As it currently stands, this question is not a good fit for our Q&A format. We expect answers to be supported by facts, references, or expertise, but this question will likely solicit debate, arguments, polling, or extended discussion. If you feel that this question can be improved and possibly reopened, visit the help center for guidance.
Closed 11 years ago.
Locked. This question and its answers are locked because the question is off-topic but has historical significance. It is not currently accepting new answers or interactions.
As my first programming language, I decided to learn Haskell. I'm an analytic philosophy major, and Haskell allowed me to quickly and correctly create programs of interest, for instance, transducers for natural language parsing, theorem provers, and interpreters. Although I've only been programming for two and a half months, I found Haskell's semantics and syntax much easier to learn than more traditional imperative languages, and feel comfortable (now) with the majority of its constructs.
Programming in Haskell is like sorcery, however, and I would like to broaden my knowledge of programming. I would like to choose a new programming language to learn, but I do not have enough time to pick up an arbitrary language, drop it, and repeat. So I thought I would pose the question here, along with several stipulations about the type of language I am looking for. Some are subjective, some are intended to ease the transition from Haskell.
Strong type system. One of my favorite parts of programming in Haskell is writing type declarations. This helps structure my thoughts about individual functions and their relationship to the program as a whole. It also makes informally reasoning about the correctness of my program easier. I'm concerned with correctness, not efficiency.
Emphasis on recursion rather than iteration. I use iterative constructs in Haskell, but implement them recursively. However, it is much easier to understand the structure of a recursive function than a complicated iterative procedure, especially when using combinators and higher-order functions like maps, folds and bind.
Rewarding to learn. Haskell is a rewarding language to work in. It's a little like reading Kant. My experience several years ago with C, however, was not. I'm not looking for C. The language should enforce a conceptually interesting paradigm, which in my entirely subjective opinion, the C-likes do not.
Weighing the answers: These are just notes, of course. I'd just like to reply to everyone who gave well-formed responses. You have been very helpful.
1) Several responses indicated that a strong, statically typed language emphasizing recursion means another functional language. While I want to continue working strongly with Haskell, camccann and larsmans correctly pointed out that another such language would "ease the transition too much." These comments have been very helpful, because I am not looking to write Haskell in Caml! Of the proof assistants, Coq and Agda both look interesting. In particular, Coq would provide a solid introduction to constructive logic and formal type theory. I've spent a little time with first-order predicate and modal logic (Mendellsohn, Enderton, some of Hinman), so I would probably have a lot of fun with Coq.
2) Others heavily favored Lisp (Common Lisp, Scheme and Clojure). From what I gather, both Common Lisp and Scheme have excellent introductory material (On Lisp and The Reasoned Schemer, SICP). The material in SICP causes me to lean towards Scheme. In particular, Scheme through SICP would cover a different evaluation strategy, the implementation of laziness, and a chance to focus on topics like continuations, interpreters, symbolic computation, and so on. Finally, as others have pointed out, Lisp's treatment of code/data would be entirely new. Hence, I am leaning heavily towards option (2), a Lisp.
3) Third, Prolog. Prolog has a wealth of interesting material, and its primary domain is exactly the one I'm interested in. It has a simple syntax and is easy to read. I can't comment more at the moment, but after reading an overview of Prolog and skimming some introductory material, it ranks with (2). And it seems like Prolog's backtracking is always being hacked into Haskell!
4) Of the mainstream languages, Python looks the most interesting. Tim Yates makes the languages sound very appealing. Apparently, Python is often taught to first-year CS majors; so it's either conceptually rich or easy to learn. I'd have to do more research.
Thank you all for your recommendations! It looks like a Lisp (Scheme, Clojure), Prolog, or a proof assistant like Coq or Agda are the main langauages being recommended.
I would like to broaden my knowledge of programming. (...) I thought I would pose the question here, along with several stipulations about the type of language I am looking for. Some are subjective, some are intended to ease the transition from Haskell.
Strong type system. (...) It also makes informally reasoning about the correctness of my program easier. I'm concerned with correctness, not efficiency.
