Learning Lisp The Bump Free Way
blog.ppenev.com
blog.ppenev.com
LispIDE is another Windows option. It is very minimal but doesn't require dealing with emacs.
One tool that got some love at one point was Cusp, a Lisp plugin for Emacs. Unfortunately, as I note here[1], it's been unloved for some time.
You know, Sublime Text integration might be possible. But Eclipse is, as near as I can tell, the canonical mainstream GUI IDE. I think getting Swank(SLIME) integrated into a mainstream IDE would add to Common Lisp's usability for the non-command-line crowd by leaps and bounds.
I see the computing world as fundamentally bimodal: people trend hard towards GUIs and mice or they trend hard towards CLIs and keyboards. Common Lisp's development environment is built by and for the second crowd. You might call them... hackers. :-) But I would dearly love to see the non-hacker computing world walk in the wonders of Lisp. I've attempted to teach several people Lisp, and the development environment is, hands down, the major stumbling point (after the parentheses whining is finished with). This is a major barrier to getting people onboarded. I suppose that means I should try to get Cusp/lispdev compiling this weekend and see what state its in. :-)
Other barriers include no modern GUI library[1]; no popular webapp framework[2], no single point of contact for the language, and a lack of a flagship thing to raise the flag around (Ruby/Rails, Python/NumPy|Django, Clojure/not-java, etc).
[1] The libraries that exist are fairly thin bindings over QT, SDL, and TK. They are far behind the sleek usability of, say, Clojure's seesaw or the even clunkiness of C# autogenerated code.
[2] Clack and children are getting there
I no longer feel lost in Emacs. I was able to anchor it to my background with AutoCad - and there are huge architectural similarities between Emacs and pre-Windows versions of AutoCad. Both have a command driven interface.
The difficulty with the Emacs tutorial is that it is written around social mores. Thus it teaches the shortcuts which distinguish emacs community insiders from outsiders [1]. Yes, no sane experienced Emacs user will type:
M-x next-line
C-n
Why not Scheme?
As for why CL is preferred - it's probably about the ecosystem, tools and libraries. No Scheme and very few languages can compete with CL in this matter. However there is Rachet, a Scheme that has a chance to approach CL in the next decade (only my opinion).
Scheme is itself not a single language; it's a subcategory of Lisps. The main Scheme standard these days is R6RS (http://www.r6rs.org/), but different implementations may or may not comply with R6RS (Racket does not, for example, even though it was formerly known as PLT Scheme).
Common Lisp is the most well-known Lisp-2, and is what's generally implied when people say 'Lisp', even though that's technically a sloppy use of terminology.
Clojure is another dialect, but people generally specify Clojure and Scheme by name when referring to them. There is no logic behind these nuances of the terminology and their inconsistent usage; it's just what's evolved over the last 50+ years.
So, to answer your question, there's no reason why one should learn Scheme over Common Lisp (or vice versa), if all you're looking to do is learn the concepts.
There are differences between the different dialects (and their many implementations), but if you're looking to expand your knowledge and not necessarily write production code, pick Scheme (using Guile) and Common Lisp (using Clisp or SBCL) and you're good to go.
So, to answer your question, there's no reason why one should learn Scheme over Common Lisp (or vice versa), if all you're looking to do is learn the concepts.
As I discovered when I was looking for TAGBODY in scheme (don't ask; it seemed to be the most convenient way of solving the problem). Then I realized I didn't need it, I could get the same effect with internal definitions (internal function = tag) and calls. A call in tail position IS essentially a goto. (More powerful though as it also binds arguments, if any.)
Scheme also looks more conceptually clean. For example, (define add +) does exactly what you think. In CL, you'd have to do something more contrived.
Another example is the member function. member checks to see if an element is in a list. In CL member takes an element, a list as well as several keyword arguments that control how the searching is conducted. One such keyword argument is test which is used to compare the element to the items in the list, and another one is key, which is called on the items in the list before they are compared to the searched element. In scheme you don't have the key argument, which is already a drag, and instead of having a test argument, you have 3 functions: member, using equal?, memq which uses eq?, and memv, which uses eqv? for comparison. If you need anything more than that, in scheme you'll have to write your own such function. in CL, you just pass two functions to member to control how it works. In that way, CL is actually much more conceptually simpler than scheme, but we have to keep myths alive, don't we :)
Another example is the CL concept of a generalized boolean. In CL nil is false, and everything else is true, with T being the canonical true value. In scheme you have special values representing true and false, which again complicates code a bit, but is more conceptually simpler. I consider this to be a great design blunder on schemes part.
