Learning about Icon
A1. What is Icon?
A2. What is Icon good for?
A3. What are Icon's distinguishing characteristics?
A4. What is the Icon program library?
A5. Where can I learn more about Icon?
A6. Where are some simple examples?
A7. How about comprehensive documentation?
Implementations
B1. What platforms support Icon?
B2. How do I get started with Icon?
B3. Is there a Unicode version of Icon?
B4. What happened to the compiler?
Administration
C1. What is the Icon Project?
C2. How often is the on-line material updated?
C3. Where did Icon come from?
C4. Where is Icon going?
Programming
D1. Why doesn't read() work with every?
D2. Why doesn't string invocation such as "foo"() work?
D3. How can I call a C function?
D4. Can I open a bidirectional pipe?
Icon is a very high level general-purpose programming language with extensive features for processing strings (text) and data structures. Icon is an imperative, procedural language with a syntax that is reminiscent of C and Pascal, but with semantics at a much higher level.
Icon has a novel expression-evaluation mechanism that integrates goal-directed evaluation and backtracking with conventional control structures. It has a string scanning facility for pattern matching that avoids the tedious details usually associated with analyzing strings. Icon's built-in data structures include sets and tables with associative lookup, lists that can be used as vectors or stacks and queues, and records.
Icon is a strongly, though not statically, typed language. It provides transparent automatic type conversion: For example, if an integer is used in an operation that requires a string, the integer is automatically converted to a string.
Most implementations of Icon have high-level graphics facilities with an easily programmed window interface.
Icon manages storage automatically. Objects are created as needed during program execution and space is reclaimed by garbage collection as needed. The sizes of strings and data structures are limited only by the amount of available memory.
As a general-purpose programming language with a large computational repertoire, Icon can be used for most programming tasks. It's especially strong at building software tools, for processing text, and for experimental and research applications.
Icon is designed to make programming easy; it emphasizes the value of programmer's time and the importance of getting programs to work quickly. Consequently, Icon is used both for short, one-shot tasks and for very complex applications.
The library is a collection of programs and procedures written in Icon. User contributions form a significant portion of the library.
Library procedures effectively augment the built-in functions available to an Icon program. A wide variety of procedures currently exists, and most graphically-based programs are built around library procedures. The core and graphics core modules are the most carefully vetted.
The programs in the library range from simple demonstrations to handy tools to complex graphical applications.
The library is a resource for both new and experienced programmers. In addition to their basic utility, its programs and procedures serve as examples of how things can be written in Icon.
The library is indexed at www.cs.arizona.edu/icon/library/ipl.htm.
There is lots of material at the Icon website, www.cs.arizona.edu/icon.
Here are some good places to start:
For a more thorough introduction to the base language, chapter 2 of the Icon graphics book (a free download) is especially good.
For a whirlwind tour of the graphics facilities, see:
For some simple text-based programs, see any of those introductory documents in the preceding question. For some simple graphics programs, see www.cs.arizona.edu/icon/gb/progs/progs.htm.
Many more examples, typically larger, are found in the Icon program library; see the indexes of Basic Programs and Graphics Programs.
Two books define the Icon language. The core language is covered in The Icon Programming Language (third edition), by Griswold and Griswold. Graphics facilities are described in Graphics Programming in Icon by Griswold, Jeffery, and Townsend. These books contain both tutorial and reference material.
Icon's internals are detailed in The Implementation of the Icon Programming Language by Griswold and Griswold. Although considerable changes have occurred since Version 6, described in the book, the basic structure of Icon remains the same. Two technical reports, IPD112 and IPD239, describe subsequent changes.
Printed copies of the Language and Graphics books are available from Jeffery Books (http://unicon.org/books/). All three books can be downloaded at no charge from the Icon books page, www.cs.arizona.edu/icon/books.htm.
A 2010 revision of the book Icon Programming for Humanists, by Alan Corré, is also available for purchase or download from Jeffery Books.
The Icon Programming Language Handbook, by Thomas W. Christopher, is available on the web at www.tools-of-computing.com/tc/CS/iconprog.pdf.
An on-line index to the Icon program library is found at www.cs.arizona.edu/icon/library/ipl.htm.
