Closed. This question is opinion-based. It is not currently accepting answers.
Want to improve this question? Update the question so it can be answered with facts and citations by editing this post.
Closed 8 years ago.
Improve this question
I'm currently working on the topic of programming-languages and interpreter-design. I have already created several programming languages but couldn't reach my goal so far:
Create a programming-language which focuses on giving the programmer a good feeling when writing code in it. It should just be fun and/or interesting and in no case annoying to write something in it.
I get this feeling when writing code in Python. I sometimes get the opposite with PHP and in rare cases when having to reinvent some wheel in C++.
So I've tried to figure out some syntactical features to make programming in my new language fun, but I just can't find any.
Which concrete features, maybe mainly in terms of syntax, do/could make programming in a language fun?
Examples:
I find it enjoyable to program in Ruby because of it's use of code blocks.
It would be nice if you could include exactly one example in your answer
Those features do not have to already exist in any language!
I'm doing this because I have experienced extreme rises in (my own) productivity when programming in languages I love (because of particular features).
You mentioned Ruby in your question. AFAIK, Ruby is the only programming language, for which Joy is an actual, stated, explicit design goal. (In fact, it is the only design goal.)
The reason that Yukihiro Matsumoto was able to design Ruby this way, is that he already knew and used tons of programming languages before he started designing Ruby and learned tons more in order to design Ruby. (Interestingly, he didn't know Python, and has said that he probably wouldn't have created Ruby if he did.)
Here's just a tiny fraction of the languages that matz has either used himself, or looked at for inspiration (or in some cases for inspiration what not to do):
CLU
Sather
Lisp
Scheme
Smalltalk
Perl
Python
Haskell
Scala
PHP
C
C++
Java
C#
Objective-C
Erlang
And I believe that this is one way that good programming languages can be designed (what Larry Wall calls postmodernist language design): Throw away everything that didn't work in the past, take everything that worked and combine that tastefully.
Of course, this requires that you actually know all those languages from which you want to "steal" and in particular, it requires that you know lots of very different languages with different paradigms, different concepts and different "feels", otherwise the idea pool from which you steal is rather small and inbred.
Consistency.
Its the feeling that you already know something when you use an API or feature you've never used before. It also makes you more productive as you don't have to learn something new for the sake of it.
I think this is also one of the Ruby 'likes', in that if you follow the naming convention, things start to 'just work' without bindings and glue and suchlike.
For example, using the STL in C++, many of the algorithms are the same for all containers - even strings. That makes it nice to use... except for those parts that do not follow the same API (eg vector of bools) then the difference is more noticable.
Two things to keep in mind are orthogonality and the principle of least surprise.
A programming language should make it easy to write correct programs and difficult (if not impossible) to write incorrect programs. For instance, in Java
long x = 2000000000 + 2000000000;
overflows, while
long x = 2000000000L + 2000000000;
doesn't. Is this obvious? I don't think so. Does anyone ever want something to overflow? I don't think so.
Hilarity.
http://lolcode.com/
Follow common practices (like using + for addition, & for bitwise/logical and)
Group logicaly-similar code in namespaces
Have an extensive string processing library
Incorporate debugging facilities
For a cross-platform language, try to minimize platform differences as much as possible
A language feature that appears simple and easy to learn surprises and delights the programmer with its unexpected power. I nominate Haskell type classes :-)
Related
Closed. This question is opinion-based. It is not currently accepting answers.
Want to improve this question? Update the question so it can be answered with facts and citations by editing this post.
Closed 9 years ago.
Improve this question
I've never programmed in a "pure" functional language. I earned my stripes on C and C++, tried Java, C#, PHP etc... but always I found myself going back to C++. Perhaps I'm a bit of a masochist, but I love the low level stuff.
I also find that I can accomplish rapid development quickly through the embedding of LUA, Python or other scripting languages (along with their focus on rapid development).
Long story short, I'm not quitting C/C++ so don't talk me out of it. However I've had little time to learn C++11 and I'm starting to feel the acceleration of the curve towards functional programming happening in the future.
My question is twofold. What language was C++11's concept of lambda functionality "borrowed" from, and what language would be the ideal one, if not that one, or if any to get a feel for "the way" to use C++11's new lambda functionality (no pun intended).
