Kent M. Pitman's Second Wind

Kent M. Pitman has already given you his first 11 answers to the questions you asked him about Lisp, Scheme, the creation of programming standards, and much more -- below are his answers to another eight (starting with answer #12). Thanks again, Kent.

12) Good texts for learning Scheme?
by drenehtsral

I have recently been working on learning Scheme in my spare time, with the eventual goal of writing a scheme based scripting system to run the guts of a massive adventure game/graphical mud sort of system, everything from environment simulation (predator/prey cycles, etc...) to 3d models (i.e. models will be geometry glued together by scripts so you could have trees that by a random seed and a growth level variable have grown over time and are unique to provide interresting landscape features). Scheme is appealing because it's simple, powerful, and adapts well to the idea of a threaded interpreter.

To further my goal of learning Scheme inside and out, I've been reading "The Little Schemer," as well as "Structure and Interpretation of Computer Programs." Do you have any other recommendations for good Scheme programming texts?

Kent M. Pitman: You can get a list of textbooks from Schemers website. If you can articulate a particular need or preference that you think should help narrow down the many available choices, I'd suggest posting a more specific inquiry to the comp.lang.scheme newsgroup.

13) Overlooked practical aspects of Lisp
by hding

Why do you think that people so often overlook many of the wonderful things in Common Lisp such as unwind-protect, the whole condition system (which you are of course closely associated with), and so on - things that make it very useful for day-to-day programming, and are there any such things that you'd particularly highlight, or conversely that you wish had become part of the standard but did not.

Incidentally, thank you for all of the insight so generously and regularly poured forth in comp.lang.lisp.

KMP: Well, people program with tools that are familiar to them. Unless Common Lisp is someone's first language, it'd be easy for them to overlook the things it contains that are not like the things they're familiar with. There's a certain irony here because often the reason people will leave a language for another language is that they've reached the limits of what they can do with the first language and they need more power. So you'd expect that they'd aggressively look for features of the new language that were different than the things they've used before. And probably some do. But you're right that others cling to the safety and familiarity of the operators they could just as well have found in the old language they left behind, and so in the process they miss out on what the language can offer them.

Fortunately, unwind-protect is finally (pardon the pun) present in Java. And some hints of the Common Lisp condition system made it into Java as well. So probably people who come to Common Lisp from Java will be inclined to seek out those capabilities. But there's a lot of other stuff there and I hope new users will indulge their curiosity and take the time to explore.

As to what we should have in the language, the main omission of note is some sort of system definition tool (the in-Lisp analog of make). It was a shame that we did a feature-freeze on ANSI Common Lisp in 1988 but didn't get the standard out until 1994, and the suggestion of including such a tool didn't come until the after-freeze period. All vendors offer such a facility, but programs would be more portable if there were a uniform solution.

There are also quite a number of things about Common Lisp that are available in the same or similar forms in nearly all implementations. Multi-tasking, sockets, database access, external function call, windowing, and so on. It wouldn't be bad to have included any of these, but the fact is that they weren't ready for standarization in 1988. At this point, though, I think other mechanisms than standards are the right way to proceed.

The Lisp community used to expect the delivery mechanism for new functionality to be a new language spec. But that requires working through consensus standards bodies. The problem is that, by their nature, standards bodies are synchronization mechanisms. The problem with synchronization mechanisms in a massively parallel world is that they slow things down. The world is not going to wait for us to slow down, so I think we need to evolve mechanisms that will keep up better with a degree of pace that is externally dictated.

I think this is an area where Lisp as a community has been slow to respond. There need to be community mechanisms for sharing the many great commercial and private packages people have been creating in Lisp, so that we can properly reap the cross-product benefits of our community's productivity. I see evidence that this is changing. The Common Lisp Open Code Collection (CLOCC) is one such mechanism that addresses open source code. I'd like to see similar mechanisms arise for the exchange of proprietary products as well.

As to my posts on the comp.lang.lisp newsgroup, thank you. I'm glad you enjoy them. Frankly, I always consider it a victory to hear I haven't bored everyone to death. In background I've been working on putting together several books on Lisp, but one never quite knows if one will finish such things. I regard comp.lang.lisp as a kind of insurance policy, assuring that at least some piece of what I have seen and done in my career gets transferred from individual memory to global group memory.