Emphasis on recursion rather than iteration. (...)
You may be easing the transition a bit too much here, I'm afraid. The very strict type system and purely functional style are characteristic of Haskell and pretty much anything resembling a mainstream programming language will require compromising at least somewhat on one of these. So, with that in mind, here are a few broad suggestions aimed at retaining most of what you seem to like about Haskell, but with some major shift.
Disregard practicality and go for "more Haskell than Haskell": Haskell's type system is full of holes, due to nontermination and other messy compromises. Clean up the mess and add more powerful features and you get languages like Coq and Agda, where a function's type contains a proof of its correctness (you can even read the function arrow -> as logical implication!). These languages have been used for mathematical proofs and for programs with extremely high correctness requirements. Coq is probably the most prominent language of the style, but Agda has a more Haskell-y feel (as well as being written in Haskell itself).
Disregard types, add more magic: If Haskell is sorcery, Lisp is the raw, primal magic of creation. Lisp-family languages (also including Scheme and Clojure) have nearly unparalleled flexibility combined with extreme minimalism. The languages have essentially no syntax, writing code directly in the form of a tree data structure; metaprogramming in a Lisp is easier than non-meta programming in some languages.
Compromise a bit and move closer to the mainstream: Haskell falls into the broad family of languages influenced heavily by ML, any of which you could probably shift to without too much difficulty. Haskell is one of the strictest when it comes to correctness guarantees from types and use of functional style, where others are often either hybrid styles and/or make pragmatic compromises for various reasons. If you want some exposure to OOP and access to lots of mainstream technology platforms, either Scala on the JVM or F# on .NET have a lot in common with Haskell while providing easy interoperability with the Java and .NET platforms. F# is supported directly by Microsoft, but has some annoying limitations compared to Haskell and portability issues on non-Windows platforms. Scala has direct counterparts to more of Haskell's type system and Java's cross-platform potential, but has a more heavyweight syntax and lacks the powerful first-party support that F# enjoys.
Most of those recommendations are also mentioned in other answers, but hopefully my rationale for them offers some enlightenment.
I'm going to be That Guy and suggest that you're asking for the wrong thing.
First you say that you want to broaden your horizons. Then you describe the kind of language that you want, and its horizons sound incredibly like the horizons you already have. You're not going to gain very much by learning the same thing over and over.
I would suggest you learn a Lisp — i.e. Common Lisp, Scheme/Racket or Clojure. They're all dynamically typed by default, but feature some sort of type hinting or optional static typing. Racket and Clojure are probably your best bets.
Clojure is more recent and has more Haskellisms like immutability by default and lots of lazy evaluation, but it's based on the Java Virtual Machine, which means it has some odd warts (e.g. the JVM doesn't support tail call elimination, so recursion is kind of a hack).
Racket is much older, but has picked up a lot of power along the way, such as static type support and a focus on functional programming. I think you'd probably get the most out of Racket.
The macro systems in Lisps are very interesting and vastly more powerful than anything you'll see anywhere else. That alone is worth at least looking at.
From the standpoint of what suits your major, the obvious choice seems like a logic language such as Prolog or its derivatives. Logic programming can be done very neatly in a functional language (see, e.g. The Reasoned Schemer) , but you might enjoy working with the logic paradigm directly.
An interactive theorem proving system such as twelf or coq might also strike your fancy.
I'd advise you learn Coq, which is a powerful proof assistant with syntax that will feel comfortable to the Haskell programmer. The cool thing about Coq is it can be extracted to other functional languages, including Haskell. There is even a package (Meldable-Heap) on Hackage that was written in Coq, had properties proven about its operation, then extracted to Haskell.
Another popular language that offers more power than Haskell is Agda - I don't know Agda beyond knowing it is dependently typed, on Hackage, and well respected by people I respect, but those are good enough reasons to me.
I wouldn't expect either of these to be easy. But if you know Haskell and want to move forward to a language that gives more power than the Haskell type system then they should be considered.