There is a large amount of additional information at the Icon web site, www.cs.arizona.edu/icon, including complete sets of back issues of the Icon Newsletter and Icon Analyst.
The current implementation of Icon is a Unix implementation. This includes Linux, BSD, Solaris, Macintosh, and the Cygwin environment under Windows. Version 9.5 of Icon has been tested on all these platforms.
Older versions are available for Windows and some other systems. An alternative Java-based implementation for Unix, Jcon, is also available.
None of these environments includes a window-based development environment. While Icon programs can open windows and use graphics, programming is done using Unix editors and other tools from a command shell.
The current version of Icon for Unix can be downloaded from www.cs.arizona.edu/icon/current. Source and binary packages are available, each with the complete Icon program library.
For older implementations, start at www.cs.arizona.edu/icon/implver.htm. Jcon is at www.cs.arizona.edu/icon/jcon.
No. Icon is defined in terms of 8-bit characters, and changing this presents several design challenges that would likely break existing programs. Modifying the C implementation is probably infeasible anyway, although a Unicode version of Jcon might be possible.
For a while, Unix distributions included both an interpreter and a compiler. The compiler was an interesting research project but it has not been maintained and is no longer supported. The interpreter is much easier to use and is generally quite fast enough, even for production applications.
The Icon Project name identifies the group that distributes and supports the Icon programming language. A non-commercial organization, the project is supported by the Department of Computer Science at the University of Arizona.
Gregg Townsend, now retired from Arizona, maintains the Icon implementation and the Icon website. Carl Sturtivant of the University of Minnesota assists with the Cygwin port.
The Icon implementation is now in maintenance mode, with new releases produced only when necessary. This typically happens every few years when changes in host platforms cause the source to stop building.
Icon is the latest in a series of high-level programming languages designed to facilitate programming tasks involving strings and structures. The original language, SNOBOL, was developed at Bell Telephone Laboratories in the early 1960s. SNOBOL evolved into SNOBOL4, which is still in use. Subsequent languages were developed at the University of Arizona with support from the National Science Foundation. Although it has similar objectives and many similar capabilities, Icon bears little superficial resemblance to SNOBOL4.
Icon implementations were developed by faculty, staff, and students at the University of Arizona, with significant contributions from volunteers around the world. An Icon history by Ralph and Madge Griswold appears in the preprints of the second History of Programming Languages Conference (HOPL-II), ACM SIGPLAN Notices, March 1993 (Vol 28, No 3).
The name Icon is not an acronym, nor does it stand for anything in particular, although the word iconoclastic was mentioned when the name was chosen. The name predates the now common use of icon to refer to small images used in graphical user interfaces. This sometimes misleads people into thinking that Icon is designed to create or manipulate icons, but there's no good solution to that problem.
We continue to use Icon on a daily basis, but no significant changes are planned. We expect to support the Unix version for the foreseeable future, and to distribute ports to other systems as supplied by volunteers.
The Unicon project is developing an object-oriented language based on Icon. For more information, see unicon.sourceforge.net. An earlier object-oriented extension to Icon, Idol, can be found in the Icon program library.
read() work with every?
every s := read() do {...}
doesn't loop because read() produces a single value and
then fails if resumed.
Other "consumer" procedures such as get() and pop()
work the same way.
Use a while loop with these procedures, and save
every for use with generators such as !x
or key(T).
"foo"() work?
String invocation works if the procedure is present;
the catch is that the linker removes unreferenced procedures.
To ensure a procedure's presence, reference it in the
main() procedure.
A simple reference suffices, as in
refs := [foo, bar, baz];
it's not necessary to actually call it.
(Why does the linker remove unreferenced procedures? Because that greatly reduces the memory requirements of programs that use the library. There was a time when this mattered.)
You can't call an arbitrary C function, but you can load and call one
that is written to Icon's specifications.
A tutorial appears in
Icon Analyst 36.
Some examples can be found in the
cfuncs and
packs/loadfuncs directories of the Icon program library.
The Jcon implementation allows Icon programs to call Java code that is written to Jcon specifications.
No, this is not possible. Although the concept is simple — write a line to a program via a pipe, then read that program's output — it probably wouldn't work. Most I/O libraries don't write anything to a pipe until they've filled a buffer, and the most likely consequence would be a deadlock, with each program waiting for the other to send more data.