My theory is that communicating abstractions is hard. If you work on your own, or in a (very) small team, you can come up with powerful abstractions that allow you to build amazing systems, quickly. However, sharing the underlying ideas and philosophy with new team members can be daunting. As systems grow, and mistakes are made, it becomes more and more likely that you run into serious problems.
This may also be why Java and similar object oriented programming languages are so successful for systems that have to be maintained for ages, by large teams of developers. There are but few abstractions and patterns, and it does not allow you to shoot yourself in the foot, nor to blow your whole leg off. Conversely, this may also be why complex frameworks, such as Spring, are not always so nice, because they introduce (too?) powerful abstractions, for example through annotations. It may also clarify why more powerful languages such as Scala, Common Lisp, Smalltalk, Haskell, etc, consistently fail to pick up steam.
Another theory is that not every developer is comfortable with abstract concepts, and that it simply takes a team of smart people to handle those.
Languages need a window of opportunity, and many of those squandered it.
Clojure won over Scala because at the time when people were loooking for an alternative JVM langauge, Clojure was more of a departure from Java and seemed to have better tooling (compile times and syntax support) than Scala.
Smalltalk and Common Lisp wasted their moment by not being cheap/free to people using micros in the 1980s.
Lisp, especially, very much wasted its moment with micros. The fact that no Lisper had the vision to dump a Lisp onto the bank switched micros (which makes GC really easy and useful) of the mid to late 1980s is a self-inflicted bullet wound. Lots of us hated doing assembly language programming but had no real alternative. This was a loss born of pure arrogance of Lispers who looked down on those micros as not being "real machines".
I weep for all the hours I wasted doing assembly language as a teenager that I could have been writing Lisp. How much software could have been written that would have been <100 lines of Lisp if only someone had written that tool?
I see way more Scala in companies last ~5y and have the impression of its ecosystem being more robust. Not uncommon for greenfields. It's longer than that I even encountered an active Clojure codebase. This is from a data-engineer perspective.
Clojure may be more popular for some niche of app startups perhaps? We are in different "bubbles" I suppose.
EDIT: Data disagrees with you also.
https://www.tiobe.com/tiobe-index/
https://redmonk.com/sogrady/2024/09/12/language-rankings-6-2...
https://survey.stackoverflow.co/2024/technology#1-programmin...
Smalltalk and Common Lisp are not individuals.
"$99 Smalltalk Announced" 1986 InfoWorld Jun 30
https://books.google.com/books?id=Wi8EAAAAMBAJ&pg=PA11&dq=Sm...
I kind of fail to see Lisp as an alternative to assembler on mid 80s micros.
Though, there were several cheap Lisps for PCs...
Rapid application technologies, methedologies, or frameworks are not unusual.
I know some wonderfully productive polyglot developers who by their own choice end up at Clojure. It doesn't have to be for everyone.
I wouldn't rule out that Clojure doesn't deserve credit. I wouldn't think it's a good idea to discredit Clojure from not having tried it myself.
I do hope someone with extensive Clojure experience can weigh in on the advantages.
How easy something is a codebase grows is something to really consider.
This product regardless of how it's built is pretty impressive. I'd be open to learning advantages and comparisons without denying it.
With time you get to understand the power of simplicity. How to break the problem and compose the solutions to achieve your intended result..
That's where the power of Clojure came in for us.
I believe once you've got to some point, dealing with abstractions is a way of life. It's either in the language, the technical requirements, or the software design.
In the end it's about designing abstraction, and community's focus on designing simple abstractions drove me in designing the whole system.
Now once I have the working system, I am fairly sure it can be implemented in any language.
ehmmm.... excuse me.... erghmm... what about Emacs? I'm sure, it absolutely can be count for a "successful system that have to be maintained for ages". For far, far longer than any Java-based project that ever existed.
Even though Elisp lacks:
- static typing
- OOP class system (until relatively recently)
- Modern package management (until ELPA/MELPA)
- Multi-threading model
- JIT compilation
Perhaps "the secret sauce" of successful software is in simplicity? Maybe some programmers just get it, and for others, it is such an obscure and mysterious entity. Some programmers write "programs to program computers", and some may have realized that they are not trying to solve purely technological problems, but they are, in fact, tackling socio-technological problems, and they write programs to communicate their ideas to fellow human beings, not machines.