I think preserving individual experiences for history is quite important. In the future, this will happen naturally due to logs kept by online collaboration tools. But I'm especially worried about the records of what happened between about 1960 (the birth of programming languages) and 1994 (the birth of the web). Most of everything in that time range is recorded on paper and will eventually be lost. Looking back from the future, I expect it to be as confusing to figure out how the information society was born as it is to look back in a telescope to see the birth of the Universe. You'll get very close, but then you'll get to a point where you can see nothing. The informational big bang. I've been working on webbing all of my old hardcopy papers, and I hope others of that era will commit to doing the same.

14) Lisp - Scheme - ML
by Tom7

I know a lot of big academic (erstwhile) lisp shops, such as CMU, have transitioned away from lisp to ML [standardml.org] and relatives [inria.fr]. Some of the reasons we might give are:

• Sophisticated type systems, catching most bugs before your program is run, ensuring safety, etc.
• Much more efficient (http://www.bagley.org/~doug/shootout/craps.shtml), partly due to compilation strategies using types
• Increased modularity and abstraction
• Pattern matching, (subjectively) more natural syntax

In fact, I'm one of those people. I've been scoffed at by lisp fans, but most had never used ML. But I have an open mind, so, in the face of more "modern" languages, what advantages do lisp and scheme offer? Do you think that these advantages are fundamentally impossible to achieve in a typed setting?

KMP: First, I assume by "typed" you mean "statically typed." I think of Lisp as "dynamically typed." I think of most machine languages as "untyped." I've heard statically typed languages sometimes called strongly typed, and I sometimes use this terminology myself out of habit, but I've grown to dislike it because it seems to me that the issue of strength ought to refer to whether you check types, not when you do. The terms "static" and "dynamic" seem to me to better get to the heart of the matter.

To quote Abraham Lincoln, admittedly somewhat out of context, "People who like this sort of thing will find this the sort of thing they like." So to somewhat flippantly re-interpret Lincoln's remarks in a modern context, applying perhaps just a bit of obligatory political spin to the result: The fact that functional languages appeal to people who like functional languages is not a proof that functional languages are of general purpose appeal.

I think the real reason that CMU (or any university with a grant-based funding model) changed its direction is good sources of funding in research depend on saying you're doing something "new and different." Such a shift doesn't imply that the thing left behind wasn't "tried and true," but only that "tried and true" is not what gets research dollars. Research must constantly stir the mix, but that doesn't imply obsolescence to what came before. So don't read too much into that.

Answering each of your points in detail might require a whole article, but I'll touch on each in brief:

• Sophisticated type systems, catching most bugs before your program is run, ensuring safety, etc. Much more efficient partly due to compilation strategies using types.

  Actually, it's funny that you both mention the CMU project and then make this comment. Before moving away from Common Lisp, the CMU crowd was successful in demonstrating to the Lisp community's satisfaction that there were enormous opportunities offered by the Common Lisp language design in terms of type inferencing that still today go untapped by implementations. This is really a market issue, not a language design issue. The fact is that although other languages do a lot more type inferencing, vendors are not getting huge numbers of bug reports saying that better type inferencing is what stands between programmers and the commercial success of their product. Over time, I think you'll see more and more interesting type analysis done, but such work is always balanced against other needs of users, such as CORBA, COM, RMI, and web interfaces, for example, such as UI toolkits and debugging options. When I observe, as I often do, that languages are political parties, this is what I mean. They are each responsive not to the needs of the world, but to the needs of their constituencies. And the Lisp constituency, while it is not oblivious to the value of type inferencing, does not see that issue as its number one priority.

• Increased modularity and abstraction.

  This is quite a multidimensional space. I think Lisp provides great opportunities for modularity and abstraction that other languages do not. And yet, there are sometimes things I can't abstract as well as I wish. An example of a minor omission: Common Lisp's CLOS doesn't do protocol abstraction as well as Zetalisp's New Flavors; among other things, one can't declare that certain unimplemented methods are required. But with the use of the macro system and the Meta-Object Protocol (MOP), one can add this kind of thing. Further, the package system is missing certain kinds of inheritance capabilities I've often wished for, but I recently sat down and did the work of writing my own versions of defpackage for my own use, adding the capabilities I wanted in a way that my own tools can use, and I had no difficulty. For the most part, I've found the limitations of Common Lisp's abstraction capabilites to be incidental, and not deep, and I've found its syntactic reorganization capabilities more than capable of making up for it.