Another example is that CL is very much an OO language, much more than an FP one. You have classes, methods, inheritance, mutable slots, and all sort of messy and complicated things that come in handy every once in a while.
I can go on for ever, but I won't even get to clojure, which is I think much more interesting. Clojure emphasizes immutability and laziness. Neither CL nor Scheme force you to be use immutability, but clojure does everywhere it makes sense to do so. Laziness is something else clojure emphasizes. And although there are libraries for CL to make it very much a clojure like language, it isn't the default.
However, Clojure strays from a number of standard common idioms in other Lisps, which, IMHO, makes Clojure a less-than-ideal candidate if you're looking to get the "full SICP" experience. It has a lot of other benefits as a language, but for a learning experience, you'll miss out on some of the idioms common to other Lisps.
But, give it a shot and see if it works for you. Once you've learned one, it's really to learn another dialect.
And, in the end, the only wrong answer to 'Which Lisp?' is the empty list[0]. :)
[0] This is an example of a super-nerdy Lisp joke which works in Common Lisp but not Clojure - in Common Lisp, 'nil', 'false', and the empty list are all the same value, (and all other values are considered "truthy"). In Clojure, they are three distinct entities, and there is a discrete "true" value.
Clojure:
user=> (cons 3 4)
IllegalArgumentException Don't know how to create ISeq from: java.lang.Long clojure.lang.RT.seqFrom (RT.java:494)
user=> (cons 3 '(4))
(3 4)
CL-USER(1): (cons 3 4)
(3 . 4)
CL-USER(2): (cons 3 '(4))
(3 4)
CL-USER(3):If you want to know more about common lisp, I have a site for that: http://articulate-lisp.com
Scheme is case sensitive by default though, which I like.
It's sort of like asking 'Why C++ and not C?'. It's the features or lack thereof that you have to want or not want that lead you to your initial decision.
Then you switch to the other to see what you're missing.
Then you make up your mind.
It has been stable for a good long time. Which is nice if you've got a giant repository of old code that you dust off and run from time to time.
CL has a standard way of specifying variable types when you want to do so, so it's possible to write high performance numerical code in a portable (across different compilers) manner. Or write macros to write portable code.
My view on macros is that being code transformers, when you write a macro you are basically writing a quick and dirty compiler. And like everything that's quick and dirty, they are full of bugs and gotchas. The fact that CL has separate function and value namespaces helps with decreasing the gotcha cross-section. The scheme community early on spent a lot of time dealing with this and developing hygienic macros, but as a lisper this just always felt a bit heavy. (This is probably just a lack of exposure).
From my perspective, the appeal of scheme as been its more pure focus on functional programming. But when I'm in that sort of mood I just reach for haskell now.
I think CL is one of the few languages that I can see myself using for a decade, continuously, without having to go back and refresh pieces to conform to the latest version & compiler. Not only that, you can essentially reprogram it to keep up with advances in the field.
1) CLOS for object-oriented programming;
2) LOOP as a DSL for expressing complex iteration (http://www.gigamonkeys.com/book/loop-for-black-belts.html);
3) Type declarations and a compiler (CMUCL/SBCL) that can take full advantage of them;
4) Better tooling (CMUCL/SBCL + SLIME);
5) More commercial support options (Clozure, LispWorks, Allegro).
R5RS doesn't even have hash tables! You can patch together alternatives for Scheme for most of those things, but the fact that they're not built-in causes problems when integrating code from multiple sources.
R6RS does, though.
> You can patch together alternatives for Scheme for most of those things, but the fact that they're not built-in causes problems when integrating code from multiple sources.
R6RS has a lot more built-in than R5RS.
Unlike Common Lisp, which has a well-defined standard.
Unlike Clojure, which, while evolving, is at least compatible with itself.
Scheme code is so hopelessly unportable,that you can not even talk about "writing Scheme". You are either writing Gauche, or Chicken, or Racket, and once you write it, your code stays in that tiny little ecosystem with not more than 10 different users each.
Of course, you do know that the Arc language is racket code right? The site you are on is using racket ;)
It's incredibly unreadable :)
The content looks good, though, but I have to admit I didn't make it through the whole thing.
On Linux the xcalib package is an absolute godsend, white backgrounds are like being yelled at to these eyes.
Windoze has a similar screen inversion application, but cannot remember what it is right now.