However, emacs is a fucking mess, and there is a reason "init.el bankruptcy" is a thing and why the most popular way to use emacs is through various frameworks such as doom or spacemacs.
In emacs, nearly everything can(and often does) mess with everything else. It is serious integration hell to actually get things to work together, and the work that goes into e.g. doom is basically all about managing that complexity through good abstractions and more rigid ways to configure and install things.
Emacs is also objectively dogshit in a lot of ways compared to most modern editors. LSP is ridiculously slow and a constant source of performance issues, many of which are probably directly related to emacs internals. Eglot seems to do better but it's a lot more limited(you can't use multiple language servers together, for example). Then there's things like the buffer being the data-structure for everything, which is sort of like modeling nearly everything as one long string. Things that would be trivial to do in most other languages or contexts are difficult and error-prone in emacs.
To be fair, I will readily say that the lack of static analysis really does bite when refactoring, though I think that good design principles and the overall productivity multiplier may offset that cost (also unique, descriptive, grep-able names!). I guess I've also seen enough C++ template spaghetti to know that static typing is no panacea either.
I don't know to what extent I'll use dynamic languages going forward, though for now I'm kind of in love with opening up a window into the computer and building up my digital sandcastles. Many of these languages also have a great FFI story, which is making me dream up cool bilingual projects that play on the strengths of both approaches.
All in all, no regrets about my adventures in dynamic-land.
Static types are great, but boy... I love my REPL. I think I'd hate to actually work while writing code. REPL-driven interactivity with Clojure allows me to treat the work like I'm playing a video game.
Because of being a Lisp based language, bringing something else to the table besides "lets replace Java", and the community being welcoming of the host environments where Clojure is a guest.
Not that I “feel” it was no problem, but there were no bugs found that could be traces down to that.
It was not a small codebase.
Sure, Clojure is dynamically typed, but it is also strongly typed. That in practice means that for example Clojurescript when compiling to Javascript enforces those type guarantees, sometimes emitting safer code than even statically typed Typescript cannot.
I've come to the conclusion that it is just a better experience using a language that already has static types for large projects, than trying to make a dynamic language have similar things. Having to wrap every function in a error boundary to get somewhat of a meaningful debug experience is just .. awful.
I mean, I love Clojure, and used it for personal and work projects for 10+ years, some of which have hundreds of stars on github. But I cannot count the time wasted to spot issues where a map was actually a list of maps. Here Elixir is doing the right thing - adding gradual typing.
(declare (type Integer m))Some narrow view. Have you tried thinking that maybe Clojure intentionally chose not to include type declarations because they can lead to a messy middle ground? After all, maybe not every feature from Common Lisp needs to be replicated in every Lisp dialect? Besides, Clojure's Spec and Malli can be far more powerful for validation as they can define complex data structures, you can generate test data from them, you can validate entire system states, and they can be manipulated as data themselves.
If CL so "far superior" like you say, why then it can't be 'hosted' like Clojure? Why Clojure has Clojurescript, ClojureCLR, ClojureDart, babashka, nbb, sci, etc.? I'm not saying that to argue your specific point. Neither of them is 'superior' to another. They both have different purposes, philosophies, and use cases. Each has its strengths, pros, and cons. And that is actually very cool.
I guess the major advantage for Closure with this style is the “persisted” data structures end up sharing some bytes behind the scenes - it’s nice the language is explicitly situated around this style, rather than TypeScript’s chaotic Wild West kitchen sink design. What I don’t understand the advantage for “state management”. Like, you build a new state object, and then mutate some pointer from prevState to nextState… that’s what everyone else is doing too.
There are times though when it’s nice to switch gears from function-and-data to an OO approach when you need to maintain a lot of invariants, interior mutability has substantial performance advantages, or you really want to make sure callers are interpreting the data’s semantics correctly. So our style has ended up being “functional/immutable business logic and user data” w/ “zero inheritance OO for data structures”.
