Swarm: applications

Latest revision as of 05:57, 16 August 2011

Wiki main page
Swarm main page
Software main page
Stable release
Development snapshot
License
Platforms
Prior releases
Documentation main page
Doc set & reference guide
User guide
FAQ
User support email list
Applications & contrib code
Applications
User-contributed code
Release Management
Checklist

We make a wide variety of Swarm models available. These illustrate the many different kinds of research questions addressed using agent-based models, and provide example ideas and software for other modelers.

[edit] Demo page

For newcomers, here is a peek at a couple of classic Swarm applications.

[edit] How to choose and run example applications

(For a discussion of Java vs. Objective-C, see the FAQ: Swarm_FAQ)

[edit] Objective-C applications

Listed below are sites where you can download Objective-C Swarm models. To run them:

Download and un-zip the archive. The archives are usually in TAR GZIP format (e.g., HeatBugs.tar.gz), which you can unzip (in a Linux terminal window or, if using Windows, in Cygwin) by typing "tar xzvf HeatBugs.tar.gz".
Change directories into the newly unzipped directory of source code (in a Linux terminal window or, if using Windows, in Cygwin).
If your Swarm installation is set up correctly, you can then compile the application by simply typing "make". (If you get a message something like "Makefile: ... /Makefile.appl: No such file or directory..." then you do not have the environment variable SWARMHOME set correctly. Keep reading...)
If compilation is successful, you will have a new executable file (e.g., "heatbugs.exe" in Windows; "heatbugs" in Linux). Then you can start the model by typing "./heatbugs.exe" (or, in Linux, "./heatbugs").
A common problem is not having your system's SWARMHOME environment variable correctly pointed to the location of your Swarm installation. SWARMHOME should typically be /usr/local/ or /usr/ if you've installed it from a binary package on a GNU/Linux system. For Windows, see Windows_binary_notes

[edit] Java applications

We provide step-by-step instructions for compiling and running the Java program Jheatbugs-3.0 on Windows XP using Swarm 2.2. Our demo covers interactive (visual) mode as well as batch mode (multiple runs with output to files).

Other Java Swarm models are listed below.

[edit] Official applications archive

Our official archive of Swarm applications is at http://ftp.swarm.org/pub/swarm/apps/. Even though these models are generally old, many of them still work in Swarm 2.2. There are far more Objective-C than Java applications in the archive.

The archive includes Java versions of the Swarm classics heatbugs and mousetrap. The versions that worked with Swarm-2.1.1 and Swarm-2.2 are currently available in the testing subdirectory of the ftp archive. Look for jheatbugs-2001-03-28.tar.gz and jmousetrap-2001-09-13.tar.gz.

We strongly encourage people to contribute new applications! The site includes a form for describing contributions.

[edit] Paul Johnson's package of Objective-C applications

Paul Johnson has generously assembled a new package of applications for Swarm-2.2. This is a revised version, correcting a couple of glitches in saving of screen snapshots. It is available from Paul's web site: http://www.ku.edu/~pauljohn/Swarm/swarmapps-objc-2.2-2.tar.gz and also at the Swarm download site: http://ftp.swarm.org/pub/swarm/apps/objc/sdg/swarmapps-objc-2.2-2.tar.gz. Please keep testing and let Paul know how it works.

The package contains classic example Swarm applications, including:

The Swarm tutorial (discussed in the Swarm User Guide)
Heatbugs
Mousetrap
ArborgamesII (a simulation of forests used in publications by M. Savage et al.)
SchellingII (Schelling's neighborhood segregation model)
Conway (the Game of Life)
sss (Swarm Sugar Scape).

We are soliciting volunteers who want to maintain some 'orphaned' applications from swarmapps-2.1.1. These apps that need love and attention are market, template, and hello-world. If you are interested in taking on one of those applications, or offering up other applications, contact PJ.

[edit] Additional links

The world-famous Artificial Stock Market has its own homepage at http://artstkmkt.sourceforge.net (Objective-C)
Paul Johnson keeps many examples of Swarm-related code, his own models, and other people's models, at: http://www.ku.edu/~pauljohn/Swarm/
Keep your eye on the Agent-based modeling community pages for projects using Swarm.
See the software resources at the ABM Resources area of the wiki for a series of template models in Java and Objective-C Swarm, and for guidance on using Swarm with integrated development environments.

