Lambda the Ultimate

Relations of Language and Thought: The View from Sign Language and Deaf Children provides an interesting angle on the Sapir-Whorf hypothesis that we periodically discuss on LtU. A small sample from Google Books is available.

...Hypothesis concerning language and thought...:

• Language equals thought. Perhaps the simplest view of language and thought is that they are essentially the same thing. This position is most frequently ascribed to American behaviorists, and especially to John Watson, who argued that thought is just subvocal speech.
• Language and thought are independent. This view, most often attributed to theorists like Noam Chomsky and Jerry Fodor, suggests that the development of language and the development of cognition are distinct, dependending on different underlying processes and experiences.
• Language determines thought. In the form usually identified with the linguistic determinism and linguistic relativity theories of Sapir and Whorf, this perspective directly entails a correlation between language skill and cognitive skill. One implication of this view is that individuals who have "inferior" (or superlative) language are expected to have "inferior" (or superlative) thought. Implicitly or explicitly, such a perspective has been used as a rationale for an emphasis on spoken language for deaf children by those who have seen sign language as little more than a set of pragmatic gestures.

...The more interesting question... is whether growing up with exposure to a signed language affects cognition in a way different from growing up with a spoken language. Indeed, that is one of the fundamental questions of this volume. While we fully agree... that any strong form of the Sapir-Whorf position appears untenable, it also seems clear that language can affect and guide cognition in a variety of ways. Much of what a child knows about the world, from religion to the habitats of penguins, is acquired through language.

Sign language is an obvious candidate for linguistic study, since the mode is visual as opposed to oral/aural. The summary of one of the authors is telling:

The conclusion that American Sign Language (ASL) is an independent, noncontrived, fully grammatical human language comparable to any spoken language has been supported by over 30 years of research. Recent research has shown that ASL displays principles of organization remarkably like those for spoken languages, at discourse, semantic, syntactic, morphological, and even phonological levels. Furthermore, it is acquired, processed, and even breaks down in ways analogous to those found for spoken languages. The similarities between signed and spoken languages are strong enough to make the differences worth investigating. In the third section of this chapter, I will argue that although there are differences in detail, the similarities are strong enough to conclude that essentially the same language mechanism underlies languages in either modality.

On a programming language level, I can't help but think that sign language offers valuable clues into the nature of visual PLs (though I haven't quite nailed down any specifics). ASL on Wikipedia informs us that signs can be broken down into three categories:

• Transparent: Non-signers can usually correctly guess the meaning
• Translucent: Meaning makes sense to non-signers once it is explained
• Opaque: Meaning cannot be guessed by non-signers

With the majority of signs being opaque. As much as those who design visual languages would like them to be intuitive - falling into the Transparent and Translucent category - I figure you still have to end up using many signs that are only meaningful internally to the language at hand.

On a personal level, I have recently been attempting to delve into ASL. I've almost got the alphabet and numbers down, and have a vocabulary of about 100 additional signs - which probably means that I'm at the proficiency level of somewhere between ankle biter and sesame street. I do find it to be a fascinating language. I noticed when I was looking at the course offerings for college (my son started university this year) that ASL is now offered for foreign language credit (wish it had been offered when I was a student all those years ago).

Just finished reading In the Land of Invented Languages, by Arika Okrent. It makes an accessible read for many topics in language.

Natural languages may be less universal than music and less precise than programming languages, but they are far more versatile, and useful in our everyday lives, than either. Ambiguity, or fuzziness of meaning, is not a flaw of natural language but a feature that gives it flexibility and that, for whatever reason, suits our minds and the way we think. Likewise, the fact that languages depend on arbitrary convention or cultural habit is not a flaw but a feature that allows us to rein in the fuzziness by establishing agreed-upon meanings at different levels of precision. Language needs its "flaws" in order to do the enormous range of things we use it for.

Aside from this passage, the book barely mentions PL's at all. But programming languages are, by definition, invented languages (and, no, Perl does not qualify as natural), I think there are many parallels to be drawn. Most language inventors don't do it for the money - creating PL's is not a way to untold wealth. And there are a thousand invented (and programming) languages, so the chance of success is rather slim (and mostly accidental). The book itself is more an informal narrative that goes between personal experience, to examining the persons behind the languages, and on to a more critical analysis of the languages spotlighted. Although there are over 500 listed languages in the appendix, there is only in depth coverage of a dozen or so. The major periods covered:

• Enlightenment: John Wilkins and his Philosophical Language are the main subject of this period. The 17th century saw the widespread adoption of mathematical conventions, and there was a belief that a language could be designed that removed ambiguity - words would convey meaning exactly as intended. That belief is still a central tenet in much PL design.
• Idealism: Here we have Zamenhoff and Esperanto trying to bring about peace, love and understanding by sharing a common language. A couple of WWs would tell us that such utopian visions were not quite achieved. But Esperanto has been the most successful invented language in terms of usage. Most of the languages of this period were designed to be easier to learn, and were a mixture of languages - rather than striking out in bold semantic/syntactic fashion. Of course, we have PLs that want to borrow features from many different sources and strive to be easy to learn. Then again, efforts to reduce the number of languages usually have the effect of just creating more languages.
• Symbols: Charles Bliss and Blissymbolics with emphasis on non-oral language in this section covering symbol language and sign language. Visual PLs is what I thought of here
• Logic: Brown and Loglan were started as a roundabout thought experiment for Sapir-Whorf. But the only answer it would provide would be: what if, instead of trying to get AI from programming languages, we used something like a programming language for speaking, writing and communicating in the large.
• Esoteric: Klingon and other Conlang's are discussed in this section, with the emphasis on language as art or puzzle. Esoteric PLs are similar in spirit.

Lot's of tangential topics that are fun (Chinese writing, Hebrew, Tolkien, etc) and covers some very colorful characters. Not sure if PL designers are quite so eccentric, though I suspect it's only because we are still early in the game for PL evolution.

Books on two of the languages that get a lot of airplay on LtU have made the finalist list for this year's Jolt awards.

Books Technical
* High Performance MySQL by Baron Schwartz, Peter Zaitsev, Vadim Tkachenko, Jeremy Zawodny, Arjen Lentz, Derek J. Balling (O'Reilly Media)
* Java Power Tools by John Ferguson Smart (O'Reilly Media)
* Programming in Scala by Martin Odersky, Lex Spoon, and Bill Venners (Artima Press)
* Real World Haskell by John Goerzen, Bryan O'Sullivan, Donald Bruce Stewart (O'Reilly Media)
* The iPhone Developer's Cookbook: Building Applications with the iPhone SDK by Erica Sadun (Addison-Wesley Professional)

Congratulations to Martin, Lex, Bill, John, Bryan, and Don!

Whether or not either book wins, it's quite a sea change that two sophisticated, statically typed functional programming languages with research origins are getting so much mainstream attention.

From the FAQ

How are the winners selected?

Our judges not only examine the standard criteria of audience suitability, productivity, innovation, quality, ROI, risk and flexibility, but also seek to honor products that are ahead of the curve. Jolt-winning products are universally useful; are simple, yet rich in functionality; redefine their product space; and/or solve a nagging problem that has consistently eluded other products and books.

As a result of a refereeing request, I took a look at the website of College Publications, and was very pleased with what I saw. It is a non-profit publisher founded by Jane Spurr and the omnipresent (at least in logic) Dov Gabbay. From their "About us" page:

...objectives...:

• To bring College Publications to a level with Oxford (OUP) and Cambridge (CUP) in terms of publishing capability.
• To provide the community with a non-profit making, highly prestigious publishing outlet that will break the monopoly that commercial publishers have.
• To publish books that can be purchased at reasonable prices, making information accessible to all.

Accordingly, College Publications can publish books at a range of 15–25 dollars per average 300 page book, giving substantially better royalties to authors than other publishers.

...

College Publications does not take copyright and authors are free to use their material elsewhere, and even to put the book on the web once sales have achieved profitability.

The existence of this publisher is very good for the kind of things that LtU stands for in much the same way that Logical Methods in Computer Science is, and supposedly community focussed organisations such as the ACM are not. Take a look at their PL offerings in their computing series, edited by Ian Mackie.

We last mentioned this book back in 2005, when the text was available as a series of drafts, and LtU user raould updated the post back in late August letting LtU readers know you can get it from MIT Press, but to be sure: the book is out. Not to play favorites, but Amazon currently has it new for $54 USD, and it retails for $75 USD. Powell's currently has one copy in stock as well (thanks Tim!).

And it's massive. At 1,322 numbered pages, it'll take me a while to get through, but I hope to post a review once I'm done. LtU readers might be interested to know that the book is based on and is now used in MIT's graduate programming languages course, 6.821.

Peteris Krumins has been posting his notes on MIT’s Introduction to Algorithms. The notes are valuable for anyone interested in working their way through the CLRS text and MIT Open Courseware videos.

I just finished watching the last lecture of MIT’s "Introduction to Algorithms" course. Having a great passion for all aspects of computing, I decided to share everything I learned...

Although not directly tied to programming languages, every PL has to eventually be able to express algorithms. Aside from Knuth, CLRS is probably the closest approximation to a comprehensive approach to algortihms. The text itself is language agnostic - the authors use their own brand of pseudo-code to describe the algorithms. This has the advantage of allowing the reader to focus on the algorithms at a higher level, rather than get bogged down in the specifics of any PL. The downside, at least in my estimation, is that the authors don't make it particularly easy to implement the algorithms in any specific PL. The pseudo code conflates common data structures (such as arrays) with properties/attributes that can be tagged with those structures. And some of the algorithms refer to variables that are outside of the scope of the function. Also, like Knuth, most of the algorithms are steeped in state, making it hard to implement them with functional programming approaches.