Whenever I read some open source TypeScript code that’s using the language like it’s Java like `class implements ISomething` ruining cmd-click go to method or an elaborate inheritance hierarchy it makes me sad.
Clojure's real super power is its reference type(s) (in particular the atom). Rich does an excellent job explaining them in this video: https://www.youtube.com/watch?v=wASCH_gPnDw&t=2278s
Seems like many similar capabilities, like a focus on immutable data structures, pure functions, being able to patch and update running systems without a restart, etc.
CIDER and nREPL is better tech than IEX though. I live in both and Clojure is much more enjoyable.
1. IEx provides a robust and interactive debugging environment that allows me to dig into whatever I want, even when running in production. I've never lost state in IEx, but that happens fairly often in CIDER and nREPL.
2. IEx uses Elixir's compilation model, which is a lot faster than CIDER and nREPL, leading to faster debugging cycles.
3. IEx is tightly integrated with Elixir whereas Clojure's tools are more fragmented.
4. IEx doesn't carry the overhead of additional middleware that CIDER and nREPL do.
I'm also not a fan of JVM deployments, so I've migrated all my code away from Clojure to Elixir during the past 10 years.
In how many man-hours/days? It's hard to know if the list is long or short only knowing that calendar time should be multiplied by three for calculating people time spent...
2 years if you count when I exploring building it in other languages.
Alongside building Vade Studio: I have been working as a contractor for 2 clients. Developing systems for them Other two developers have been managing their college curriculum as well.
I am not sure how to do the math around it, but anecdotally I don't think this would be possible in any other environment.
A few years ago, I worked in a small group (6 devs) for a retail business. We had all sorts of different third-party integrations - from payment processors to coupon management and Google Vision (so people wouldn't upload child porn or some other shit through our web and mobile apps). The requirements would constantly change - our management would be like: "Hey guys, we want this new program to launch, can we do it next week?", then a day later: "Turns out we can't do it in the state of Tennessee, let's add some exceptions, okay?", then some time later: "Folks, we really want to A/B test this but only for a specific class of customers..." Jesus, and it wasn't just an "occasional state of affairs" every once in a while; it was a constant, everyday, business-as-usual flow. We had to quickly adapt, and we had to deploy continuously. We had several services, multiple apps - internal, public, web and mobile, tons of unit and E2E tests, one legacy service in Ruby, we had Terraform, containers, API Gate, load balancers, etc.
I can't speak for myself, but a couple of my peers were super knowledgeable. They used all sorts of tools and languages before. I remember my team-lead showing me some Idris features (tbh, I don't even remember anymore exactly what) and asking my opinion if we should find a way to implement something like that, and I couldn't hide my confusion as I didn't know nothing about Idris.
Numerous times, we had discussions on improving our practices, minimizing tech debt, etc. And I remember distinctly - many times we have speculated how things would've turned out if we used some other stack, something that's not Clojure. We would explore various scenarios, we even had some prototypes build in Golang, Swift and Kotlin. And every single time, we found compelling and practical reasons for why Clojure indeed was the right choice for the job.
Sure, if we had a larger team, maybe we could've done it using a different stack. But it was a startup, and we had just the six of us.
I only use google for email logins for services I don't take seriously and am willing to lose.
Your concern makes sense though and we'll be considering it in the next feature rollout.
Because Clojure treats data as first-class citizens, we could build our own lightweight conflict resolution system using pure functions that operate on these transactions.
1. Passed as arguments
2. Returned from functions
3. Stored in variables
4. Manipulated directly
5. Compared easily
Unlike some languages where data needs special handling or conversion, Clojure lets you work with data structures directly and consistently throughout your program.
This philosophy extends to how Clojure handles data transformations. For example, transducers are composable algorithmic transformations that can work on any data source - whether it's a collection, stream, or channel. They treat the data transformation itself as a first-class value that can be composed, stored, and reused independently of the input source.
This first-class treatment of both data and data transformations makes Clojure particularly powerful for data processing and manipulation tasks.
