1721]

Nice video, like every time ! Keep up

[@MilanJovanovicTech]

Thanks, will do!

[@ajirahman5289]

Thanks for the demo. Waiting for new videos. Could you share how can do xunit for this railway programming please. Could you please share an intro about what version of c# and the core.

[@MilanJovanovicTech]

Sure, I can make a short video about writing tests for this. All of the videos are C# 11 / .NET 7 going forward.

[@UristMcFarmer]

Well, other than the cons you listed, I disagree with everything you've said. However, I don't have the wherewhithall to discuss the finer points right now. I just want to say that if one is truely committed to writing Functional code in .NET, one should probably switch to using F# so they don't have to constantly write the boiler plate backing needed in C#

[@MilanJovanovicTech]

That's fine, we don't all have to agree 😁 I'd stick with C# for other reasons than functional programming though

[@amitconnect]

Good one...liked to learn more about c# functional programming...will you please share some more insight into this?....liked your work....love to learn more from you....thanks 👍

[@MilanJovanovicTech]

Definitely making more videos

[@relhimp]

Two chars: Rx

[@MilanJovanovicTech]

There's Rx.NET

[@duongphuhiep]

when you (developers) have to write a "forest" of "if" statements (for example an Authorization service which would have to check a lot of different rules and conditions). With time, we will add more rules (so more "if") in the forest and our code will become harder to maintain and to "unit test". I think that the "Railway programming" pattern might solve this problem. What do you think?

[@MilanJovanovicTech]

I see ROP as a way to get rid of exceptions - or express failure in a functional way The if statements are still "there", they're just abstracted behind the Either monad (Result)

[@kevinaubuchon5979]

Great video.

[@MilanJovanovicTech]

Glad you enjoyed it

[@autograf89]

Thank you for the video Milan! ROP looks great with linear workflows. Could you elaborate on more complex cases where an object requires to validate multiple parameters on creation?

[@MilanJovanovicTech]

Absolutely, there are solutions for that also!

[@diegodelafuente1739]

¡Gracias!

[@MilanJovanovicTech]

Thank you very much, Diego, I really appreciate it! :)

[@mohammedalbosifi3975]

Niiiiiiiiooccccccccccccce.... Thank you

[@MilanJovanovicTech]

Most welcome 😊

[@calblui]

Good class,tks!

[@MilanJovanovicTech]

Thank you, glad you liked it 😁

[@JustMe-gp4so]

Hello, Milan. Great example, i like it. But i have some thoughts: That example works with primitive conditions. So what if we need to process some async assertions that works with db(for example)? I heard about fluent validation, it allows you to inject you service in ctor, write your rule and call MustAsync or CustomAsync. What you need to do (or change)in that scenario with extensions to archive that goals?

[@MilanJovanovicTech]

Yes, it's possible. I'll record a video soon talking about that!

[@mahmoudalaskalany]

Great One Milan Keep It Up and can we use something like this to create automated workflows

[@MilanJovanovicTech]

What do you mean by automated workflows?

[@mahmoudalaskalany]

@@MilanJovanovicTech I mean something like (Elsa workflow framework) Can we use this way to create something like this or it is different aproach

[@scott98390]

Still waiting for more videos in this series

[@MilanJovanovicTech]

Check out Function Composition video

[@MilanJovanovicTech]

And one more is coming in a week

[@scott98390]

@@MilanJovanovicTech will do!

[@majormartintibor]

Good stuff. Have you ever used the FluentResult library?

[@MilanJovanovicTech]

Heard about it, didn't try it. I like coding some stuff on my own. Maybe I should consider creating my Nuget package for ROP 😅

[@majormartintibor]

@@MilanJovanovicTech your approach def has the advantage that you get a deeper understanding of the concept instead of just blindly using something.

[@Brendan2Alexander]

Love the railway approach here. Tho I find these fluent statements hard to debug. Is there a way to step thru each step?