• Pattern matching.

  I think you're right that Lisp doesn't do pattern matching. Whether or not that's a good or bad thing is subjective. I think there are people who like pattern matching and people who don't. In fairness to Lisp, though, on the few occasions in my career where I've felt a strong need for pattern matching, I've been able to implement it easily. And, importantly, Lisp's syntactic adaptability has allowed me to make my personal implementation look as natural in the programs I write as if it were natively provided by the language; most other languages don't give me the syntactic control to be able to add new functionality in a way that feels appropriate to the language. So personally, I don't find this a strong negative; rather, I see it as an opportunity for you to create a layered library that supports the needs of yourself and others like you.

• (Subjectively) more natural syntax.

  I don't think you can make the case that much of any language has "natural" syntax. COBOL and HyperTalk gave this the fairest shot and there's a big difference even between them and any natural language. I personally find Lisp syntax remarkably natural in that it focuses on symbols that you could say out loud, marking them minimally to indicate grouping. Other languages contain lots of special-purpose markers like commas, semicolons, asterisks, and braces/brackets/parens that are used in quite nitpicky ways. All this to say that you're right on this one: it's subjective. And as such, I hope I can fairly dismiss this as an even draw.

15) Lisp in Mathematics Programming
by An Anonymous Coward

Gregory Chaitin has a book called "The Limits of Mathematics." In it he claims that mathematicians should love Lisp because Lisp is basically set theory, and all mathematicians love set theory. I wholeheartedly agree with this, one only needs to look at Chaitin's Lisp programs to realize how quickly and succinctly one can arrive at astonishing incompleteness results in mathematics. So we know Lisp is great for stuff like this, really researching a mathematical subject. Do you see Lisp continuing in this direction, showing and discovering theorems, or will it move into industry? Or has it moved into industry, and we just don't know it? Do the likes of NASA and JPL use Lisp and Scheme religiously? I would bet so.

KMP: Lisp may have started out as a way of addressing abstract topics like math (logic, calculus, prime numbers, etc.) and artificial intelligence, but it long ago made the transition to commercial applications. Both Scheme and Common Lisp have been and continue to be used in real-world applications that might surprise you. These include (but are certainly not limited to) applications in airline scheduling, commercial database manipulation, computer operating systems, bioengineering, web services, document format translation, and, yes, even space exploration. Franz, Inc. has created quite a nice page of Lisp success stories that I think expand on this much better than I could in the space I've allowed myself to answer this question. And speaking of NASA/JPL, they did a comparative study of Lisp vs Java and C++ that some might find interesting.

16) Scheme in CS
by An Anonymous Coward

It seems many of the more popular CS programs in the world use Scheme as a teaching language. A lot of times, students complain about this, saying they'd prefer to learn about C or another language that is considered "apt for industry." I used to be like this too, but have now discovered the error of my thinking. How have you convinced others that while the latest programs might not be written in Scheme, that it is worth a student's time to learn Scheme. Many seem stuck to the point that if they won't use it outside of school, they shouldn't learn it. How can we convince them otherwise, to become scholarly citizens instead of drones?

KMP: I think the thing to explain to a student is that the world is ever changing and that one cannot put ones eggs all in one basket. Furthermore, modern environments are often quite heterogeneous, with different languages and systems being used together cooperatively. Especially for a CS student, who often has the luxury of time that a person in the job world does not, I think it's worth taking time to learn as many different languages as possible. This not only exposes the students to alternate ways of thinking, but it also prepares the student to quickly change modes of thought or languages of expression later. Once on the job, one often can't afford the ramp-up time to learn a new language at the point it becomes necessary to use. Better to already know it and just have to "brush up".

One is much more likely to consider alternative approaches if one has a sense of what is involved in them; it's very easy to fear the unknown, even when the unknown might be of great help. So get to know as many things as you can while you can. Common Lisp and Scheme, which I regard as two very different languages, by the way, should definitely be among the things every student studies. But they should not be the only things the person studies. Like it or not, there is a lot the professional programmer needs to know to be really successful not just tomorrow, but for a lifetime.

As Oliver Wendell Holmes is often quoted as saying, "A mind stretched to a new idea never returns to its original dimensions." In order to stretch a student's mind, I recommend they make a list of "kinds of languages" and then learn as many different kinds as they can. Here are some that come to mind, though I'm sure others with different experience than me might reasonably contribute still others.