That's why Clojure often finds strong adoption in data analytics, fintech and similar domains. The ability to treat both data and transformations as first-class citizens makes it easier to build, for example: reliable financial systems where data integrity is crucial.
The two sentences around the one you quoted should answer the question as well:
> With Clojure, we modeled the entire collaboration system as a stream of immutable data transformations. Each user action becomes a transaction in our system.
> When conflicts occur, our system can merge changes intelligently because we're working with pure data structures rather than complex objects.
In particular with clojure, data structures tend to be immutable and functions tend to not have side effects. This gives rise to the benefits the article talks about, though is not without its own drawbacks.
- immutability and persistent data structures (makes code easier to reason about [the data]; enables efficient concurrency - no locks; some algorithmic tricks that makes it very performant despite having to create copies of collections),
- seq abstraction - unlike other Lisp where sequence functions are often specialized for different types, Clojure simplifies things by making baked-in abstraction central to the language - all core functions work with seqs by default. it emphasizes lazy sequences as a unified way to process data, i.e., memory efficiency and infinite sequences, etc.
- rich standard library of functions for data transformation
- destructuring - makes code both cleaner and more declarative
- emphasis on pure functions working on simple data structures
The combination of these features makes data processing in Clojure particularly elegant and efficient.
I think having a clear example would help in understanding.
Subscribe the newsletter to know when it's live.
Incredible history, I feel like Clojure makes magic. What I like about functional programming is that it brings other perspectives of how things CAN work!!
Congratulations by the life change
I'm unable to understand this mindset. All the time I read things like "Developers love complexity because it feeds their egos" but I've never encountered a situation in which added complexity made me more proud of the work. Just the opposite: being able to do more made me more proud of the work I put in, and complexity was the price I paid for that ability. The greatest hacks, the ones that etch people's names into history, are the ones -- like Unix and the Doom engine -- that achieve phenomenal feats with very little code and/or extreme parsimony of design. This is no more true than in that famous ego-stroking/dick-measuring contest of programming, the demoscene. My favorite example being the 4k demo Omniscent: https://www.youtube.com/watch?v=G1Q9LtnnE4w
Being able to stand up a 100-node K8s cluster to provide a basic web service, connected to a React SPA front end to provide all the functionality of a Delphi program from the 90s doesn't stroke the ego of any programmer I know of; but it might stroke their manager's ego because it gives them an opportunity to empire-build and requisition a larger budget next year.
Big long lived code bases are all about this battle against complexity and the speed at which you can add new or update features largely comes down to how well you’re doing at management of complexity.
Being able to use kubernetes for infrastructure, grafana, prometheus, etc Elasticsearch for search, Mongodb as database, redis as caching layer.
Knowing all these tech and being able to say you know these very well used to massage my developer ego...
Now I am much more like: Use one system to best of it's capability. Use Postgres. Mostly you won't need anything else.
I never resisted to urge to try out something new and shiny in production in earlier days.
Now I mostly use boring technologies and things I am comfortable with running in production.
Collecting knowledge of various technologies-of-the-now isn't really a point of pride in the field, unless you're aggressively doing résumé driven development, which is a short term win, long term lose strategy.
But yeah, I'm glad you've come to see that. And yes, just use Postgres. As I like to put it, unless you have a really good reason not to use an RDBMS, use an RDBMS. But we're living in an era where bootcamp grads come out knowing only JavaScript, and MongoDB lets you make queries that resemble JavaScript, so it becomes that thing of "real professionals choose technologies for which there is a large talent pool".
Not really, in an OO language state could have been stored in some data structure as well, with a way to serialize and deserialize. E.g. React made this very popular.
Specifically, I find language evangelists particularly likely to be closer to .5x than 5x. And that's before you even account for their tendency to push for rewriting stuff that already works, because "<insert language du jour here> is the future, it's going to be great and bug free," often instead of solving the highest impact problems.
There are fewer of them, they ask for more money, but they really are exceptional. Especially Rust devs right now because there are not a lot of jobs you only find the most passionate and the most brilliant in that space. A short window though which will close as Rust gets more popular to startups, take advantage of it now.