PS: I'm honestly not too happy about the new "bloated" additions to C++. I liked C++ how it was, it's starting to feel like the language is becoming bloated. I won't clam that to be a fact; I hear you have to have experienced a functional language to "get it".
It honestly seems like there is a new heavyweight in town. First it was just "procedural" programming, then came the OOP paradigm shift, while now it seems like things are heading towards the "functional" way of doing things.
Of course procedural programming is still alive and well (inside classes), I have to wonder where the lambda way will fit in (properly used) to class/oop design. Will it just be a replacement for the procedural part? Make OOP a thing of the past (pfft)? Or something else entirely (say, a functional event system generating events for objects encapsulating procedural code)?
I would try to curtail your opinions until you have more rigorous experience of the issues involved.
To paraphrase Bjarne Stroustrup: Functional programming has had a lot of airtime in academia over the last several decades, yet the number of deployed functional systems in industry remains about zero.
More concretely to your question, a lambda is just a short-hand syntactic way to declare a singleton functor object (a class with an operator() function) that captures variables from its enclosing scope as member variables. I wouldn't consider it a "functional programming" concept, any more so than any other entity in C++.
Functional programming generally involves immutable data types (objects that dont change once constructed) and pure functions (functions that have output that depends purely on their input, and nothing else).
If you are interested in functional programming there is a free online course (MOOC) starting right now called Functional Programming Principles in Scala, that serves as a very good and highly regarded introduction to the subject from one of the top Swiss universities.
I can't speak about lambdas in C++11, but I know that part of the rationale for adding lambdas to Java 8 is to enable transparent concurrency support out of the box. How? It provides a (lazy) Stream interface where you can switch between parallel and sequential processing simply by calling parallel and sequential (these methods return new streams, and do not have side effects on existing streams).
If you look at the methods in Stream, you'll quickly notice that without a lambda facility, they would be an extreme pain to use. Have a look at some examples of what you can do with streams in combination with lambdas.
It should be possible to implement a similar library for C++11, if there isn't already such a library.
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 11 years ago.
Improve this question
Disclaimer: Yes I know this will take 3 years, at least.
I am looking forward to writing a new interpreted programming language. I have a quite solid idea of what I want in terms of dynamicness, syntax, object model, etc, etc.
Now that I have the idea, I have a few questions before I start:
Should I begin writing the full specification and then implement, or write them both all along?
I'm still doubting between C and C++. C++ would allow for more clean design and faster development while C would (maybe) ensure portability to more platforms (microprocessors?). Performance is a must.
Should I try to interest people for the project before the first working prototype so they can cooperate (the end product will be a liberal license anyway), or keep working alone until I have something that runs?
How modular should it be? I am sure that I won't immediately start working on a bytecode interpreter but something easier to implement but slower thing first, so modularity is a must in order to be able to extend later, but I guess overdoing it will hamper performance and clearity.
The answers to your questions depend largely on why you're doing this- the primary reason. Are you trying to create the next Ruby, or is this a learning exercise?
Specification: If this is a personal project, this is not as important. PHP gets a bad rap for having been developed "on the fly," yet many people use it every day. A more complete spec will probably help get people involved if/when you want help.
If you want cross-platform and performance, C is the way to go.
If you want people to join in, prove something first. Write a killer-cool application with your language and blog/talk about why your language is different/special/better.
Modularity of what, the language itself or the compiler? If you want to extend the language, a good spec will help (see #1.) The compiler should be designed with all the best practices in mind, which should help make it extensible.
I hear the Dragon Book is good for learning to develop compilers.
Your specification will be broken unless you write it hand-in-hand with the implementation.
If you think C++ would give you cleaner design and faster development, you should probably use it.
You will have difficulty getting anyone interested in a project unless there is something that runs and demonstrates what is unique about your language.
If you think your language will ever require a byte-code interpreter (and you do say "Performance is a must") you should investigate the capabilities of existing byte-code interpreters before you finalize your language design.
I think you have set yourself too many goals. You say "performance is a must" but in a comment reply you say your goal is "to learn a lot about language design" and that it is "pretty unlikely" that you'll use it in a real project. New programming languages are created to solve problems; more precisely, they're created to help people express solutions to problems in better ways. Designing a language without using it seriously, intensely, continually is like writing software without any test cases: you're likely to wind up with something unusable.
