ACM Sigplan Notices, Volume 31, Number 8, August 1996
Most people are surprised when I tell them what language we use in our undergraduate AI programming class. That's understandable. We use GAWK. GAWK, Gnu's version of Aho, Weinberger, and Kernighan's old pattern scanning language isn't even viewed as a programming language by most people. Like PERL and TCL, most prefer to view it as a `scripting language.' It has no objects; it is not functional; it does no built-in logic programming. Their surprise turns to puzzlement when I confide that (a) while the students are allowed to use any language they want; (b) with a single exception, the best work consistently results from those working in GAWK. (footnote: The exception was a PASCAL programmer who is now an NSF graduate fellow getting a Ph.D. in mathematics at Harvard.) Programmers in C, C++, and LISP haven't even been close (we have not seen work in PROLOG or JAVA).
There are some quick answers that have to do with the pragmatics of undergraduate programming. Then there are more instructive answers that might be valuable to those who debate programming paradigms or to those who study the history of AI languages. And there are some deep philosophical answers that expose the nature of reasoning and symbolic AI. I think the answers, especially the last ones, can be even more surprising than the observed effectiveness of GAWK for AI.
First it must be confessed that PERL programmers can cobble together AI projects well, too. Most of GAWK's attractiveness is reproduced in PERL, and the success of PERL forebodes some of the success of GAWK. Both are powerful string-processing languages that allow the programmer to exploit many of the features of a UNIX environment. Both provide powerful constructions for manipulating a wide variety of data in reasonably efficient ways. Both are interpreted, which can reduce development time. Both have short learning curves. The GAWK manual can be consumed in a single lab session and the language can be mastered by the next morning by the average student. GAWK's automatic initialization, implicit coercion, I/O support and lack of pointers forgive many of the mistakes that young programmers are likely to make. Those who have seen C but not mastered it are happy to see that GAWK retains some of the same sensibilities while adding what must be regarded as spoonful of syntactic sugar. Some will argue that PERL has superior functionality, but for quick AI applications, the additional functionality is rarely missed. In fact, PERL's terse syntax is not friendly when regular expressions begin to proliferate and strings contain fragments of HTML, WWW addresses, or shell commands. PERL provides new ways of doing things, but not necessarily ways of doing new things.
In the end, despite minor difference, both PERL and GAWK minimize programmer time. Neither really provides the programmer the setting in which to worry about minimizing run-time.
There are further simple answers. Probably the best is the fact that increasingly, undergraduate AI programming is involving the Web. Oren Etzioni (University of Washington, Seattle) has for a while been arguing that the "softbot" is replacing the mechanical engineers' robot as the most glamorous AI test bed. If the artifact whose behavior needs to be controlled in an intelligent way is the software agent, then a language that is well-suited to controlling the software environment is the appropriate language. That would imply a scripting language. If the robot is KAREL, then the right language is turn left; turn right. If the robot is Netscape, then the right language is something that can generate Netscape -remote 'openURL(http://cs.wustl.edu/~loui) with elan.
Of course, there are deeper answers. Jon Bentley found two pearls in GAWK: its regular expressions and its associative arrays. GAWK asks the programmer to use the file system for data organization and the operating system for debugging tools and subroutine libraries. There is no issue of user-interface. This forces the programmer to return to the question of what the program does, not how it looks. There is no time spent programming a binsort when the data can be shipped to /bin/sort in no time. (footnote: I am reminded of my IBM colleague Ben Grosof's advice for Palo Alto: Don't worry about whether it's highway 101 or 280. Don't worry if you have to head south for an entrance to go north. Just get on the highway as quickly as possible.)
There are some similarities between GAWK and LISP that are illuminating. Both provided a powerful uniform data structure (the associative array implemented as a hash table for GAWK and the S-expression, or list of lists, for LISP). Both were well-supported in their environments (GAWK being a child of UNIX, and LISP being the heart of lisp machines). Both have trivial syntax and find their power in the programmer's willingness to use the simple blocks to build a complex approach.
Deeper still, is the nature of AI programming. AI is about functionality and exploratory programming. It is about bottom-up design and the building of ambitions as greater behaviors can be demonstrated. Woe be to the top-down AI programmer who finds that the bottom-level refinements, `this subroutine parses the sentence,' cannot actually be implemented. Woe be to the programmer who perfects the data structures for that heap sort when the whole approach to the high-level problem needs to be rethought, and the code is sent to the junk heap the next day.
