Language is not important

Some people like to make fun of me and like to call me a hipster, especially ever since I've started working with Functional Programming languages. The longer you're working with them, the more opinionated you get about the ways to build the software, and the more you start missing certain features that generally have very little to do with Functional Programming (such as pattern matching and list comprehensions), but are extremely useful in a day-to-day life.

I've been doing Clojure since 2011 and have started looking at Erlang and Haskell a couple of years ago. Up till the date, I still haven't written anything reasonably large in Erlang (just a toy implementation of Cassandra CQL binary protocol, but that's still rather small).

With Haskell, I've done quite some things related to FFI (not sure whether that counts more for C than for Haskell though), started writing a pet project , which will take a long time to finish and polish. I've even accidentally managed to send two pull requests to GHC. But I'd never call myself even an intermediate Haskeller. I'm just an amateur and do it for sheer fun.

Having drunk FP Kool Aid

On Twitter, someone (if you know who it was, please send me a link) posted his (in that particular case I'm quite sure I got that gender thing right) impressions about writing a first reasonably-sized Haskell project. And it was a mixture of fun and pain. I can say that my experience was rather similar.

When you just start working with Haskell, you start using (and abusing) the do-notation and the Scary Monads start look rather imperative. Many people (especially ones that aren't planning to build a great Haskell career), stop there and decide they've seen it all, and know the Monads.

If you stick around Haskell for long enough, you realize that you don't really need Monad everywhere, and the majority of things can actually be accomplished with whatever Functor and Applicative offer. You don't feel stuck when you see the type signatures, and you start really enjoying that making a massive refactoring in your project can be done in relatively small amount of time, and in majority of cases, "when it compiles, it works".

It's hard to say how much time it takes to get from the first to the second stage. For some people, it's a couple of days. For some, it may be months. For me, it was rather long. I felt myself lost in front of all those Monads and wanted to give up pretty much every day. When I tried asking people about help, I was either getting answers that were rather generic and didn't make sense for me (probably made a lot of sense for the person on the other side of the wire). Although once I got an awesome 30-minute lecture on Monad Transformers on IRC, and I will never stop being thankful for that.

In general, it's close to impossible to just sit and start programming Haskell. The syntax is simple, but you have to catch up on the theory. But I now tend to think that the concepts behind all the cool things in Haskell are much more important than Haskell itself for me.

For example, I now know that it's possible (and even understand how) to write programs that have no nil values, or exceptions. Reading a tutorial online will maybe help you to stop fighting the thought of the impossibility of such, but won't really make you sit and stare at ceiling for 20 minutes, realizing how cool that is and resenting not knowing it for your whole life. But both these things are only relevant for me on a conceptual level. I can build them into any language I use. I don't need Haskell for that.

hears raging voices saying it's impossible without an awesome type checker

Category Shmategory

I knew that Haskell is also a gateway drug into the world of Abstract Algebra and Category Theory. So I decided to start with the books in my native language and have gotten myself familiar with graphs, isomorphisms, binary operations, all the laws, theorems, axioms and so on, carrying a book with me all the time, reading proofs while sipping coffee during the short break.

It's a great fun, I must say. And many things were extremely interesting. So as I started feeling that I understand the algebraic concepts and should move towards the roots of Haskell, I found that I can understand everything there, too.

Except for I didn't understand why the nested let bindings should take 6 pages of formulas and proving to explain, and even the simplest types are expressed in the terms so abstract that it felt necessary to google for that GitHub repo Russian government banned GitHub over.

I must admit that I didn't do Ph.D. in either of those topics. It's all been just my hobby, and I will some day continue studying Abstract Algebra for fun. But for now, it all feels much too abstract and I see absolutely no application for it in my life.

Financial Pyramid

For all those months of learning, when I felt lost and every line of code was taking minutes and sometimes hours to write, I kept listening to the people who were saying that Haskell is a great investment, and I should keep pressing on, because it will pay me back tenfold.