That said, the video lectures and the accompanying notes above are good resources for any that want to self-study CLRS. Here are the notes thus far:

• Lectures 1 and 2: Analysis of Algorithms
• Lecture 3: Divide and Conquer
• Lectures 4 and 5: Sorting
• Lecture 6: Order Statistics
• Lectures 7 and 8: Hashing
• Lectures 9 and 10: Search Trees
• Lecture 11: Augmenting Data Structures
• Lecture 12: Skip Lists
• Lectures 13 and 14: Amortized Analysis and Self-Organizing Lists
• Lecture 15: Dynamic Programming
• Lecture 16: Greedy Algorithms
• Lectures 17, 18 and 19: Shortest Path Algorithms
• Lectures 20 and 21: Parallel Algorithms
• Lectures 22 and 23: Cache Oblivious Algorithms

SICP gets many nods when it comes to introductory texts to programming and the study of PLs. I've been slowly working my way through SICP in a number of different PLs, most notably Oz and Alice ML. In that process, I've come across Eli Bendersky's methodical solutions to the SICP Exercises in a series of blog posts. His review of SICP is instructive of the role of the exercises:

A word about exercises in SICP. They are numerous, some of them are hard, but the exercises are the best way to really understand what the book tries to teach. In a manner, they’re as integral part of the book as the text itself. The exercises are very well prepared and lead the reader through the examples coded by the authors into greater understanding of the topics taught. At times, it feels like the exercises are specifically designed to force you to think about the essence of the topics, and not just grasp them superficially.

Highly recommended reading for anyone that is working their way through SICP. Unlike my own work, which concentrates solely on code, his explanations are quite good. He uses mostly Common Lisp for the solutions, though resorting to Scheme when it makes for more concise solutions.

Translating code from one programming language to another is a black art. Even if successful in capturing functionality, each PL has its own styles, idioms and community morals. Doing automated translations (which I have done) has more misses than hits. Doing it manually gets you closer but it can require an inordinate amount of time to get it just right. Even so, PL translations are something that I personally enjoy as it is particularly instructive in teaching the strengths and limitations of expressing different concepts (though I usually catch flak for violating the social values of the target language).

My latest postings into this gray area are translations of the remaining Little Books to Oz - consisting of The Little Schemer, The Seasoned Schemer, The Little MLer and A Little Java, A Few Patterns (previous LtU post on The Reasoned Schemer in Oz). The Little Books are the antithesis to recipe books. There's not much code here that can be plugged into a project. The aim is to systematically teach programming thought processes. The books are useful for those wanting to learn Scheme (or ML). But the lessons are also useful even if those are not your particular language(s) of choice. Such didactic material may not be everyone's cup of tea, but they do represent a unique manner in which to teach (and still hoping for The Little Haskeller).

Along a similar line, I've started in on Introduction to Algorithms in Oz. Previously, I made a weak attempt at Knuth which I'll get back to one of these years, but found that translating MIX to higher level languages was tedious and time consuming. The CLRS book is a bit easier to translate, but the language they chose to express algorithms in doesn't seem to map to any exact known programming language in the universe. The language is concise, which was their aim, but it takes some shortcuts and has some peculiarities. Also, like Knuth, the algorithms are very much oriented to having mutable state. (Purely Functional Data Structures is in my queue). Anyhow, I find it interesting that the authors of the two best known book(s) on algorithms chose to invent their own language rather than use an existing PL.

Paul McJones alerts us that the ACM posted PDF versions of some books in its Classic Books Series, which are available to anyone who creates a free ACM Web Account.

Among the currently available books, LtU readers are likely to be particularly interested in Hoare and Jones's Essays in computing science, Adele Goldberg and David Robson's Smalltalk-80: the language and its implementation, and Dahl, Dijkstra, and Hoare's Structured programming.

Long time readers will also know that I highly recommend Papert's Mindstorms: children, computers, and powerful ideas to anyone interested with the effect computers might have on education. Papert's Logo remains to this day the best children oriented programming language, but even if you disagree with me about this, his book is a must read.

A draft textbook by Theodore Sider aimed at philsophy graduate students that while not as technical as computer scientists are used to, may be of interest due to the explicit discussion of extensions (e.g., modal operators), deviations (e.g., multi-valued logic) and variations (such as the Sheffer Stroke) on basic propositional logic.

The book includes chapters on counterfactuals and two-dimensional modal logic that may include material new to PLT wonks.