AI programming requires high-level thinking. There have always been a few gifted programmers who can write high-level programs in assembly language. Most however need the ambient abstraction to have a higher floor.
Now for the surprising philosophical answers. First, AI has discovered that brute-force combinatorics, as an approach to generating intelligent behavior, does not often provide the solution. Chess, neural nets, and genetic programming show the limits of brute computation. The alternative is clever program organization. (footnote: One might add that the former are the AI approaches that work, but that is easily dismissed: those are the AI approaches that work in general, precisely because cleverness is problem-specific.) So AI programmers always want to maximize the content of their program, not optimize the efficiency of an approach. They want minds, not insects. Instead of enumerating large search spaces, they define ways of reducing search, ways of bringing different knowledge to the task. A language that maximizes what the programmer can attempt rather than one that provides tremendous control over how to attempt it, will be the AI choice in the end.
Second, inference is merely the expansion of notation. No matter whether the logic that underlies an AI program is fuzzy, probabilistic, deontic, defeasible, or deductive, the logic merely defines how strings can be transformed into other strings. A language that provides the best support for string processing in the end provides the best support for logic, for the exploration of various logics, and for most forms of symbolic processing that AI might choose to call reasoning'' instead of logic.'' The implication is that PROLOG, which saves the AI programmer from having to write a unifier, saves perhaps two dozen lines of GAWK code at the expense of strongly biasing the logic and representational expressiveness of any approach.
I view these last two points as news not only to the programming language community, but also to much of the AI community that has not reflected on the past decade's lessons.
In the puny language, GAWK, which Aho, Weinberger, and Kernighan thought not much more important than grep or sed, I find lessons in AI's trends, Airs history, and the foundations of AI. What I have found not only surprising but also hopeful, is that when I have approached the AI people who still enjoy
programming,
some of them are not the least bit
surprised.
by Ronald P. Loui
Associate Professor of CSE
Washington University in St. Louis
(Pre-publication draft; copyright reserved by author. A subsequent version of this document appeared as
IEEE Computer, vol. 41, no. 7, July 2008).
This article's main purpose is to review the changes in
programming practices known collectively as the "rise of
scripting," as predicted in 1998 IEEE COMPUTER by Ousterhout.
This attempts to be both brief and definitive, drawing on many of
the essays that have appeared in online forums. The main new idea
is that programming language theory needs to move beyond semantics
and take language pragmatics more seriously.
To the credit of this journal, it had the courage to publish the signal
paper on scripting, John Ousterhout's "Scripting: Higher Level Programming
for the 21st Century" in 1998. Today, that document rolls forward with an
ever-growing list of positive citations. More importantly, every major
observation in that paper seems now to be entrenched in software practice
today; every benefit claimed for scripting appears to be genuine
(flexibility of typelessness, rapid turnaround of interpretation, higher
level semantics, development speed, appropriateness for gluing components
and internet programming, ease of learning and increase in amount of
casual programming).
Interestingly, IEEE COMPUTER also just printed one of the most canonical
attacks on scripting, by one Diomidis
Spinellis, 2005, "Java Makes Scripting Languages Irrelevant?" Part of
what makes this attack interesting is that the author seems unconvinced of
his own title; the paper concludes with more text devoted to praising
scripting languages than it expends in its declaration of Java's progress
toward improved usability. It is unclear what is a better recommendation
for scripting: the durability of Ousterhout's text or the indecisiveness
of this recent critic's.
The real shock is that the academic programming language community
continues to reject the sea change in programming practices brought
about by scripting. Enamored of the object-oriented paradigm,
especially in the undergraduate curriculum, unwilling to accept the
LAMP (Linux-Apache-MySQL-Perl/Python/Php) tool set, and firmly believing
that more programming theory leads to better programming practice, the
academics seem blind to the facts on the ground. The ACM flagship,
COMMUNICATIONS OF THE ACM for example, has never published a paper
recognizing the scripting philosophy, and the references throughout the
ACM Digital Library to scripting are not encouraging.
Part of the problem is that scripting has risen in the shadow of
object-oriented programming and highly publicized corporate battles
between Sun, Netscape, and Microsoft with their competing software
practices. Scripting has been appearing language by language,
including object-oriented scripting languages now. Another part of the
problem is that scripting is only now mature enough to stand up against
its legitimate detractors. Today, there are answers to many of the
persistent questions about scripting: is there a scripting language
appropriate for the teaching of CS1 (the first programming course for
majors in the undergraduate computing curriculum)? Is there a scripting
language for enterprise or real-time applications? Is there a way for
scripting practices to scale to larger software engineering projects?
