Auditory Characteristics of Algorithms and Data

Sonnet is an audio-enhanced debugger under development in the Mathematical Sciences Department at IBM Research. The issues in which we are interested include the use of sophisticated yet easy to understand sounds to aid in understanding program execution, how to shift the programmer's focus from the narrow line-oriented view of a program to a global gestalt or holistic view, and finally how to provide a light-weight graphical userinterface that allows run-time interaction rather than post-mortum analysis. An important goal of Sonnet is that it should be easy to predict what sounds will be generated in advance of execution. Should an unexpected sound be produced, the user may then investigate more closely in the hope of finding the anomaly. Sonnet supports a visual programming language that allows arbitrary run-time actions to be attached to running code. Data from the program being debugged can also be accessed and modified. Run-time actions are constructed by dragging primitive and blackbox components onto a canvas, connecting them to each other and to the running code. A special set of primitives are available to control electronic synthesisers so that the programmer can listen to aspects of the code and data. In this document, I describe the visual programming language, its implementation and, as far as is possible without audio equipment, some results of listening to algorithms. God may be subtle but He is not malicious. Nature hides her secrets through her intrinsic grandeur but not through deception. Albert Einstein