It depends; exact meaning, application and personal preference play a big role.
However my criteria for selecting a language for use in a professional context:
0: fit to task - obviously the language has to be able to do the job - to take this seriously you must define the job and what its requirements are and map those against the candidate languages
1: hiring and recruiting - there must be a mainstream sized talent pool - talent shortages are not acceptable - and I don't buy the argument that "smart people are attracted to non mainstream languages which is how we find smart people", it is simply not true that "most smart people program with Scala/Haskell/Elixir/whatever" - there's smart and smarter working on the mainstream languages.
2: size of programming community, size of knowledge base, size of open source community - don't end up with a code base stuck in an obscure corner of the Internet where few people know what is going on
3: AI - how well can AI program in this language? The size of the training set counts here - all the mainstream languages have had vast amounts of knowledge ingested and thus Claude can write decent code or at least has a shot at it. And in future this will likely get better again based on volume of training data. AI counts for a huge amount - if you are using a language that the AI knows little about then there's little productivity related benefits coming to your development team.
4: tools, IDE support, linters, compilers, build tools etc. It's a real obstacle to fire up your IDE and find that the IDE knows nothing about the language you are using, or that the language plugin was written by some guy who did it for the love and its not complete or professional or updated or something.
5: hiring and recruiting - it's the top priority and the bottom and every priority in between. If you can't find the people then you are in big trouble I have seen this play out over and over where the CTO's favorite non-mainstream language is used in a professional context and for years - maybe decades after the company suffers trying to find people. And decades after the CTO moved on to a new company and a new favorite language.
So what is a mainstream language? Arguable but personally it looks like Python, Java, JavaScript/TypeScript, C#, Golang. To a lesser extent Ruby only because Ruby developers have always been hard to find even though there is lots of community and knowledge and tools etc. Rust seems to have remained somewhat niche when its peer Golang has grown rapidly. Probably C and C++ depending on context. Maybe Kotlin? How cares what I think anyway its up to you. My main point is - in a professional context the language should be chosen to service the needs of the business. Be systematic and professional and don't bring your hobbies into it because the business needs come first.
And for home/hobbies/fun? Do whatever the heck you like.
Smart people can be trained in any language and become effective in a reasonably short period of time. I remember one company I worked at, we hired a couple of fresh grads who'd only worked with Java at school based on how promising they seemed; they were contributing meaningfully to our C++ code base within months. If you work in Lisp or Haskell or Smalltalk or maybe even Ruby, chances are pretty good you've an interesting enough code base to attract and retain this kind of programmer. Smart people paired with the right language can be effective in far smaller numbers as well.
The major drawback, however, is that programmers who are this intelligent and this interested in the work itself (rather than the money or career advancement opportunities) are likely to be prickly individualists who have cultivated within themselves Larry Wall's three programmer virtues: Laziness, Impatience, and Hubris. So either you know how to support the needs of such a programmer, or you want to hire from a slightly less intelligent and insightful, though still capable, segment of the talent pool which means no, you're not going to be targeting those powerful languages off the beaten track. (But you are going to have to do a bit more chucklehead filtering.)
> if you are using a language that the AI knows little about then there's little productivity related benefits coming to your development team.
This is vacuously true because the consequent is always true. The wheels are kind of falling off "Dissociated Press on steroids" as a massive productivity booster over the long haul. I think that by the time you have an AI capable of making decisions and crystallizing intent the way a human programmer can, then you really have to consider whether to give that AI the kind of rights we currently only afford humans.
The amount of knuckleheads that Ive had to interview just to get a single coherent developer is mind boggling (remote first).
Rust Common Lisp Go Ruby/Elixir C++ Python C# Typescript Java Javascript
Clojure is my personal favorite language, and I am planning to build very small team. So it would work for us.
Small correction: finding experienced people is difficult. There's no shortage of engineers who only briefly tried Clojure and would love to use it at their full-time gig.
An example: When existing technology team members can run an interview and decide if someone could learn what they're working with.