I intend to review the recent history briefly for those who have not yet
joined the debate, then present some of the answers that scripting
advocates now give to those nagging questions. Finally, I will describe
how a real pragmatism of academic interest in programming languages would
have better prepared the academic computing community to see the changes
that have been afoot.
1996-1998 are perhaps the most interesting years in the phylogeny of
scripting. In those years, perl "held the web together", and together
with a new POSIX awk and GNU gawk, was shipping with every new Linux.
Meanwhile javascript was being deployed furiously (javascript bearing no
important relation to java, having been renamed from "livescript" for
purely corporate purposes, apparently a sign of Netscape's solidarity with
Sun, and even renamed "jscript" under Microsoft). Also, a handoff from
tcl/tk to python was taking place as the language of choice for GUI
developers who would not yield to Microsoft's VisualBasic. Php appeared
in those years, though it would take another round of development before
it would start displacing server-side perl, cold fusion, and asp. Every
one of these languages is now considered a classic, even prototypical,
scripting language.
Already by mid-decade, the shift from scheme to java as the dominant CS1
language was complete, and the superiority of c++ over c was unquestioned
in industry. But java applets were not well supported in browsers, so the
appeal of "write once, run everywhere" quickly became derided as "write
once, debug everywhere." Web page forms, which used the common gateway
interface (cgi) were proliferating, and systems programming languages like
c became recognized as overkill for server-side programming. Developers
quickly discovered the main advantage of perl for cgi forms processing,
especially in the dot-com setting: it minimized the programmer's
write-time. What about performance? The algorithms were simple, network
latency masked small delays, and database performance was built into the
database software. It turned out that the bottleneck was the
programming. Even at run-time, the network and disk properties were the
problems, not the cpu processing. What about maintenance? The developers
and management were both happy to rewrite code for redesigned services
rather than deal with legacy code. Scripting, it turns out, was so
powerful and programmer-friendly that it was easier to create new scripts
from scratch than to modify old programs. What about user interface?
After all, by 1990, most of the programming effort had become the writing
of the GUI, and the object-oriented paradigm had much of its momentum in
the inheritance of interface widget behaviors. Surprisingly, the
interface that most programmers needed could be had in a browser. The
html/javascript/cgi trio became the GUI, and if more was needed, then
ambitious client-side javascript was more reliable than the browser's java
virtual machine. Moreover, the server-side program was simply a better
way to distribute automation in a heterogeneous internet than the
downloadable client-side program, regardless of whether the download was
in binary or bytecode.
Although there was not agreement on what exact necessary and sufficient
properties characterized scripting and distinguished it from "more
serious" programming, several things were clear:
scripting permitted rapid development, often regarded as merely
"rapid prototyping," but subsequently recognized as a kind of agile
programming;
scripting was the kind of high-level programming that had always
been envisioned, in the ascent from low-level assembly language
programming to higher levels of abstraction: it was concise,
and it removed programmers from concerning themselves with
many performance and memory management details;
scripting was well suited to the majority of a programming task,
usually the accumulation, extraction, and transformation of data,
followed eventually by its presentation, so that only the
performance-critical portion of a project had to be written in a
more cumbersome, high-performance language;
it was easier to get things right when source code was short, when
behavior was determined by code that fit on a page, all types were
easily coerced into strings for trace-printing, code fragments
could be interpreted, identifiers were short, and when the
programmer could turn ideas into code quickly without losing
focus.
This last point was extremely counterintuitive. Strong typing, naming
regimen, and verbosity were motivated mainly by a desire to help the
programmer avoid errors. But the programmer who had to generate too many
keystrokes and consult too many pages, who had to search through many
different files to discover semantics, and who had to follow too many
rules, who had to sustain motivation and concentration over a long period
of time, was a distracted and consequently inefficient programmer. Just
as vast libraries did not deliver the promise of greater reusability, and
virtual machines did not deliver the promise of platform-independence, the
language's promise to discipline the programmer quite simply did not
reduce the tendency of humans to err. It exchanged one kind of frequent
error for another.
Scripting languages became the favorite tools of the independent-minded
programmers: the "hackers" yes, but also the gifted and genius
programmers who tended to drive a project's design and development. As
Paul Graham noted (in a column reprinted in "Hackers and Painters" or this), one
of the lasting and legitimate benefits of java is that it permits managers
to level the playing field and extract considerable productivity from the
less talented and less motivated programmers (hence, more disposable).