• A block-structured language, such as Algol or Pascal or Scheme.
• A line-oriented language, such as Fortran or BASIC.
• A rule-based logic language, such as Prolog or EMYCIN.
• An Englishy language such as HyperTalk or Cobol.
• A "stack" language such as PostScript.
• A "line noise" language (heavy emphasis on one-letter operators) like APL or Teco.
• A dynamic object-oriented language, such as Common Lisp or Smalltalk .
• A strongly-typed, statically analyzed language such as ML or Haskell.
• A dynamically-typed functional programming language, such as Scheme.
• A string or macro processing language, such as Perl or m4 or TeX or Teco.
• A database access language, such as SQL.
• An abstract high-level, assembly language, such as Java.
• A concrete high-level, assembly language, such as C.
• A low-level, traditional assembly language.
• A scripting language, such as Javascript.
• An interface-definition language such as Visual Basic, HyperTalk, or Javascript.
• A document-structuring language such as XSL or TeX.
• A language with a modern error system, such as Common Lisp or Dylan.
• A reflective/introspective language such as Common Lisp or Java.
• A symbolic programming language, such as Common Lisp or Scheme.
• A security-oriented language, such as Java or MOO.
• A language where both programs and data are fully persistent, such as MOO or HyperTalk.

I'm confident that if people make a sincere effort to learn each of these kinds of languages on their own terms, the Lisp community will get its share of converts. But further, I would hope that the people who come to the Lisp community bring with them the ideas of other communities so that we can incorporate, rather than hide from, the good ideas that we need in order to continually be moving forward.

17) A question for Kent
by MarkusQ

Do you have a maclisp manual I could borrow?

KMP:For those not familiar with Maclisp, it's a defunct dialect of Lisp that predated and strongly influenced the design of Common Lisp.

I've been working on webbing The Revised Maclisp Manual, which I had published on paper back in the early 1980's. It's not quite ready to go out yet, but should finally be ready sometime in the not terribly distant future. Probably a month or two. Watch the site maclisp.info for more information.

18) Open Implementations
by Martin Pomije

What is your opinion of the idea of Open Implementations from Gregor Kiczales? Do you think that his idea could help Lisp be more widely used?

You can see him giving a lecture about this idea here. [microsoft.com] The video is only available in Windows Media format on this site.

KMP: I hadn't seen Gregor Kiczales's talk on Open Implementations, so I enjoyed watching it. Thanks for the pointer!

The talk made me think back to various related ideas I've seen batted around for a long time, the earliest of which that I can recall is a short paper on something called "Capsules" (an object system where classes were allowed to have multiple implementations) by Richard Zippel back in the late 1970s or early 1980s at MIT. Often, especially in a university environment, people will make up such a concept, bat it around for a bit, and then go on to something else. There are some very interesting ideas there and I'm glad to see that they're being pursued seriously, especially by someone as thoughtful and talented as I know Gregor to be.

As a formalized area of study, this topic of "aspect-oriented programming" is new to me. It reuses some old ideas in new ways, and introduces some new ones along with it. I'm only just barely becoming conversant in the terminology, so I can't really speak to it from a theoretical point of view. But it looks promising. And from a practical point of view, I can note that I'm getting daily on-the-job training in it through my consulting relationship with The Software Smith. They're using Lisp as a vehicle to apply the principles of aspect-oriented programming, and the results they get are quite spectacular.

19) What was up with CLisp's "loop" form?
by Jayson

Did you can have anything to do or know who had anything to do with the "loop" form in Common Lisp? Why does it look and feel just like a FOR loop on C (from the Graham book):

(loop for x = 8 then (/x 2)
until (< x 1)
do (princ x))

This is one of by biggest minor nags about CLisp and I am very curious what was going through the committee's collective head. Didn't anybody balk at this enough to at least get the syntax cleaned up?