If you want to try your hand at language design, then find a problem---one that you care about---that existing languages won't let you solve the way you want. Then do whatever you can to get a working implementation and start writing and running programs using it. You don't need a hand-crafted JIT compiler with a runtime written in highly bummed assembly code. If you target the JVM or .NET, you get a very high-performance GC, scalable threading system, libraries, and lots of other good stuff for free, even if it interferes with that awesome idea you had for ______.
On the other hand, if you just want to make something run fast, don't try to design a language at the same time. Just find one that you like, learn about implementation strategies, and see if you can do better.
Closed. This question is opinion-based. It is not currently accepting answers.
Want to improve this question? Update the question so it can be answered with facts and citations by editing this post.
Closed 4 years ago.
Improve this question
One of the key properties to designing comprehensible software (and, indeed, designing anything at all) is to develop a good set of abstractions. These days, those abstractions include things like functions, classes, interfaces, recursion, and higher-order functions. But what else is there? How can we further abstract our designs, so that I needn't think about anything but my immediate, direct goal? What novel abstractions have yet to be leveraged by existing technologies?
Also note that most of the items on my list (with the exception, perhaps, of recursion) are also tools used for code reuse. Code reuse is not the subject of this question, and is not what I see as a necessary aspect of a good abstraction. Functions are useful as abstractions because they hide what they are doing behind a descriptive name, not because I can call them from several different places.
A poorly-formed idea: Is a driver function that only calls a sequence of other functions, without maintaining any state of its own, really the same as a function? We write it as a function, and call it as a function, but perhaps it represents a different concept? This is reflected in some languages by making a distinctions between procedures returning values and procedures not returning values. But maybe there's a better way to view that difference, some different way to abstract the sequence of relatively unrelated steps?
So to reiterate, how can future programming languages better facilitate abstraction?
A powerful absraction tool, Lisp macros. Why not look into the past and present? :)
They can use self-exposing semantics to better allow metaprogramming of the environment/language presented as the end-user interface. Mutable language semantics.
Some areas that I think are potentially fruitful:
Intentional Programming, or something similar. Charles Simonyi's company Intentional Software has been keeping pretty quiet for a while but is now starting to show some promising early demonstrations.
Functional Programming: ideas from functional programming are increasingly finding their way into more mainstream languages like Python, C# (Linq, lambdas, etc.) and even C++ (lambdas in C++ 0x). F# is becoming a first class .NET language with full support in Visual Studio. The rise of multi core development is another factor driving the wider adoption of functional concepts.
Domain Specific Languages (DSLs): closely related to the ideas behind Intentional Programming, Microsoft seem to be putting some effort into supporting DSLs as part of the .NET ecosystem.
Much more sophisticated IDEs. There are already some positive developments with refactoring tools in IDEs like Visual Studio and IntelliJ but I think there's a lot of room for progress in this area. Moving away from dumb text source files towards something more like an abstract syntax tree representation could make it much easier to work at a higher level of abstraction. Again, this connects with many of the ideas behind Intentional Programming.
By having built in detection of stupid ideas that, when tripped, lock the developer out of the IDE and refuse to let them code ever again.
OOP facilitates abstraction quite nicely. It's developers that come up with poorly formed ideas.
Let's see, how about if we make abstraction mandatory for every data type, and then provide ways of generalizing our abstractions over type parameters? Wait! I've just reinvented CLU. Do I get a Turing Award?
Anyone interested in the role of abstraction in programming should study CLU.
Eiffel code proofs. (warning: link to PDF!)
Functional programming, aspect oriented programming, design by contract and generally everything that takes us away from the dark age of imperative programming.
Also, I hope non - managed software development will cease to exist. C++ and other low level stuff makes me sad. :-(
I like my LINQ, my lambda operator, my extension methods and my fluent interfaces. Oh, and I love PostSharp.NET. And F#, but I guess it's very hard NOT to love F#. :-)
I will give an indirect answer. Before we can develop better constructions in programming languages, we must first understand the theory of abstraction.
Oh yes, there is an actual theory which predates modern computing, it is called category theory.
Closed. This question needs to be more focused. It is not currently accepting answers.
Want to improve this question? Update the question so it focuses on one problem only by editing this post.
Closed 7 years ago.
