[martyalain]

Thank you for your clever code. This morning I improved a little the JS code in http://lambdaway.free.fr/lambdaspeech/index.php?view=lispolo... and I plan to learn, understand and insert your code in the fft() function.

Yes there is a relationship between the JS version and its translation into lambdatalk. My project is to replace the array based version by a list based version so that I can replace in this page http://lambdaway.free.fr/lambdaspeech/?view=PLR the inefficient unary numeration based implementation of numbers (using standard Church numbers or just lists) by a decimal position numeration. Standard multiplication of words seen as polynoms being O(n^2) I need to go further and implement fast multiplication. So my interest in FFT.

As you could see in http://lambdaway.free.fr/lambdaspeech/meca/JS.js, the lambdatalk's interpreter is a regular expression window running on the code (not an AST) and replacing in situ expressions by their values. A kind of Turing machine. I like the idea of overcoming limits of JS numbers using nothing but words and simple substitutions on words.