There was a corollary to this difference between the mundane and the
liberating:
scripting was not enervating but was actually renewing:
programmers who viewed code generation as tedious and tiresome in
contrast viewed scripting as rewarding self-expression or
recreation.
The distinct features of scripting languages that produce these effects
are usually enumerated as semantic features, starting with low I/O
specification costs, the use of implicit coercion and weak typing,
automatic variable initialization with optional declaration, predominant
use of associative arrays for storage and regular expressions for pattern
matching, reduced syntax, and powerful control structures. But the main
reason for the productivity gains may be found in the name "scripting"
itself. To script an environment is to be powerfully embedded in that
environment. In the same way that the dolphin reigns over the open ocean,
lisp is a powerful language for those who would customize their emacs,
javascript is feral among browsers, and gawk and perl rule the linux
jungle.
There is even a hint of AI in the idea of scripting: the scripting
language is the way to get high level control, to automate by capturing
the intentions and routines normally provided by the human. If recording
and replaying macros is a kind of autopilot, then scripting is a kind of
proxy for human decisionmaking. Nowhere is this clearer than in simple
server-side php, or in sysadmin shell scripting.
So where do we stand now? While it may have been risky for Ousterhout to
proclaim scripting on the rise in 1998, it would be folly to dismiss the
success of scripting today. It is even possible that java will yield its
position of dominance in the near future. (By the time this essay is
printed, LAMP and AJAX might be the new darlings of the tech press;
see recent articles in Business Week, this IEEE COMPUTER, and even James
Gosling's blog where he concedes he was wanting to write a scripting
language when he was handed the java project. Java is very much in full
retreat.) Is scripting ready to fill the huge vacuum that would be
produced?
I personally believe that CS1 java is the greatest single mistake in the
history of computing curricula. I believe this because of the empirical
evidence, not because I have an a priori preference (I too voted to shift
from scheme to java in our CS1, over a decade ago, so I am complicit in
the java debacle). I reported in SIGPLAN 1996 ("Why gawk for AI?") that
only the scripting programmers could generate code fast enough to keep up
with the demands of the artificial intelligence laboratory class. Even
though students were allowed to choose any language they wanted, and many
had to unlearn the java ways of doing things in order to benefit from
scripting, there were few who could develop ideas into code effectively
and rapidly without scripting. In the intervening decade, little has
changed. We actually see more scripting, as students are happy to
compress images so that they can script their computer vision projects
rather than stumble around in c and c++. In fact, students who learn to
script early are empowered throughout their college years, especially in
the crucial UNIX and web environments. Those who learn only java are
stifled by enterprise-sized correctness and the chimerae of just-in-time
compilation, swing, JRE, JINI, etc. Young programmers need to practice
and produce, and to learn through mistakes why discipline is needed. They
need to learn design patterns by solving problems, not by reading
interfaces to someone else's black box code. It is imperative that
programmers learn to be creative and inventive, and they need programming
tools that support code exploration rather than code production.
What scripting language could be used for CS1? My personal preferences
are gawk, javascript, php, and asp, mainly because of their very gentle
learning curves. I don't think perl would be a disaster; its imperfection
would create many teaching moments. But there is emerging consensus in
the scripting community that python is the right choice for freshman
programming. Ruby would also be a defensible choice. Python and ruby
have the enviable properties that almost no one dislikes them, and almost
everyone respects them. Both languages support a wide variety of
programming styles and paradigms and satisfy practitioners and
theoreticians equally. Both languages are carefully enough designed that
"correct" programming practices can be demonstrated and high standards of
code quality can be enforced. The fact that Google stands by python is an
added motivation for undergraduate majors.
But do scripting solutions scale? What about the performance gap when the
polynomial, or worse the exponential, algorithm faces large n, and the
algorithm is written in an interpreted or weakly compiled language? What
about software engineering in the large, on big projects? There has been
a lot of discussion about scalability of scripts recently. In the past,
debates have simply ended with the concession that large systems would
have to be rewritten in c++, or a similar language, once the scripting had
served its prototyping duty.