KMP: The example you cite is quite simplistic and if this were the only reason for using LOOP, we wouldn't have it. Lisp has a number of other iteration operators for doing simple loops like this. However, the reason for using LOOP is that it can represent much more complicated arrangements of iteration paths and collection techniques. I used to grumble a lot myself about how "un-Lispy" LOOP seemed, but over time I come to the belief that the benefits outweigh the costs. A loop like this:

(loop for x from 0
for y in some-list
when (good-result? y)
collect (list x y))

is easy to write and maintain, and much easier to explain than the equivalent, but more Lispy:

(do ((x 0 (+ x 1))
(y-list some-list (cdr y-list))
(result '()))
((null y)
(nreverse result))
(let ((y (first y-list)))
(when (good-result? y)
(push (list x y) result))))

The Common Lisp community likes to offer the traditional Lispy notations for places where they enhance readability, but we also offer alternative notations for situations where we've learned there's a call for it. We leave the choice of which style to use up to the individual taste of the programmer. Common Lisp is not a minimalist language offering only one way to do things or rigidly attempting to force people into a single programming paradigm.

By the way, this is a fundamental difference between the ANSI Common Lisp design philosophy and the Scheme design philosophy. The introduction to the Scheme specification states:

Programming languages should be designed not by piling feature on top of feature, but by removing the weaknesses and restrictions that make additional features appear necessary. Scheme demonstrates that a very small number of rules for forming expressions, with no restrictions on how they are composed, suffice to form a practical and efficient programming language that is flexible enough to support most of the major programming paradigms in use today.

By contrast, the charter for X3J13, the group that designed ANSI Common Lisp, stated the following in the X3J13 charter:

It will codify existing practice, provide additional features to facilitate portability of code among diverse implementations, and establish normative Common Lisp programming practice. The committee will begin with the language described in Common Lisp: The Language by Guy L. Steele Jr. (Digital Press, 1984), which is the current de facto standard for Common Lisp. Whenever there is a proposal for the standard to differ from Common Lisp: The Language, the committee shall weigh both future costs of adopting (or not adopting) a change and costs of conversion of existing code. Aesthetic criteria shall be a subordinate consideration.

In other words, the Scheme community is a very conservative community that is highly focused on keeping its language specification as highly aesthetic and minimal in size as possible. By contrast, the Common Lisp community is an industrial standard that is concerned with messier issues of compatibility, portability, and commercial need; while the Common Lisp community cares about aesthetics, it does not allow aesthetics to dominate practicality as a design criterion.

The relevance of this here is that the Lisp family of languages is made up of a number of smaller communities who share a few core ideas, but really have some very divergent points of view. Each is worthy of study in its own right. One should not, having looked at Scheme, assume they have good intuitions about Common Lisp, nor vice versa.

Comments

Large-scale Lisp projects?

[Anonymous Coward]

One-offs and small utilities are all fine and dandy, but are there really people implementing large systems using Lisp?

Thanks Kent!

[justins]

The quality of the information you've given in your answers here and in your other responses is really fantastic. This is infinitely more effective than most language "advocacy" and it has been really interesting.

repost: Classic /. language rant (author unkown)

[Ars-Fartsica]

You Work in a Fashion Industry

I've spent the last several years trying to explain to colleagues why
they should start using another obscure-but-good language, Eiffel, to no avail.
Here is what I have learned. Note that this is not about the pros and cons of
particular languages or paradigms, its about the way the programming language
industry actually works.

The language industry is dominated by network effects. There are major
costs with using a minority language, and for an individual project these
completely outweigh the benefits, even when the benefits are very large. Hence
it is generally far better to stay with a majority language. The costs of a
minority language include:

Support. Sure, you can get a GPL compiler for most languages, but on a project
you don't want to have your coders digging into the code trying to
fix a bug, you want them writing code. Support is something you outsource.

Performance. Every minority language claims to be faster than C, but often
isn't in practice. Whatever the truth, C and C++ are at least known
quantities. Maybe the minority language will be faster, maybe slower. If its
faster, well gee so what. If its slower then you have a major problem.

Tool support. These days even small projects start by drawing UML diagrams and
then converting these automatically into class templates. CASE
tool vendors don't support minority languages. Ditto for testing and
documentation tools. Little things like tying your compiler to your
configuration control manager might potentially be major headaches. Again, its
more risk that the PM can do without.

Nobody ever got fired for buying C/C++/Java. If you are a PM this is a major
issue. Every language is going to bring some headaches, but if you have chosen
a minority language then these headaches can be turned into an excuse for
project failure, and hence for hanging you out to dry.

