The Yorick Programming Language

Yorick is an interpreted language for numerical analysis used by scientists on machines from Linux laptops to Cray supercomputers.

Linux leverages a vast amount of academic software, either easy ports of existing UNIX packages or, increasingly in recent years, software that is ready to run under Linux. One example is Yorick, and this article is an attempt to provide a brief overview of the nature and capabilities of this system.

Yorick is not just another calculator. Readable syntax, array notation and powerful I/O and graphics capabilities make Yorick a favorite tool for scientific numerical analysis. Machine-independent I/O, using the standard NetCDF file formats, simplifies moving applications between hardware architectures. Yorick is an interpreted language developed by David H. Munro at Livermore Labs. Implemented in C, it is freely distributed under a liberal copyright. Yorick runs on a vast range of machines, from 486SX Linux Laptops (in my case) to Cray YMP supercomputers.

Who uses Yorick? The majority of users are physicists, many with access to the most powerful computers in the world. Specific applications include Astrophysics, Astronomy, Neurosciences, Medical Image Processing and Fusion Research.

In this article I will discuss the basics of running Yorick, describe the key array operations, and briefly discuss array operations, programming and graphics. I hope that this quick look is enough to get the more mathematically inclined readers to give Yorick a try.

Basic Operations

When invoked without arguments, Yorick presents a typical command-line interface. Expressions are evaluated immediately, and the result is displayed. Primitive types include integers, floating-point values and strings. All the built-in functions and constants you would expect to be present are present. Variable names are unadorned, with no leading $ character and need not be pre-declared. C-style comments are supported.

One might not expect an interpreted language to be suitable for numerical analysis, and indeed, this would be the case if arrays were not built into the language. Arrays are first-class objects that can be operated on with a single operation. Since the virtual machine understands arrays, it can apply optimized compiled subroutines to array operations, eliminating the speed penalty of the interpreter.

Arrays can be created explicitly:

> a1 = [1.1, 1.2, 1.3, 1.4, 1.5]

Elements can be accessed singly or as a subset, with 1 being the origin. Parentheses indicate the indexing operation, and a single index or a range of indices can be specified.

> a1
> a1(2)
> a1(1:3)
Since array operations are built into the language, functions applied to the array are automatically applied to all elements at once.
> sqrt(a1)
Arrays are not limited in dimension. The rank (number of indices) of an array is not limited to one (a vector) or two (a matrix), but can be as large as desired. Arrays of rank 3 can be used to represent the distribution of a parameter across a volume, and an array of rank 4 could model this over time.

Yorick also provides a simple but effective help system. Executing the help command describes the help system. Executing it with a command name as an argument provides information on that command.

Yorick provides a complete programming language that closely matches C in terms of control flow, expressions and variable usage. For example, the statement:

> for(i=1; i<10; i++) { print,1<<i; }

will print the powers of two just as you would expect. Function declarations, introduced with func, also work as expected:

> func csc(x) {
> return 1/sin(x);
> }
There are differences—variables need not be declared, and arrays are much more powerful than in C. The major difference is in function invocation. Passing arguments to a function in parentheses causes an evaluation and printing of the result; however, passing arguments separated by commas simply executes the function and does not return the result. Since in most cases intermediate results are not required, many scripts contain function calls of the form f,x,y rather than the more familiar f(x,y).

Having a programming language close to C allows easy migration between Yorick for prototyping and C for final implementation. However, as several Yorick users have indicated, moving to C is often unnecessary—the Yorick program proved to be fast enough to get the job done with a minimum of programming effort.

If C extensions are required, a straightforward framework allows extending the Yorick command language with whatever new operations are necessary.