@@ Line 8: / Line 8: @@
 ==How to choose and run example applications==
+(For a discussion of Java vs. Objective-C, see the FAQ: [[Swarm_FAQ]])
-===Objective-C or Java?===
-You need to decide whether you will write your Swarm programs using Objective-C or Java. Aside from obvious points (such as the fact that you might already be more familiar with one of the languages), here are the pros and cons of the two languages:
-*In the world of software development, at least one hundred times as much programming is done in Java as in Objective-C &mdash; see http://www.cs.berkeley.edu/~flab/languages.html. So you'll find more people and books that can help you with Java than with Objective-C. You'll also find more third-party libraries so, for example, if your model is three-dimensional, you might choose to abandon Swarm's visual interface (which uses TCL) in favor of Java's Java3D technology.
-*Java was created over ten years later than Objective-C, and presumably takes advantage of the advances in language design that occurred after Objective-C was created.
-*Probably more Swarm programs have been written in Objective-C than in Java, so there are more Swarm-specific programming models in Objective-C than in Java. On the other hand, given the size of the universe of all individual-based models that anyone might like to program in Swarm, unless you know of a relevant existing model right now, you're not likely to find anything more useful than the standard Swarm demos available in both languages.
-*Although expertise in Swarm is rare, and expertise in Objective-C is rare, the latter tends to go with the former. So for the Swarm side of programming, you may find it easier to obtain help in Objective-C, while for the pure model building (programming the behavior of your agents), Java will, as mentioned above, provide more helpful resources.
-*If one language is faster at run time, it's likely to be Objective-C. However, the difference is likely to be slight. You should not accept claims about speed without seeing a relevant benchmark. Try googling ''benchmark java objective-c''.
-*Java's treatment of floating-point arithmetic is not a full implementation of the IEEE 754 standard, in that you cannot check whether floating-point errors have occurred or change the rounding direction. In Objective-C this is not a limitation so long as your platform supplies C libraries to access the floating point unit (e.g. fpgetsticky() and fpgsetround()). This is a key advantage of Swarm over other pure-Java libraries. For more on this issue, there's a [http://dx.doi.org/10.1016/j.envsoft.2004.10.011 general paper] on how floating-point arithmetic issues are relevant to agent-based models, and [http://jasss.soc.surrey.ac.uk/8/1/5.html another] showing how these issues affected two specific published models. You may also wish to [http://www.macaulay.ac.uk/fearlus/floating-point/javabugs.html vote for] some requests for enhancements to Java logged with the Sun website (in particular bug IDs [http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=5095286 5095286] and [http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=5095600 5095600]).
-*Since the Java interface to the Swarm engine is generated automatically, identical functionality is available in both languages. Thus functionality would never be a reason for choosing one language over the other.
 ===Objective-C applications===
@@ Line 31: / Line 15: @@
 *Download and un-zip the archive. The archives are usually in TAR GZIP format (e.g., HeatBugs.tar.gz), which you can unzip (in a Linux terminal window or, if using Windows, in Cygwin) by typing "tar xzvf HeatBugs.tar.gz".
-*Change directories into the newly unzipped directory of source code.
+*Change directories into the newly unzipped directory of source code (in a Linux terminal window or, if using Windows, in Cygwin).
-*If your Swarm installation is set up correctly, you can then compile the application by simply typing "make".
+*If your Swarm installation is set up correctly, you can then compile the application by simply typing "make". (If you get a message something like "Makefile: ... /Makefile.appl: No such file or directory..." then you do not have the environment variable SWARMHOME set correctly. Keep reading...)
 *If compilation is successful, you will have a new executable file (e.g., "heatbugs.exe" in Windows; "heatbugs" in Linux). Then you can start the model by typing "./heatbugs.exe" (or, in Linux, "./heatbugs").
 *A common problem is not having your system's SWARMHOME environment variable correctly pointed to the location of your Swarm installation. SWARMHOME should typically be /usr/local/ or /usr/ if you've installed it from a binary package on a GNU/Linux system. For Windows, see [[Windows_binary_notes]]
@@ Line 68: / Line 52: @@
 *Keep your eye on the [[ABM_Resources|Agent-based modeling community pages]] for projects using Swarm.
 *See the [[ABM_Resources#Software_Resources|software resources at the ABM Resources area of the wiki]] for a series of template models in Java and Objective-C Swarm, and for guidance on using Swarm with integrated development environments.
+----