The enterprise question is the easier of the two. Just as the individual
programmer reaps benefits from a division of labor among tools, writing
most of the code in scripts, and writing all bottleneck code in a highly
optimizable language, the group of programmers benefits from the use of
multiple paradigms and multiple languages. In a recent large project, we
used vhdl for fpga's with a lot of gawk to configure the vhdl. We
used python and php to generate dynamic html with svg and javascript for
the interfaces. We used c and c++ for high performance communications
wrappers, which communicated xml to higher level scripts that managed
databases and processes. We saw sysadmin and report-generation in perl,
ruby, and gawk, data scrubbing in perl and gawk, user scripting in bash,
tcl, and gawk, and prototyping in perl and gawk. Only one module was
written in java (because that programmer loved java): it was late, it was
slow, it failed, and it was eventually rewritten in c++. In retrospect,
neither the high performance components nor the lightweight code
components were appropriate for the java language. Does scripting scale
to enterprise software? I would not manage a project that did not include
a lot of scripting, to minimize the amount of "hard" programming, to
increase flexibility and reduce delivery time at all stages, to take
testing to a higher level, and to free development resources for
components where performance is actually critical. I nearly weep when I
think about the text processing that was written in c under my managerial
watch, because the programmer did not know perl. We write five hundred
line scripts in gawk that would be ten thousand line modules in java or
c++. In view of the fact that there are much better scripting tools for
most of what gets programmed in java and c++, perhaps the question is
whether java and c++ scale.
How about algorithmic complexity? Don't scripting languages take too long
to perform nested loops? The answer here is that a cpu-bound tight loop
such as a matrix multiplication is indeed faster in a language like c.
But such bottlenecks are easy to identify and indeed easy to rewrite in
c. True system bottlenecks are things like paging, chasing pointers on
disk, process initialization, garbage collection, fragmentation, cache
mismanagement, and poor data organization. Often, we see that better data
organization was unimplemented because it would have required more code,
code that would have been attempted in an "easier" programming language
like a scripting language, but which was too difficult to attempt in a
"harder" programming language. We saw this in the AI class with heuristic
search and computer vision, where brute force is better in c, but complex
heuristics are better than brute force, and scripting is better for
complex heuristics. When algorithms are exponential, it usually doesn't
matter what language you use because most practical n will incur too great
a cost. Again, the solution is to write heuristics, and scripting is the
top dog in that house. Cpu's are so much faster than disks these days
that a single extra disk read can erase the CPU advantage of using
compiled c instead of interpreted gawk. In any case, java is hardly the
first choice for those who have algorithmic bottlenecks.
The real reason why academics were blindsided by scripting is their lack
of practicality. Academic computing was generally late to adopt Wintel
architectures, late to embrace cgi programming, and late to accept Linux
in the same decade that brought scripting's rise. Academia understandably
holds industry at a distance. Still, there is a purely intellectual
reason why programming language courses are only now warming to
scripting. The historical concerns of programming language theory have
been syntax and semantics. Java's amazing contribution to computer
science is that it raised so many old-fashioned questions that tickled the
talents of existing programming language experts: e.g., how can it be
compiled? But there are new questions that can be asked, too, such as
what a particular language is well-suited to achieve inexpensively,
quickly, or elegantly, especially with the new mix of platforms. The
proliferation of scripting languages represents a new age of innovation in
programming practice.
Linguists recognize something above syntax and semantics, and they call it
"pragmatics". Pragmatics has to do with more abstract social and
cognitive functions of language: situations, speakers and hearers,
discourse, plans and actions, and performance. We are entering an era of
comparative programming language study when the issues are higher-level,
social, and cognitive too.
My old friend, Michael Scott, has a popular textbook called PROGRAMMING
LANGUAGE PRAGMATICS. But it is a fairly traditional tome concerned with
parameter passing, types, and bindings (it's hard to see why it merits
"pragmatics" in its title, even as it goes to second edition with a
chapter on scripting added!). A real programming pragmatics would ask
questions like:
how well does each language mate to the other UNIX tools?
what is the propensity in each language for programmers
at various expertise levels to produce a memory leak?
what is the likelihood in each language that unmodified
code will still function in five years?
what is the demand of a programmer's concentration, what is the
load on her short-term memory of ontology, and what is the support
for visual metaphor in each language?
There have been programming language "shootouts" and "scriptometers" on
the internet that have sought to address some of the questions that are
relevant to the choice of scripting language, but they have been just
first steps. For example, one site reports on the shortest script in each
scripting language that can perform a simple task. But absolute brevity
for trivial tasks, such as "print hello world" is not as illuminating as
typical brevity for real tasks, such as xml parsing.
Pragmatic questions are not the easiest questions for
mathematically-inclined computer scientists to address. They refer by
their nature to people, their habits, their sociology, and the
technological demands of the day. But it is the importance of such
questions that makes programmers choose scripting languages. Ousterhout
declared scripting on the rise, but perhaps so too are programming
language pragmatics.