But I'm a Soviet Kid. I still can remember that MMM story from after-soviet time, when you were asked to invest for a greater promise, so I'm very skeptical when, after a long time of working with the language, and being very proficient with it, being able to write whatever I want, I never feel certain or secure about what I am doing.

Maybe that has something to do with the Mysteries around Haskell and superhuman abilities of the people who do it, that I was imagining back when I just started. To be honest, I still feel this way. I have to close my eyes for a second when someone mentions endomorphism or bijection to tie it all up with the ongoing conversation. I just feel cheated, since in my head, by now I should have stopped investing and could already start collecting my dividends.

But since that's a rather philosophical post, I'm not putting any concrete information such as dates here. It doesn't matter how much I learned. It's been long enough for everyone I know to get bored with a subject, and long enough for me to start pondering about where all the time went. I can only speak for myself, and if it was different for you - I'm very glad.

But if you feel the same way I do, and, when looking around, only see the people who find those discussions Abstract Algebra exciting (and not tiring), and wonder if it all was worth the effort, be my guest. We can create a club of people who at least didn't stick around Abstract Algebra long enough to become dangerous.

Back to the Basic(s)

Fast forward all that time, it's still much easier for me to make a little web server in Clojure or Python, I still do most of juggling with data in Clojure REPL, and still enjoy writing Java and C code (depending on circumstances).

Do I think that the time was wasted? No, not at all. I think that the time was well spent. I've learned a whole bunch of things, and I'm still planning to continue writing my little database in Haskell. I just don't want to invest years of my life into a demanding technology.

Moreover, it's impossible to grasp what Functors, Applicatives and Monads (and all kinds of Semigroups, Monoids and other fancy things) are from the "Monad Tutorials". They kind of scratch the surface of usefulness from time to time, but in majority of cases they're either misleading or not really helping to understand the whole power of the concept and how to apply it in real life.

In that regard, learning Haskell helped more than anything else. You don't just think of a Monad as a "side-effect" or a "container/context", you think of it as a major building block of your application. And you can start applying the same concept in your Java / Ruby / Clojure code, or any other language. Without diving into it head first and spending enough time for it to click in your head, I don't believe you will be proficient enough to use it anywhere.

Learning things

It's always hard to choose what to learn next since there are so many things out there that are very interesting. Many people enthusiastic about Functional Programming Languages are ready to teach their peers, and invest their personal time into making it all bigger. But you have to always bear in mind: it's a lot of effort and work. And it will require effort not only from your side (as a teacher), but also from the people who learn, which is quite normal and applies to anything, starting from learning how to bike and up to quantum physics.

Most of the time, when people talk about using the cool stuff at work, they also mention some bonds or obligations: the one who brought FP in should be teaching and supporting the others. I have spent 2 years helping people learn Clojure, Haskell, and other FP Languages. And I can say that it takes a lot of time, and usually return does not excite you so much.

If you enter the World Of Haskell with the great enthusiasm, the opportunities are pretty much endless. You start with Haskell itself, and you can practice and master it for years and years. If you feel fancy, you can start learning Category Theory (or Homeopaty Type Theory) and Abstract Algebra. I'm still not sure whether you have to have a great mathematical mind in order to learn and understand it all (most likely yes), I've only scratched the surface.

Once I've been told by the Haskell Enthusiast that I want to reap what I saw too quickly, which I admit is true. But, unfortunately, there are too many things applicable in everyday life I'd like to learn, so I'm putting my 4b$7rac7 $7uff endeavor on a pause.

Language doesn't matter

After I've started doing Clojure, I thought I won't be ever able to program Java again, because Functional Programming and all these things, you know. But after doing both Clojure and Haskell (and many other languages in between), I can say that language doesn't matter that much. I now know where I can make mistakes, and write my Clojure code the way I could have at least some poor man's type checking, and write my Java code the way I could be at least remotely sure about what it does after it compiled.

