This is the home page of J&+ ("with J added", "J plus", or J+) language and YACTS (yet another continuous time simulator) application, based on J+.
J+ is a functional non-imperative simulation language with lazy evaluation, based on J programming language, a dialect of APL. A program in J+ is a collection of (possibly functional) J assignment statements just like a set of formulae on a sheet of paper. The interpreter, implemented as a software library, knows the interdependencies between these formulae and is able to compute any of the defined quantities with minimal effort, keeping track of the values, which were already computed. The driver program (such as YACTS) may ask the interpreter to compute certain quantities and set the values for others (propagating the changes to dependents).
YACTS uses J+ to interpret scripts, defining systems of ordinary differential equations, which it solves. Additionally, it defines a number of J verbs, essential for numerical computing. Please see the online version of YACTS manual page for more information, including a more detailed introduction to J+ language.
The current flagship application of YACTS is the Landau-Lifhitz-Gilbert equation (LLG) solver. This is real application with performance, comparable to LLG solvers, expressed as thousands of lines of C++ code. In a few dozen (of non-comment) lines it contains both simulation (including the definition of vector cross-product) and visualization (able to plot pretty Line Integral Convolution vector fields, superimposed on a color map of local energy). Please see README file in the "examples" sub-directory of J+ distribution for instructions how to run it. The LLG solver was the original motivating application for writing YACTS and J+, but the resulting system is certainly capable of doing much more.
The strong point of YACTS scripts is that they are self-contained and terse (thanks to usage of J). They are comparable in information density to mathematical formulae. Look, for example, at the 7 lines, expressing the stray fields of rectangular prism (verb hp in llg script) and formula (1) in the classical work by A. Aharoni [J. Appl. Phys. 83(6), 1998, p. 3432]. Note that (1) expresses only one component of the field at a single point, while hp evaluates all three, in arbitrary number of locations at the same time. Also, unlike (1), hp is much more formal and can be directly evaluated by computer. It is not hard to imagine that J+ scripts can be publishable in scientific journals directly (e.g. in an Appendix), providing the reader with the value of seeing exactly how the particular result was computed.
Internally J+ library consists of several levels. The first level (classes jarray, jengine) defines direct interface between J and C++, providing the caller with ability to execute J sentences, directly working with J arrays (in a type-safe manner as facilitated by jarray_of_type template), define J verbs and adverbs as C++ functions. The second layer is jplus class, built on top of jengine. It parses and executes J+ scripts with methods to request and set values of variables, defined therein. The third layer is yacts class, built on top of J+, which adds meaning to a few specific variables, defines the library functions for numerical computation and visualization, solves the ODE, manages the trajectory file (class trjfile) and provides options for parallel processing (e.g. on a cluster) of the computed trajectory.
Quck bugfix release 0.4.4 repairs (parallel) processing of saved frames.
Quick followup release 0.4.3, extending the API for making it possible (in future, when J dll supports this) to have multiple instances of jengine. See the NEWS file for details.
The quick followup release 0.4.2 focuses in memory management improvements, for the details see the NEWS file, included in the distribution.
The current release of YACTS is of alpha quality. The initial release is numbered 0.4.1 for the sole purpose of encouraging you to try it ;-), in fact, it is more like 0.0.1 (but it does solve LLG well !). The internal interfaces are more or less well-defined, though. These are worth looking at and commenting. The implementation is ugly (and in some places very ugly). There is a rudimentary testsuite, containing slightly more than 100 tests. Having the testsuite around will help to rewrite the ugly parts in future. The weakest point at the moment is error handling and reporting. User errors mostly cause assertions to fail and YACTS to exit. The error handling must definitely be improved.
Portability: YACTS uses GNU autoconf and should, in theory, be portable. In practice, however, it was only compiled on a handful of my own computers, all running 32 bit Ubuntu Lucid Lynx. Compilation on other platforms is not guaranteed (64 bits will definitely require major porting), but you can try and send me patches.
Dependencies: YACTS depends on a number of external libraries to do its job, on Ubuntu Linux the required build dependencies can be installed with the following command:
apt-get install g++ libc6-dev make \ autoconf automake libltdl-dev libtool \ libsundials-serial-dev libfftw3-dev libreadline-dev \ txt2man doxygen graphviz man2htmlSome of these (like readline or txt2man) can be missing, which will automatically turn off some of the features (not recommended). On other platforms try the equivalents.
License: YACTS is released under the terms of GNU General Public License, which is compatible to the licenses of all the external code it uses.
version 0.4.4 (15 July 2012), tar.gz,
2964696 bytes, md5: 41463f40a1423f62f86be430fda13ea8.
version 0.4.3 (29 June 2012), tar.gz, 2964919 bytes, md5: 5e10bebdfa2c037df98a9edd5bf319e1.
version 0.4.2 (27 June 2012), tar.gz, 2938950 bytes, md5: 54d5bf8fb7419398f362f359c13603c3.
version 0.4.1 (25 June 2012), tar.gz, 2917162 bytes, md5: 226fd41851842b9db0544816164fa37a.
© 2012 Konstantin L. Metlov <email@example.com>