Acknowledgements
I have to thank Charlie Comstock for contributing many ideas and
references over the past two years that have shaped my views,
especially the commitment to the idea of pragmatics.
About the Author
Prof. Dr. Loui and his students are the usual winners of the department
programming contest and have contributed to current gnu releases of gawk
and malloc. He has lectured on AI for two decades on five continents,
taught AI programming for two decades, and is currently funded on a
project delivering hardware and software on U.S. government contracts.
References
Graham, P., Hackers and Painters, O'Reilly, 2004.
Levinson, S., Pragmatics, Cambridge University Press, 1983.
Ousterhout, John K., Scripting: Higher Level Programming for the
21st Century, IEEE Computer, March 1998.
Shannon, Christine, Another breadth-first approach to CS I using Python,
Proceedings of the 34th SIGCSE technical symposium on
Computer science education, Reno, 2003.
Scott, Michael L., Programming Language Pragmatics, Morgan Kaufman 2000.
Spinellis, Diomidis. Java makes scripting languages irrelevant? IEEE
Software, 22(3):70-71, May/June 2005.
Zelle, J.M., Python as a First Language, 13th Annual Midwest Computer
Conference, 1999.
Java Is Dead, Long Live Java!" The Future of Java By: Bryan W. Taylor
(check out the very last line... hardly a defense of java!)
http://au.sys-con.com/read/169595.htm
(For tutorial material on Awk, see Learning Awk page.)
R. Loui loui@ai.wustl.edu is Associate Professor of Computer Science, at Washington University in St. Louis. He has published in AI Journal, Computational Intelligence, ACM SIGART, AI Magazine, AI and Law, the ACM Computing Surveys Symposium on AI, Cognitive Science, Minds and Machines, Journal of Philosophy.
Whenever Ronald Loui teaches GAWK, he gives the students the choice of learning PERL instead. Ninety percent will choose GAWK after looking at a few simple examples of each language (samples shown below). Those who choose PERL do so because someone told them to learn PERL.
After one laboratory, more than half of the GAWK students are confident with their GAWK skills and can begin designing. Almost no student can become confident in PERL that quickly.
After a week, 90% of those who have attempted GAWK have mastered it, compared to fewer than 50% of PERL students attaining similar facility with the language (it would be unfair to require one to `master' PERL).
By the end of the semester, over 90% who have attempted GAWK have succeeded, and about two-thirds of those who have attempted PERL have succeeded.
To be fair, within a year, half of the GAWK programmers have also studied PERL. Most are doing so in order to read PERL and will not switch to writing PERL. No one who learns PERL migrates to GAWK.
PERL and GAWK appear to have similar programming, development, and debugging cycle times.
Finally, there seems to be a small advantage for GAWK over PERL, after a year, for the programmers willingness to begin a new program. That is, both GAWK and PERL programmers tend to enjoy writing a lot of programs, but GAWK has the slight edge here.
Here are a few short programs that do the same thing in each language. When reading these examples, the question to ask is `how many language features do I need to understand in order to understand the syntax of these examples'.
Some of these are longer than they need to be since they don't exploit some (e.g.) command line trick to wrap the code in for each line do X. And that is the point- for teach-ability, the preferred language is the one you need to know LESS about before you can be useful in it.