Improve this question
I've been programming for several years now and since then I have learned several concepts and techniques that have made me a better programmer (i.e. OOP, MVC, regex, hashing, etc). I feel too that by been able to learn several languages (basic, pascal, C/C++, lisp, prolog, python) I have widen my horizons in a very possitive way. But since some time ago I feel like I'm not learning any new good "trick". Can you suggest some interesting concept/technique/trick that could make me retake the learning flow?
A good paradigm shift always allows you to see things differently and become a better developer. I would suggest you read up on functional programming and maybe learn a functional language like Haskell or Scheme.
YAGNI (You Ain't Gonna Need It) and DTSTTCPW (Do The Simplest Thing That Could Possibly Work)
It's easy to spend a lot of time thinking about edge cases, and find that you've implemented something that's completely useless. I believe that a far better approach is to knock out a simple prototype, and then poke and prod it until you understand the domain well enough to create production code.
Recognize, however, that your prototype is going to evolve into production code whether you like it or not. So write it with that in mind.
Learning how to use your IDE and tools. This to me resulted in a far greater productivity increase.
For examples:
learning how to use a source level debugger
using tools like purify/boundschecker
fxcop
etc. I realize I am dating myself, but those were big steps. There are many more.
Any time you can change the way you think about a problem or solve a problem without having to undo previous work is HUGE gain. Process, tools, etc all can help with that. Don't limit yourself to finding silver bullet techniques for productivity gains.
Watching productive people work and getting them to tell you what they are doing and why is also invaluable.
If I'm honest, using, and learning a great framework like .NET has really increased my productivity.
I'm often amazed what people are willing to reinvent due to their ignorance that the very same function already exists in the framework.
AGILE and especially Test Driven Development. Best thing to happen to software development since the invention of Object Oriented Design.
Concerning coding, I'd say design patterns and architecture patterns are always nice to look at and can help you write cleaner/better code.
For methodology I would advice Agile development that is great. There are a numerous number of techniques and methods (I'm personally fan of extreme programming) and reading that can keep you busy and improve your general approach.
Finally I'd say learn new languages like Ruby
Design patterns
SCRUM process
DiSC assessment (and understanding of how it applies to collaborative s/w development)
StackOverflow.com (of course!)
Google
... other stuff too, I'm sure
Design Patterns. Learning how to break dependence upon implementation and inheritance, and depending on interfaces (contracts) instead changed the way I think about programming.
Debugging. Once I figured out how to actually step through the code and go line-by-line, examining the underlying state, it revolutionized how I troubleshoot code.
Practice, practice practice: I didn't realize how important it is to keep working on my skills apart from work until a relatively short time ago. Mistakes and solutions I make at home make me a better programmer at work, and vice a versa. Learning should never stop if you want to be good at something, and programming isn't an exception.
If I had to pick just one, I'd say Test-Driven Design, aka TDD: write unit tests (and check that they fail) before you incrementally add features.
Try to learn to see things from the user's standpoint.
For example:
learn how to write meaningful error messages
learn how to produce usable applications
learn some basic speed-optimization techniques
Remember that the user sees your application, not your code.
VIM Quick Reference Card. After I started using advanced vim (macros, plugins) I have stopped doing any repetitive actions during coding manually.
Apart from that, Scrum and working at night, when noone interrupts You gave me the highest benefit.
If you want to expand your experience into web programming, you should try and get a good handle on the HTTP Request/Response paradigm. This will make creating web apps much easier on you because you understand the underlying framework.
(http)://en.wikipedia.org/wiki/Hypertext_Transfer_Protocol
I would look at some of the newer languages that combine OO and functional elements, like C# or Scala.
Learning Smalltalk has helped me become more productive. It is an easy language to learn and things can be built extremely quickly. For a stunning productivity aid check out Seaside, it's a framework for building web applications. Moreover, if you have only been used to curly brace languages Smalltalk will also make you smile!
I was helped by the following paradigms in this order:
1) bottom-up programming
2) top-down programming (C, Pascal)
3) object-oriented programming (Smalltalk, Java)
4) functional programming (lisp, Mathematica)
with some logic programming thrown in (prolog).
nHibernate hands down. The fact that I dont need to write database functionality for my business objects is very useful and time saving.
High level understanding, creating good abstractions with proper dependencies, is what pays off in long term. For example, Law of Demeter is an important guideline. I recommend also reading Eric Evan's Domain Driven Design
Code generators. They're the best thing in software engineering.
Would you like to write all your projects in asm? Nope, let's generate it from C++. Or from something sat above the JVM which diligently generates the necessary machine code.