Trained staff in a minority language are going to be rare. This does not
necessarily make them more expensive (nobody else wants them), but it
does make recruitment much harder and more uncertain. Alternatively you have to
train all your existing people in the new language. And for Functional
Languages its not just another syntax, its a whole new way of thinking. The
industry went through this with OO languages, and many PMs have vivid memories
of reams of non-OO obfuscated C++ written by a bunch of C hackers who had been
sent on a one week C++ course. Getting your head around a new paradigm can take
months, and this is time that the project just does not have.

So, overall the PMs want to go with popular languages, not for PHM
reasons, but for entirely rational local reasons. But rational local decisions
turn into globally arbitrary decisions, as the entire herd gallops off in a
random direction chosen only because most of the herd thought that most of the
herd were headed that way.

The lesson of this is that if you want to introduce a language, you don't
concentrate on making it a good language, you try to persuade the herd of
programmers, PMs and tool vendors that your language is the Next Big Thing. The
important point here is not how much the language will do for productivity,
quality and cost, it is to create the perception that everyone else thinks that
this language will be the next big thing.

There are two ways to do this. One way is to tackle the whole industry at once.
For an object lesson in how to do this, see Java. For an object lesson
in how not to do it, see Eiffel. Believe me, I know all about this. I have
spent a long time giving presentations extolling the technical virtues of
Eiffel, only to have my audience say "Yes, but in the Real World....". In the
Real World what counts is the network effects. And you know what? My audiences
were right. It has taken me a long time to realise this.

The other more interesting and more promising way to introduce a new
language is to identify a niche market and attack that. Once you have taken
over your niche you can expand to nearby niches and start to build momentum.
Python is doing exactly this in web serving, for example. Web serving is a good
niche because lots of people do it, and productivity and quality generally
count for more than raw performance. Projects also tend to be small, so
experiments are not the Career Limiting Moves they are for large projects.
Education can also be a useful niche if you can afford to take the long view,
which is how Pascal, Basic and Unix got started.

Re:Okay, enough Lisp advocacy

[mrpotato]

I agree so much.

I'm familiar with a few languages. I begun with C++, thought it was nice and fun. Then I learned Java: I thought it was great not to have to worry about memory allocation and pointers, etc. I thought Java was fun too. But damn the Java community is dry. Then I learned Perl: I was in love. You can build things quick, and the community around the language is just so alive, diverse and interesting. You can get help anytime, people just _love_ to teach you (and show off ;-).

And these time, I'm into Scheme. It's very different, very fun and rewarding. You get to see the world from a new angle. You get to stretch your mind.

And everytime I learn a new language it just helps me to better understand the others. And don't forget the fun-factor: it is what drives me to try new stuff, to learn more. Ah well, just random rambling from a happy CS student ;-)

Re:NASA comparative study

[frank_adrian314159]

I'm not that impressed with the NASA study... Perl would also likely kick C++ and Java's ass in this task.

If you took the time to read the paper, you'd notice that the study was based on an earlier one that did include Perl. They both did, indeed, kick C++'s ass.

However, Lisp still did somewhat better than Perl in terms of time spent coding the task, had less variability in the time it took to code the task, and used significantly fewer runtime resources than Perl, IIRC.

don't you think it a little churlish to denigrate Java as 'an abstract, high level assembly language'?

It's not?

Precedence and Associativity cause Unreadable Code

[SimHacker]

If you're just talking about readability, Lisp is much more readable than most other languages. Especially "popular" ones like C++ and Perl.

And readability has a whole lot to do with maintainability. Programs should not just be written for the computer to understand. Much more importantly, other people should be able to clearly read and unambiguously understand the code.

I've read and written quite a lot of Lisp and other Lisp-like languages, not only in school, but also in the real world. I've spent much time reading, understanding and modifying huge complex Lisp systems written by other people, and I don't think it's unreadable at all. Well written Lisp code is extremely elegant, and a joy to read.

Complaining about parenthesis is trivial, misguided nit picking, and totally misses many much more important points.

Don't like parenthesis? Then let's talk about how all the subtle levels of operator precedence and associativity rules effect the readability and writability of C, C++, Java and Perl code!

Can't remember the difference in precedence between "==" and "<="? Then you have to either look it up (and require the poor bastards reading your code to either remember it or look it up themselves), or use an extra level of parenthesis!

Now whose code has lots of unnecessary parenthesis? Not the Lisp code. Every paren is in there for a reason. But most C, C++, Java and Perl code is eather riddled with unnecessary parenthesis, or totally unreadable, brittle and unmaintainable, because it subtly depends on operator precedence and associativity.