hello world
PERL:
print "hello world\n"
GAWK:
BEGIN { print "hello world" }
One plus one
PERL
$x= $x+1;
GAWK
x= x+1
Printing
PERL
print $x, $y, $z;
GAWK
print x,y,z
Printing the first field in a file
PERL
while (<>) {
split(/ /);
print "@_[0]\n"
}
GAWK
{ print $1 }
Printing lines, reversing fields
PERL
while (<>) {
split(/ /);
print "@_[1] @_[0]\n"
}
GAWK
{ print $2, $1 }
Concatenation of variables
PERL
command = "cat $fname1 $fname2 > $fname3"
GAWK
command = "cat " fname1 " " fname2 " > " fname3
Looping
PERL:
for (1..10) { print $_,"\n" }
GAWK:
BEGIN {
for (i=1; i<=10; i++) print i
}
Pairs of numbers
PERL:
for (1..10) { print "$_ ",$_-1 }
print "\n"
GAWK:
BEGIN {
for (i=1; i<=10; i++) printf i " " i-1
print ""
}
List of words into a hash
PERL
foreach $x ( split(/ /,"this is not stored linearly") )
{ print "$x\n" }
GAWK
BEGIN {
split("this is not stored linearly",temp)
for (i in temp) print temp[i]
}
Printing a hash in some key order
PERL
$n = split(/ /,"this is not stored linearly");
for $i (0..$n-1) { print "$i @_[$i]\n" }
print "\n";
for $i (@_) { print ++$j," ",$i,"\n" }
AWK
BEGIN {
n = split("this is not stored linearly",temp)
for (i=1; i<=n; i++) print i, temp[i]
print ""
for (i in temp) print i, temp[i]
}
Printing all lines in a file
PERL
open file,"/etc/passwd";
while (<file>) { print $_ }
GAWK
BEGIN {
while (getline < "/etc/passwd") print
}
Printing a string
PERL
$x = "this " . "that " . "\n";
print $x
GAWK
BEGIN {
x = "this " "that " "\n" ; printf x
}
Building and printing an array
PERL
$assoc{"this"} = 4;
$assoc{"that"} = 4;
$assoc{"the other thing"} = 15;
for $i (keys %assoc) { print "$i $assoc{$i}\n" }
GAWK
BEGIN {
assoc["this"] = 4
assoc["that"] = 4
assoc["the other thing"] = 15
for (i in assoc) print i,assoc[i]
}
Sorting an array
PERL
split(/ /,"this will be sorted once in an array");
foreach $i (sort @_) { print "$i\n" }
GAWK
BEGIN {
split("this will be sorted once in an array",temp," ")
for (i in temp) print temp[i] | "sort"
while ("sort" | getline) print
}
Sorting an array (#2)
GAWK
BEGIN {
split("this will be sorted once in an array",temp," ")
n=asort(temp)
for (i=1;i<=n;i++) print temp[i]
}
Print all lines, vowels changed to stars
PERL
while (<STDIN>) {
s/[aeiou]/*/g;
print $_
}
GAWK
{gsub(/[aeiou]/,"*"); print }
Report from file
PERL
#!/pkg/gnu/bin/perl
# this is a comment
#
open(stream1,"w | ");
while ($line = <stream1>) {
($user, $tty, $login, $junk) = split(/ +/, $line, 4);
print "$user $login ",substr($line,49)
}
GAWK
#!/pkg/gnu/bin/gawk -f
# this is a comment
#
BEGIN {
while ("w" | getline) {
user = $1; tty = $2; login = $3
print user, login, substr($0,49)
}
}
Web Slurping
PERL
open(stream1,"lynx -dump 'cs.wustl.edu/~loui' | ");
while ($line = <stream1>) {
if ($flag && $line =~ /[0-9]/) { print $line }
if ($line =~ /References/) { $flag = 1 }
}
GAWK
BEGIN {
com = "lynx -dump 'cs.wustl.edu/~loui' &> /dev/stdout"
while (com | getline line) {
if (flag && line ~ /[0-9]/) { print line }
if (line ~ /References/) { flag = 1 }
}
}
Download from LAWKER.
Look at the first line of each file for something that looks like thos:
#!/usr/bin/gawk -f
Replace this with the full path to the local version of Gawk.
Developers
Ronald Loui (programmer and designer)
Organization
Washington University in St. Louis
Country
USA
Domain
Text-based game simulation.
Contact
Ronald P. Loui
Email
r.p.loui@gmail.com
Description
Ronald Loui writes: Most people are surprised when I tell them what language we use in our undergraduate AI programming class. That's understandable. We use GAWK. GAWK, Gnu's version of Aho, Weinberger, and Kernighan's old pattern scanning language
This code manages a CGI interface to a process that simulates a soccer game, polling for inputs from two student programs.
A repeated observation in this class is that only the scripting programmers can generate code fast enough to keep up with the demands of the class. Even though students were allowed to choose any language they wanted, and many had to unlearn the java ways of doing things in order to benefit from scripting, there were few who could develop ideas into code effectively and rapidly without scripting.
In the puny language, GAWK, which Aho, Weinberger, and Kernighan thought not much more important than grep or sed,
I find lessons in AI's trends, Airs history, and the foundations of AI.
What I have found not only surprising but also hopeful, is that when I have approached the AI people who still enjoy
programming,
some of them are not the least bit surprised.
Awk
Was written for gawk in 1995 but should run on almost any awk dialect; some css positioning commands will not work in all browsers; try IE6.
Platform
Was written on Redhat Linux with multiple hardware platforms in mind.