Duplicating the same source code all over the place, but stuck with a language that insists on the line noise? Use macros.
Want to use lambdas in a language that doesn't have them? Work out how to fake the anonymous name and variable scoping required then generate the boilerplate.
None of the readily available languages quite fit your pattern of thought, desired syntax or even semantics? Write a compiler for a new one.
Better languages are nice. Better design patterns are nice. Emacs is awesome. But compilers are where all the power lies in our field. I suspect the only reason they aren't mentioned in any of the other answers is that we can't imagine programming without any.
Copy/paste technique
Closed. This question needs to be more focused. It is not currently accepting answers.
Want to improve this question? Update the question so it focuses on one problem only by editing this post.
Closed 8 years ago.
Improve this question
Functional languages are good because they avoid bugs by eliminating state, but also because they can be easily parallelized automatically for you, without you having to worry about the thread count.
As a Win32 developer though, can I use Haskell for some DLL files of my application? And if I do, is there a real advantage that would be taken automatically for me? If so, what gives me this advantage? The compiler?
Does F# parallelize functions you write across multiple cores and CPUs automatically for you? Would you ever see the thread count in Task Manager increase?
How can I start using Haskell in a practical way, and will I really see some benefits if I do?
It seems like the book Real World Haskell is just what you're looking for. You can read it free online.
F# does not contain any magic pixie dust that will pass functions off to different CPUs or machines. F#/Haskell and other functional programming languages make it easier for you to write functions that can be processed independent of the thread or CPU they were created on.
I don't feel right posting a link here to a podcast I participate in. It seems a little off, but in the Herding Code episode, where we talked with Matt Podwysocki, we asked the same question and he gave some interesting answers. There are also a lot of good links relating to functional programming in that episode. I found one link titled "Why Functional Programming Matters". That may provide some answers for you.
This might also be interesting:
"Real World Functional Programming"
Examples are in F# and C#, but the theory is fairly generic.
From what I've read (pre-release) it is definitely interesting, but so far I think it is making me want to stick more and more with C#, using libraries like Parallel Extensions.
You didn't mention, but I'm assuming, that you're using C++. One potentially easy way to get into functional is via C++/CLI to F#. C++ contains "magic pixie dust" (called IJW: It Just Works) to allow you to call into and out of managed code. With this, calling F# code is almost as simple as it is from C#.
I've used this in one program (FreeSWITCH), which is written entirely in C/C++. With a single managed C++/CLI (use the /clr switch), it magically transitions into managed code, and from there, I can go load my F# plugins and execute them. To make things even easier for deployment, F# can statically link all its dependencies, so you don't need to deploy the F# runtime files. One other thing that makes CLR code attractive is that you can pass managed code (delegates) to C code, and the runtime automatically makes a thunk for you.
If you decide to go the Haskell way, the feature you'll be looking for is FFI: Foreign Function Interface. However, I don't think it'll give you the same level of integration as C++/CLI with F#.
I'm currently learning Haskell myself. When you start out learning it, it doesn't seem very intriguing, because the learning experience is nothing like learning a language like C#.
It's a whole new world, but I noticed I could write very very complex expressions in just a few lines of code. When I looked back at the code, it was much more concise; it was small and tight.
I'm absolutely loving it! You can indeed write real-world programs that will be smaller, easier to maintain, and much more complex than most other languages allow. I vote for you to learn it!!
Since you mention Win32 and DLLs, I presume you're working with unmanaged code. In that case, GHC will work very well for you. Late last year I wrote a DDE server under Windows using FFI to talk to the Microsoft DDE libraries, and, surprisingly, it was an extremely pleasant experience (especially given that I'm a Unix guy).
Haskell's FFI is powerful (even supporting, e.g., callbacks into Haskell functions from C or other libraries), and having Haskell's type checking when writing C-level code is like a dream come true.
That last point is one of the major advantages of Haskell: the type system is amazing. That said, it's like any powerful tool; it needs time and effort to make good use of it.
So yes, it is possible to start out writing small bits of code in Haskell that link into the rest of your code (though you may find it easier to start with small Haskell programs that link to your other code), and it's well worth spending a fair amount of time learning about this and using it wherever you can. You may end up like me, planning a fairly major project tightly integrated with Windows code (in my case, a sophisticated Excel add-in) in Haskell.