Most competent programmers put in lots of extra parenthesis just to be sure, because code is changed over time by other people. Relying on operator precidence causes many subtle, hard to find bugs, because it's extremely easy to make and miss practically invisible mistakes. And text editors like Emacs can't give you any help with operator precedence and associativity, like they can with parenthesis.

The arguments for operator precedence that claim people expect it are ridiculous. It's an utterly arbitrary and capricious linguistic artifact foolishly copied from mathematics (which most people don't understand), and taken to a ridiculous extreme (with many additional subtle levels of precedence for operators that simply don't exist in standard mathematical notation).

http://www.howstuffworks.com/c14.htm
C contains many operators, and because of the way in which operator precedence works, the interactions between multiple operators can become confusing.

x=5+3*6;

X receives the value 23, not 48, because in C multiplication and division have higher precedence than addition and subtraction.

char *a[10];

Is a a single pointer to an array of 10 characters, or is it an array of 10 pointers to character? Unless you know the precedence conventions in C, there is no way to find out. Similarly, in E.11 we saw that because of precedence statements such as *p.i = 10; do not work. Instead, the form (*p).i = 10; must be used to force correct precedence.

The following table from Kernigan and Richie shows the precedence hierarchy in C. The top line has the highest precedence.

[I am forced to spell out the operators in order to get past slashdot's lameness filter -- and I totally agree that languages with lots of ridiculous noisy punctuation are lame. Save it for cursing in comic strips.]

Operators : Associativity
"open paren" "open bracket" "minus" "dot": Left to right
"exclamation mark", "plus", "minus", "plus-plus", "minus-minus", "star", "ampersand", "(type-cast)", "sizeof": Right to left
(in the above line, "plus", "minus" and "star" are the unary forms)
"star", "slash", "percent": Left to right
"plus", "minus": Left to right
"less-less", "greater-greater": Left to right
"less", "less-equal", "greater", "greater-equal": Left to right
"equal-equal", "exclaimation-equal": Left to right
"ampersand": Left to right
"caret": Left to right
"pipe" : Left to right
"ampersand-ampersand": Left to right
"pipe-pipe": Left to right
"question-colon": Left to right
"equal", "plus-equal", "minus-equal", "slash-equal", "percent-equal", "ampersand-equal", "caret-equal", "pipe-equal", "less-less-equal", "greater-greater-equal": Right to left
"comma": Left to right

Using this table, you can see that char *a[10]; is an array of 10 pointers to character. You can also see why the parentheses are required if (*p).i is to be handled correctly. After some practice, you will memorize most of this table, but every now and again something will not work because you have been caught by a subtle precedence problem.

-Don

Re:LISP LISP LISP

[brucehoult]

Actually, I'd argue that the genericity of lisp's syntax is as much a hindrance as it is a help, emacs parenthesis-matching aside.

The above quoted lisp sexp could mean almost anything or nothing, depending on the context in which that sexp occurs. (Is it data? Is it code? In what evaluation context will it be processed?). Java, C, and most other languages at least give you more distinct contextual tokens to guide you in your understanding.

I like Lisp a lot, but I tend to agree with this point of view. I'm simply more comfortable infix-syntax for simple arithmetic. Prefix syntax is great for function calls -- and I don't really care whether the parens go around the whole thing (Lisp), after the function name and aroudn the arguments (most languages), or aren't there at all (ML/Haskell/Logo). Hell, I even cope happily with postfix function calls (Postscript/Forth/GML).

But when it comes to control structures I really, really prefer a keyword-rich syntax and explicit markers saying just what it is that you've just come to the end of. e.g. if/then/elseif/fi or any of a number of equally reasonable alternatives.

This is why I think that Dylan is currently the best language out there. It does pretty much everything that Common Lisp does, but in a more familiar syntax and with a lot of CLs historical cruft cleaned up. Dylan was designed by a bunch of Common Lisp gurus with the intention of designing the replacement for CL. And I think they suceeded technically.

Dylan currently has two implementations, one open source batch compiler for Unixy systems and one commercial IDE for Windows (with a free personal edition). Both systems compile to fast machine code and can compete with C in all but the most hardcore applications -- certainly far far better than Java or Perl or Python.

Functional Developer is a quite mature system with nice IDE and debugger and lots of libraries.