Uses
Intended to be run on close server to minimize delays.
Lines
605 lines in main cgi with several small aux control programs.
DevelopmentEffort
Minimal compared to development effort, but potentially will require css for new browsers.
MaintenanceEffort
Number of person-months since, including enhancements
Current
2=Evaluation.
Users
50 students in artificial intelligence project classes had to use some version of this code over seven years
This is an evolutionary algorithm and visualization of a clustering
algorithm that could be turned from O(n^4) to O(nlogn) with a few
judicious uses of constants. Later developments added other
interactive devices, including progress meters and mouse-and-click
behavior.
Contact
Ronald Loui
Email
r.p.loui@gmail.com
Description
The code is an excellent example of the power of Awk as a prototyping
tool: after getting the code running, with the least development time,
a quirk was observed in the code that allowed a reduction from O(n^4)
to O(nlogn).
Two of the n's are lost
(n^2) by noticing that when there is a swap, the delta in the scoring
function falls off by the squared distance from the point of a swap.
So if you just set a constant, such as 10 or 20, or 100, based on the
expected size of your clusters, then you can stop calculating the
scoring function when you get past that constant.
The other n comes
from either fixing the size of the matrix, and occasionally flushing
new candidates in and out, or else by sampling over a subset of the n
when you calculate the score.
The nlogn remains
because there is a sort every now and then.
Awk
Gawk
Platform
Intended for fast servers, 1+ ghz.
Uses
Html.
Lines
158.
Development Effort
One weekend.
Maintenance Effort
None.
Current
2=Evaluation.
Use
2=in-House use.
Users
5
DateDeployed
2004.
Dated
Feb 2009.
References
Streaming Hierarchical Clustering for Concept Mining
Looks, M.; Levine, A.; Covington, G.A.; Loui, R.P.; Lockwood, J.W.; Cho, Y.H.
Aerospace Conference, 2007 IEEE
Volume , Issue , 3-10 March 2007 Page(s):1 - 12
Digital Object Identifier 10.1109/AERO.2007.352792
Ronald Loui (programmer and designer), Anne Jump (adversary)
Organization
National Science Foundation grant at Washington University in St. Louis
Country
USA
Domain
Prototype of a new idea for cognitive modelling (in artificial intelligence/economics/organizational behavior)
Contact
Ronald P. Loui
Email
r.p.loui@gmail.com
Description
Program generates a game board upon which players take turn searching or declaring according to a protocol. It is based on the same game bimatrix made famous by people like von Neumann and Nash, but invents a new approach to negotiation based on process instead of solution.
Awk
Was written for gawk in 1997 but should run on almost any awk dialect
Platform
Was written on Redhat Linux with multiple hardware platforms in mind
Uses
Was intended to be self-contained
Lines
658 lines, of which 39 are comments
DevelopmentEffort
One day, 6-8 hours total
MaintenanceEffort
Two revisions are available, mainly to permit programs to negotiate instead of humans, and to provide a web-based dashboard to monitor the events
CurrentStatus
2=Evaluation
Use
2=in-House use
Users
50 students in artificial intelligence project classes had to use some version of this code over three yeears
DateDeployed
October 1997
Dated
January 2008
References
There is a draft article (unpublished), and several talks,
e.g.
The paper in Harper and Wheeler, Probability and Inference: Essays in Honour of Henry E. Kyburg Jr. (Paperback), Publisher: College Publications (23 April 2007) ISBN-10: 1904987184 ISBN-13: 978-1904987185 also refers to the theory implemented here. Diana Moore's thesis on negotiation and draft article http://citeseer.ist.psu.edu/11983.html contains some precursor ideas.
This was written for the AI course, and for several investigations, including the determination of whether it is a good idea to bat the pitcher in the 8th spot. One hypothesis that emerges from this program that deserves further study is that the most potent offense is one that spreads rather than concentrates the batting threats.
Awk
Gawk around 2002
Platform
Linux around 2002
Uses
None
Lines
409
DevelopmentEffort
Approximately one day
MaintenanceEffort
Further simulators were developed for improved domain modeling and for successive addition of functionality; no other code maintenance was required.
CurrentStatus
1=Prototype
Use
1=Personal use
Users
About 50 students used this program over three years in AI classes, and two undergraduate theses and one Master's thesis on evolutionary computing made use of this simulator.
DateDeployed
October 2002
Dated
January 2009
References
None, but see Tony LaRussa's comments on batting order while managing the St. Louis Cardinals