As you didn't mention any restrictions besides your subjective interests and emphasize 'rewarding to learn' (well, ok, I'll ignore the static typing restriction), I would suggest to learn a few languages of different paradigms, and preferably ones which are 'exemplary' for each of them.
A Lisp dialect for the code-as-data/homoiconicity thing and because they are good, if not the best, examples of dynamic (more or less strict) functional programming languages
Prolog as the predominant logic programming language
Smalltalk as the one true OOP language (also interesting because of its usually extremely image-centric approach)
maybe Erlang or Clojure if you are interested in languages forged for concurrent/parallel/distributed programming
Forth for stack oriented programming
(Haskell for strict functional statically typed lazy programming)
Especially Lisps (CL not as much as Scheme) and Prolog (and Haskell) embrace recursion.
Although I am not a guru in any of these languages, I did spend some time with each of them, except Erlang and Forth, and they all gave me eye-opening and interesting learning experiences, as each one approaches problem solving from a different angle.
So, though it may seem as if I ignored the part about your having no time to try a few languages, I rather think that time spent with any of these will not be wasted, and you should have a look at all of them.
How about a stack-oriented programming language? Cat hits your high points. It is:
Statically typed with type inference.
Makes you re-think common imperative languages concepts like looping. Conditional execution and looping are handled with combinators.
Rewarding - forces you to understand yet another model of computation. Gives you another way to think about and decompose problems.
Dr. Dobbs published a short article about Cat in 2008 though the language has changed slightly.
If you want a strong(er)ly typed Prolog, Mercury is an interesting choice. I've dabbled in it in the past and I liked the different perspective it gave me. It also has moded-ness (which parameters need to be free/fixed) and determinism (how many results are there?) in the type system.
Clean is very similar to Haskell, but has uniqueness typing, which are used as an alternative to Monads (more specifically, the IO monad). Uniqueness typing also does interesting stuff to working with arrays.
I'm a bit late but I see that no one has mentioned a couple of paradigms and related languages that can interest you for their high-level of abstraction and generality:
rewriting systems, like Maude or ELAN;
Constraint Handling Rules (CHR).
Despite its failure to meet one of your big criteria (static* typing), I'm going to make a case for Python. Here are a few reasons I think you should take a look at it:
For an imperative language, it is surprisingly functional. This was one of the things that struck me when I learned it. Take list comprehensions, for example. It has lambdas, first-class functions, and many functionally-inspired compositions on iterators (maps, folds, zips...). It gives you the option of picking whatever paradigm suits the problem best.
IMHO, it is, like Haskell, beautiful to code in. The syntax is simple and elegant.
It has a culture that focuses on doing things in a straightforward way, rather than focusing too minutely on efficiency.
I understand if you are looking for something else though. Logic programming, for instance, might be right up your alley, as others have suggested.
* I assume you mean static typing here, since you want to declare the types. Techincally, Python is a strongly typed language, since you can't arbitrarily interpret, say, a string as an number. Interestingly, there are Python derivatives that allow static typing, like Boo.
I would recommend you Erlang. It is not strong typed language and you should try it. It is very different approach to programming and you may find that there are problems where strong typing is not The Best Tool(TM). Anyway Erlang provides you tools for static type verification (typer, dialyzer) and you can use strong typing on parts where you gain benefits from it. It can be interesting experience for you but be prepared, it will be very different feeling. If you are looking for "conceptually interesting paradigm" you can found them in Erlang, message passing, memory separation instead sharing, distribution, OTP, error handling and error propagation instead of error "prevention" and so. Erlang can be far away from your current experience but still brain tickling if you have experience with C and Haskell.
Given your description, I would suggest Ocaml or F#.
The ML family are generally very good in terms of a strong type system. The emphasis on recursion, coupled with pattern matching, is also clear.
Where I am a bit hesitant is on the rewarding to learn part. Learning them was rewarding for me, no doubt. But given your restrictions and your description of what you want, it seems you are not actually looking for something much more different than Haskell.
If you didn't put your restrictions I would have suggested Python or Erlang, both of which would take you out of your comfort zone.