Gwydion Dylan is a bit less polished, but it's still good enough that a team using it managed 2nd place this year in this year's ICFP 72 hour programming contest.

All previous winners at this contest have been written in either C or else one of the Hindley-Milner statically typed languages (SML/OCaml/Haskell), all of which have very fast execution speeds with a good compiler.

You don't necessarily have to have the best possible langauge to win at ICFP, but you must have no major weaknesses -- you have to have a language and compiler that lets you develop and debug complex code in a very short time, and the end result must run fast. Much like in the real work :-)

Common Lisp and Scheme have never yet won a prize here (though I believe Common Lisp could), and it seems that the large number of Java and Perl and Python entries produced every year simply don't run fast enough to do well. On the other hand, the vast majority of C and C++ programs get eliminated because of bugs (occasionally one gets through!).

Having said all that, the one-line summary is this: if you like the idea of Common Lisp but just can't stand all those parens, take a close look at Dylan instead.

Do your customers know?

[Ars-Fartsica]

Do your customers know that you are saddling them with code that will be much more expensive for them to maintain than a "majority" language?

Seriously, should you pass on and they need to hire someone else to maintain your LISP, they will certainly pay 2x or 3x the cost of a C programmer, as good LISP programmers are far more rare.

This isn't a joke - I know of companies that have acquired LISP code that has become a complete albatross as they struggle to hire or train people to maintain it - here's a hint - not many programmers want to bother with LISP as they see it as career-limiting.

All in all, I think offering LISP solutions to your customers is a lousy value proposition in the long term.

Re:If you learned it, you'd like it?

[NeonSquare]

Most of his arguments boil down to "if you learned LISP, you'd like it." But that's bogus. There are some (many?) of us who used LISP and discarded it. Most people who went through a good computer science school in the days before Java used LISP.

Most people who learned Lisp at some school seem to dislike it first. Learning something and getting something learnt is sometimes something completely different. The Lisp that you can learn on most schools is insanely crippled down to months of hacking low-level, useless list-fiddling programs.

Another problem is that most of these people did not see any need for many of the high-level features offered by lisp and therefore see no urge to use/learn/understand it.

The object system in Common LISP was an afterthought, and it shows. It's really a macro package kind of thing. LISP has a strong history of overly elaborate macros; ever see the MIT LOOP macro? These have the usual problem; what you're looking at in the source isn't what's running, and debugging macros (especially those written by others) is hell.

Many people have a bad taste in mind if a notion like "macro" is mentioned. It remembers them on crippled things like C Preprocessor macros , stupid little script languages or embedded pseudo-languages like VBA. But in Lisp a macro is in fact some kind of generalisation over functions. The reasoning is rather simple: 1) Programs are written to solve problems. 2) Writing a program is a problem 3) Solving 2 with 1 means simply to write a program that writes a program. 4) If still too complex goto 2

Most other high-level languages have banished macros, or discourage their use.

I don't think that that is really true. Most other high-level languages never had macros. Often enough the designers of other high-level languages do not even know what macros (in the Lisp way of sense) are...

Sometimes it takes more than one try

[alispguru]

Most of his arguments boil down to "if you learned LISP, you'd like it." But that's bogus.

I guess what Kent should have said more explicitly here was "most people's idea of Lisp comes from bad teaching by people who don't know and use the langauge. If you got to know Lisp properly, you'd like it more."

I was first exposed to Lisp as an undergrad, in probably the worst possible way. It was as part of the canonical language survey course, taught by someone who didn't really know Lisp, using an atrocious little monograph titled "Lisp for Fortran Programmers". List processing using only PROG, SETQ,and GO *shudder*. No macros, no arrays, no structures. And the implementation we used was essentially a batch interpreter - no emacs, no debugger, just a raw command-line read-eval-print loop.

When I saw Lisp again, in my second year of graduate school, it was taught by someone who used it to get real work done, with an implementation that had most of the features I have come to know and love.

A large fraction of current Lisp users (judging from comments in comp.lang.lisp) are people who started out knowing other languages, became frustrated with their instability or limitations, and went out looking for alternatives. Their choice of Lisp was not made in ignorance.

Lisp is not perfect, but it is an excellent balance of features in progrmming space. Lisp allows you to do imperative code, functional code, and OO code, without forcing you exclusively down one road or another. It is still one of the languages of choice when you want to do something truly new, and it will likely remain so.