It's really great that I can have IO encoded in the type signature, but that was never been a selling point for me. There are many other things I appreciate much more (like ADTs, Pattern Matching, Tuples etc). But they're generic and universal and do not belong to any specific piece of technology. I'm also sure that sooner or later many things will get better in mainstream programming languages. Whenever the humanity is ready for that.

Knowing concepts turned out to be much more important than being able to use a certain implementation (or syntax) for me. Of course, I will never be able to do Pattern Matching in Java that'd look at least remotely sane, but in the majority of cases I can also live with it. At least I can guarantee immutability when I need it, and I now know and understand very well how to work with immutable data structures without making your code look awful or complicated.

So it all kind of boils down to simple and primitive things: being able to verify that you haven't generally done anything stupid (typos, arity checks), being able to make sure something isn't going to be modified concurrently or invisibly, being able to compose things using Functions, using more explicit and less verbose syntax and having enough freedom to still use a more sensible language whenever it makes sense.

Grass is always greener on the other side

Other than that - the tools that you're using aren't that important. The most important stuff is going on in your brain. Despite the current hype and global love to immutability, the concept has been around for a long time. Pure functions were known for the longer time than computers existed. And trust me, there will be many new languages that will implement things in a better, more interesting way, and bring you ever-more fun. Although, most of the time, I do not buy the argument "older is better".

Except for chasing new languages won't get you anywhere. Yes, there will be a next big thing every two to five years. Yes, it's going to be extremely exciting to learn it all and check it out. But let's just stop for a minute and accept the fact that concepts are more important than a concrete implementation. Every implementation is flawed, and there will always be languages that do certain things better than yours.

For example, no one can still beat Erlang in Binary Pattern Matching, and the way Actors are implemented there. If you check out OTP - you'll want it under your tool belt every day.

There are people who prefer OCaml, Rust, Go (that happens, too, don't ask me), even Ruby, Python, Scala (another secret), and they'll be solving hard and interesting problems with these languages. And guess what - these people may be as conceptually proficient as you are. And they will certainly write some awesome software with the tools of their preference.

When the mathematician is reading papers, it's not important for him whether the others are using the same exact notation. People may use the most obscure notations for their works because they have to optimize for what they do. The same way it works in our industry. It will never (never say never) converge to the single, commonly accepted language (even if it's multi-paradigm and is running on JVM).

Keep pressing on

Don't take away a fun of learning new things from yourself, but try to always be skeptical about everything you've been promised. Learn the concepts, but don't become a fanatic, since you'll lose a contact with reality and won't be able to both learn and understand the things outside of something you made your comfort zone.

It's obviously not necessary to dive deep into Category Theory to work with Haskell, but somehow everyone who wants to be a part of the crowd is doing it. I thought that I'd rather invest same time into other things: algorithms, Linux kernel, statistics and linear algebra.

If it feels for you that this post is a rant - you're tuning yourself much too negatively. Or if it felt as if learning new languages isn't important for your career - no, it is extremely important.

In the end, my own takeaway from writing it all is quite simple: take it slow and take it easy. Don't try to match all the cool kids out there. It may be that people writing about all the smart things have to check in their little notebook if they've spelled it right. There will be always someone who knows it all better, but you should never feel intimidated by that. Also, don't feel bad that you can't use every fancy thing you've read about at your work: it's not necessary, and it's all happening and living in your head before it gets translated into code in some language.


If you liked this post, you may also like Show me your compiler and I'll tell you who you are post of mine.

If you're beginning with Haskell, you may like What monad type signatures look like in my imagination.

"Homeopaty" wasn't a typo. It was a joke. Also, real Homeopathy is spelled with th. Was inserted with the best intention of making the post easier to read, and never to make the actual theory unsound.

Published on December 8, 2014

If you like my content, you might want to follow me on Twitter to subscribe for the future updates!