In my experience, strong typing + emphasis on recursion means another functional programming language. Then again, I wonder if that's very rewarding, given that none of them will be as "pure" as Haskell.
As other posters have suggested, Prolog and Lisp/Scheme are nice, even though both are dynamically typed. Many great books with a strong theoretical "taste" to them have been published about Scheme in particular. Take a look at SICP, which also conveys a lot of general computer science wisdom (meta-circular interpreters and the like).
Factor will be a good choice.
You could start looking into Lisp.
Prolog is a cool language too.
If you decide to stray from your preference for a type system,you might be interested in the J programming language. It is outstanding for how it emphasizes function composition. If you like point-free style in Haskell, the tacit form of J will be rewarding. I've found it extraordinarily thought-provoking, especially with regard to semantics.
True, it doesn't fit your preconceptions as to what you'd like, but give it a look. Just knowing that it's out there is worth discovering. The sole source of complete implementations is J Software, jsoftware.com.
Go with one of the main streams. Given the resources available, future marketability of your skill, rich developer ecosystem I think you should start with either Java or C#.
Great question-- I've been asking it myself recently after spending several months thoroughly enjoying Haskell, although my background is very different (organic chemistry).
Like you, C and its ilk are out of the question.
I've been oscillating between Python and Ruby as the two practical workhorse scripting languages today (mules?) that both have some functional components to them to keep me happy. Without starting any Rubyist/Pythonist debates here, but my personal pragmatic answer to this question is:
Learn the one (Python or Ruby) that you first get an excuse to apply.
In formal specifications based on abstract algebraic types and equational theory you use formulas of equational theory to specify theory. System which will satisfy those constraints is called in formal logic a model.
Modeling is process of creating a model, which abstracts of some aspects, which are unnecessary details for a specific case. So concrete system has to adhere to created model in observed aspects.
Programming is a process of creating a program which will have specific behaviour - will perform specific algorithms - and programming languages through different paradigms enable us to think in a certain specific way, which abstracts of some details, usually machine specific ones.
So could we be doing all those things at the same time, because they are principially the same? Is declarative programming the nearest attempt to do that? Could we use some sort f programming languages which will be good for programming as well as for modeling and specification?
The scientist who has done the most to advance this point of view is Tony Hoare. Tony, along with his colleague Edsger Dijkstra, advocated nondeterministic programming languages so that there would be a smoother path from specification to implementation. Tony definitely wanted a single language for both specification and implementation. For more on this view, read his book on the Alegbra of Programming. Tony also did the seminal work on proving correctness of abstractions. All of this work was done in the context of simple, imperative languages with structured control flow and classic, side-effecting procedures. So there is not any connection with declarative programming of necessity. And historically, work on functional programming (the main branch of declarative programming) has followed more from Backus's Turing lecture on "liberating programming from the von Neumann bottleneck"; functional programming has been about programming productivity as much as anything else.
What we discovered since Hoare is that formal specifications and formal modelsl are very expensive. The expense hasn't been shown to be justified except in very special circumstances, like "if the software doesn't work, the patient will die" or "if the software doesn't work, the plane will crash." Informal models and specifications are quite useful, and much cheaper to produce and work with. There is still interesting research going on around the fringes on modelling, model checking, and so on. One of my personal favorites is the Alloy language done by Daniel Jackson's group at MIT. There's also great stuff done at Microsoft Research and plenty of good stuff elsewhere. There's some work in declarative programming as well, but it too is of the "cheap and cheerful" variety rather than a comprehensive, programmatic approach like Hoare's. One of my favorites there is Claessen's and Hughes's QuickCheck, which provides a way to state formal properties and explore them by random testing. No proofs or theorems, but still jolly useful.
In summary, you describe an agenda of doing formal models, specifications, and programs, all within a single framework. There is still plenty of good work going on piecemeal, but the unified agenda has been abandoned.
Is there high-level language out there for describing algorithms, that's geared towards specification, rather than implementation?
The idea would be to have a machine-readable archive of standard algorithms, with machine-readable annotations on trade-offs, and variants.
I'm thinking of something like CycL / OpenCyC, but for algorithms and programming patterns.