[@MilanJovanovicTech]

Debugging is a bit tricky, yeah. But you can step into each function just fine. The bigger issue is you don't have local variables to check intermediate values 😅

[@tore28]

You can capture each step into a variable of Ensure and see the state clearly while debugging. Simon Painter shows this in one of his talks, was it in Ndc London?

[@MilanJovanovicTech]

Could you share the link? Watched many of his talks, but not sure what you're referring to

[@tore28]

​@@MilanJovanovicTech Sorry, I cannot recall. There is some tips on page 152 of "Functional programming with C#" Simon Painter wrote in 2023. I see that each step is stored into a local variable instead of chaining. An alternative would be "Tap" or "Tee" method that can output information of the intermediate state. For example outputting state to the console. I am still learning FP, it is so elegant pattern in C# and such a joy to learn more of.

[@svorskemattias]

For anyone who would like to dig deeper into this I highly recommend scott wlaschins content. His book "domain modeling made functional" and site Fsharpforfunandprofit contains a huge amount on the topic. He also has very nice talks on youtube. Ive succssfully implemented many of the patterns he describe in my C#-projects.

[@MilanJovanovicTech]

Scott is the man :)

[@prabhpahul]

Hi, nice video, do we have github repository? To follow your videos, specifically interested in Result class, and Domain Error implementations, thank you again.

[@MilanJovanovicTech]

I only share the code with my Patreon supporters, but I have a lot of similar examples on my GitHub that you can take a look at :)

[@reinhardlackner1925]

watch some of Simon Painters NDC talks - he's utilizing functional C# quite a bit and currently writing an O'Reily Book on this topic

[@maynardpaye]

seems similar to fluent validation?

[@MilanJovanovicTech]

It's all functional programming in some form

[@krccmsitp2884]

At least both use the concept of method chaining, but ROP in special or method-chaining in general don't need to stop there. You can apply either on various problem areas.

[@alexalexander3252]

Milan, I would appreciate if you clear one point for me. As far as I understand, we write extension methods for classes we can not change. And here you write an extension method for Email class and then edit this class. What about O from SOLID? Isn't it easier to write the method inside this Email class? I am a newbie if anything :)

[@MilanJovanovicTech]

Where is the Email extension method? 🤔

[@saifeddinebenromdhane7553]

Amazing, any idea on whether the code will be published on Github or not, i would love to dig into it. Thank you

[@MilanJovanovicTech]

Patreons only, as usual 😁 But in the spirit of giving, let me see if I have any open source examples

[@MilanJovanovicTech]

I think you can get started here, the Controllers follow ROP: github.com/m-jovanovic/event-reminder/tree/main/EventReminder.Services.Api/Controllers The MediatR handlers don't unfortunately, you'll have to figure it out 😅

[@saifeddinebenromdhane7553]

I do appreciate both, your good work and giving spirit, Thank you.

[@debtpeon]

A switch would be easier to write and debug.

[@MilanJovanovicTech]

Easier to debug, probably. Functional vode is easier to write

[@BlazorPlate]

Is it thread-safe?

[@MilanJovanovicTech]

Yes

[@krccmsitp2884]

Since the shown implementation uses pure functions, absolutely.

[@friedrichhayek4862]

Why don't just use Haskell. Where all of this is mandatory.

[@MilanJovanovicTech]

Because I'm good at C#

[@JeffChentingwei628]

Is it possible to do with async/await in ROP ?

[@MilanJovanovicTech]

Yes, I'll show that in of the videos!

[@JeffChentingwei628]

@@MilanJovanovicTech amazing! Thanks !

[@OrgesKreka]

What keyboard are you using ?

[@MilanJovanovicTech]

Logitech G413 Carbon

[@OrgesKreka]

Thank You

[@cocoscacao6102]

F# video?

[@MilanJovanovicTech]

Unlikely, I don't develop with F#

[@angelldark6426]

Mr.Milan, do you have this project on Gidhab?

[@MilanJovanovicTech]

Nope