These are not exact fits, but they are somewhat close: Maude and CASL. They are both machine-readable (with decent tools) specification languages which also allow some forms of code. Some people swear by AsmL, but I'm not a fan.
Soon I think that Agda 2 will fulfill that niche quite well, but it's still in full development. One can also say that Coq already fulfills that role (because of program extraction), but I am not a fan either.
The glib answer: Yes. It's called English.
The serious answer: No, as you've probably already assumed after months of this question going unanswered. I don't believe there's any real consensus, or even consideration, over what such a language would entail. Programming patterns and algorithms are always changing and evolving, so designing a language that would be able to describe all future patterns would be very difficult, if not impossible.
Pseudocode, possibly. Or flowcharts. For machine readable, it would have to be something like a meta-language. Lisp and Forth come to mind. Languages that are essentially self-defining, with a small set of primitives to build up the higher layers of abstraction.
Closed. This question is off-topic. It is not currently accepting answers.
Want to improve this question? Update the question so it's on-topic for Stack Overflow.
Closed 10 years ago.
Improve this question
What is the real benefit of creating a new programming language? It is highly unlikely that you are going to actually use it.
In short, how will the process of creating a new language make you a better programmer?
You will understand the decisions behind language design and garner a better overall understanding of the compromises made between readability, performance, and reliability.
Your familiarity with concepts such as recursion, closures, garbage collection, reference management, typing, data structures and how these things actually work will increase. Most programmers will utilize resources and language features better.
Similar to the way we learn new ways to code solutions when we use other languages, when we write our own languages, we explore new ways to create solutions. See Metaprogramming. Contrary to the what the question suggests, Domain Specific Languages are used in many environments.
If you're writing a compiler, you'll learn more about how computers work than you ever did before. (Depending on your goal, perhaps more than you intended to learn)
When I wrote my own sort routines in school, even re-implementations of good ones, it really drove home some of the weaknesses of some of the algorithms.
In short, there's an order of magnitude of difference in a programmer who knows how to use tools, and a programmer who knows how to make tools.
I can speak from experience here ...
Fun, Domain specific problem solving, Complexity in context
I love creating new languages for fun, and for tackling domain specific problems. A very simple example might be Wikipedia markup or something as complex as Erlang which specializes in concurrent processing.
Many general purpose languages are similar, because they are general purpose. Sometimes you need a more accurate abstraction of the mechanics of the problem you are solving. Another example would be the M4 macro language.
Remember a language is not magic, it is just a collection of defined grammatical structures with implied semantics. SQL is a good example of a language for a purpose, with that purpose defined in it's syntax and semantics.
Learning how languages work, what makes a language parsable, what makes semantics sensible and the implementation of this, I think can make you a better programmer.
compilers embody alot of theory that underpins computer science:
Translation, abstraction, interpretation, data structures, state .... the list goes on. Learning these things will make you understand the implications of your program and what goes on under the hood. You can of course learn things independently but compilers are a great context to learn complex topics such as DFA/NDFA automata, stack-based parsers, abstract syntax trees ....
compilers are beautiful machines I think :)
Multiple reasons:
bragging rights
economic incentives
extreme boredom
dissatisfaction with the hundreds of existing languages
untreated insanity
desire to implement language that facilitates new design concepts (like languages that make design patterns more straightforward to incorporate)
other reasons, perhaps
I think Jeff Attwood answers this well in this Coding Horror post -- though he's talking about a more general issue (why create any new library, framework, etc, when other artifacts in the same design space already exist), I suspect that exactly said broader viewpoint gives him a different and interesting perspective.
I will add that if you write a semantics, so that your language is an actual language and not merely what happens to be accepted by some particular implementation, you will learn an enormous amount about how to describe computational behaviors precisely:
You will learn what kinds of behaviors are and are not easy to describe—and prove correct.
You will learn how to trade off different kinds of formalisms for describing different kinds of features.
You will ultimately be a better programmer because the formalism and proof techniques you will learn will apply to all kinds of problems: locking techniques, safety properties in kernels, lock-free data structures, network protocols, and information security, to name just a few. All these areas are amenable to the same kind of formal treatment that is given to a programming language.
To pick just one example, if you give your language a static type system and you then prove that a well-type program is guaranteed to be memory-safe, you will learn just as much (on a different dimension) as you will by writing an interpreter or compiler.
EDIT: If you want to learn this stuff I think the easiest starting point is Benjamin Pierce's series of two books on Types and Programming Languages. There is also a graduate textbook by Glynn Winskel which is a little harder but more oriented toward semantics and proof techniques.
Creating Domain Specific Languages is very valuable. Instead of thinking only about general purpose languages, consider creating so-called "little languages" that clearly express abstractions in your project.
For example, in a recent project I decided to use a Command Pattern to drive a Service Layer. I found some repetition in my command code, so I wrote a little compiler that accepts a simple language that expresses commands and emits command implementations in the "underlying" language.
For the same reason that taking a Compiler Construction course at university will benefit you even if you never write a single compiler in your whole life. It's a look under the hood, if you may.
In addition to what altCognito said, which is a theoretical/academic perspective, some highly specialized languages are created to solve specific problems efficiently when existing "general-purpose" languages are either extremely inefficient for your task or there just isn't an easy-to-use existing alternative.
Granted, that such cases tend to be rare and if your first instinct on encountering a problem is "I need a new language for this.", then it is most likely you're missing something. There needs to be a fairly substantial gap in "available" tech and and your needs to warrant such an undertaking.
I think there are really two conceptually different answers to this. First, you gain an understanding of how compilers transform your code into executable code. This can help you make better decisions about how to structure your code to optimize (or allow it to be optimized) better. If, for instance, you knew that a certain construct would prohibit the compiler from inlining a code block or unrolling a loop, then you could avoid that if performance became a real concern.
Second, all current languages were invented (or derived) at some point in history. For each one of these, the likelihood that it would actually be used was potentially small, yet here they are. They all found their reason for being in the fact that someone wanted to do something that wasn't possible or easy to do in an existing language and decided to do something about it. Laziness (or the desire to let the computer do the work for you) is the mother of invention.
Just for fun... and then you'll realize that you cannot make anything better than all the languages that you thought they sucked xD (so you stop complaining about them).
how will the process of creating a new language make you a better programmer?
You're right, you may or may not use the language, but at the least the experience you will gain from doing it will benefit you to understand the implementation of programming languages and of certain things that you will be able to apply to future computation problems that you run into.
Writing a compiler or interpreter requires a very firm understanding in computer science theory. And if you're compiling to machine code instead of to another language, it requires a firm understanding in hardware design as well.
In addition to that, knowing how to design a compiler means you will have a better understanding of languages in general, and the languages you work with specifically. You will have a better appreciation for syntax and trade-offs the language designers took when they wrote their specification.
It's not that writing compilers makes you a better programmer. It's the deep understanding of language theory and compiler design that makes you better.
Mostly you do this for fun or to broaden your comprehension of a subject.
I disagree that creating new language influences performance - performance of what? IMHO execution speed should not depend on the language constructs but what the language is translated to - which is something different: like creating a syntax for a language and writting a compiler/virtual machine for it.
Because a talking frog is pretty neat.
I want a managed language that permits tinkering with its internals as standard practice. Kind of like Ruby's duck punching on a wider scale.
I should, as the client of a library, be able to swap out library functions that don't do what I want.
That's what drives me crazy with .NET. There are bugs in the framework Microsoft will not fix and thanks to GAC signing I cannot. And even if it were not for GAC signing, hotpatching a global library is a bad idea (might break some other application).
I for one don't care about how compilers work, don't care about learning new languages, and don't care about using scripting languages like perl and javascript. I'm much more interested in the ways big programs are constructed (or should be constructed). There are still no good solutions for making LARGE software as easy to use as prototyped code. Programming languages are not helping with that. They solve trivial problems like sorting and memory deallocation, and leave you struggling alone with problems that really matter (that keep you or your firm from losing money).