#music programming

[michaelgogins]

I Made a Lot of Mistakes...

I have been programming computers in order to algorithmically compose music since 1979 or 1980, that's 46 years. And I am still doing it....

During this period, I made a lot of mistakes that have ended up costing me a great deal of wasted time and effort. I will discuss these here, in the faint hope what I have learned might save somebody else some pain.

However, first I must say, I am not sorry I chose this path. In spite of the wasted time and effort, it has enabled me to make music that I simply would not have been able to make without the software I have written.

So, what were these mistakes, and how could they have been avoided? (I have discussed some of this in my article here on How to Program, and I will be updating that article soon to reflect my experience with artificial intelligence.)

Constantly changing the infrastructure; for example, using first Fortran, then Pascal, then C, then BASIC, then C++, then Python, then Lua, then JavaScript, when using the most capable language (C++) only, and wrapping the C++ library to provide implementations usable in other languages, would have saved me a lot of time. Also, working first on Windows, then on Linux, then on macOS, when moving as soon as possible to macOS would (in hindsight, of course!) have been the wiser move.

Trying to reinvent the wheel; that is, for example, implementing chord space code in Python, JavaScript, and C++, when doing it only in C++ and then wrapping it for Python and JavaScript is the right approach (which I have now adopted).

Too often responding to requests from other users, when I myself am the major user, for example to break up projects into smaller parts, which of course require more time to maintain.

[addiekinstudios]

I am streaming some more development on a non-linear music generator.

Watch this video on Twitch Live

[synaesmedia]

My Best Chord Progression Tool Yet!!!

[michaelgogins]

Pointers for Writing Wrappable C++

This post concerns what I have learned in the course of writing some fairly large bodies of C++ library code that is then wrapped in some dynamic language to be used for doing algorithmic composition. 

I am referring to Csound and especially to CsoundAC.

I have used various facilities to generate bindings from C++ to Python, Java, Lua, Common Lisp, and JavaScript. I have used Swig, Python ctypes, Common Lisp cffi, and now Emscripten’s Embind template library (similar to boost::python which I did experiment with once).

There’s a lot to be said for writing a library in C++, which is a truly powerful language that also compiles to very fast code. However, at the moment I am committed to using both SWIG for Python and Emsripten’s Embind for providing JavaScript interfaces to CsoundAC compiled for WebAssembly.

What I have learned is this:

Go ahead and use sophisticated third party libraries such as Eigen.

But do not trouble to expose those libraries to your dynamic language as well. Instead, provide a C++ interface using mostly elementary data types (numbers and std::strings, usually) to operate with objects from those libraries as much as required, but not more.

You can make exceptions for std::vector and std::map, both SWIG and Embind can parametrize these types and expose them.

Do not use overloaded method or function names, use distinguished names for all functions.

Do not return values in the argument by reference, but rather if you must by address.

[garadinervi]

Paul Doornbusch, How Australia played the world’s first music on a computer, «The Conversation», June 26, 2016. Plus: Paul Doornbusch, Computer Sound Synthesis in 1951: The Music of CSIRAC, «Computer Music Journal», Vol. 28, No. 1, Spring, 2004, pp. 10-25, The MIT Press, Cambridge, MA (pdf here). Plus: Paul Doornbusch, The Music of CSIRAC: Australia's First Computer Music, Common Ground, 2005. Plus: Trevor Pearcey Interview about CSIRAC Music, October 1996 (video)

(image: CSIRAC (CSIR Automatic Computer) before it was put on display at Museum Victoria) (via reaktorplayer)

[magictransistor]

♆

[neckraisin]

Guess who spent hours programming a really cool song and then accidentally deleted it

that’s right, it was me.

the struggle is real

[uniquesquared]

HOW JUST BLAZE VERY NEARLY BECAME A PROGRAMMER

All music can actually be defined on paper without the use of scales and meter.  In fact, a simple tone like an A note on the piano, can be written mathematically and reproduced many times.  The struggle with early electronic music was to find way to control sound without having to break it down to the simplest of axioms.  Programs like Ableton can tackle complex tasks, and when a single song is produced – it requires a lengthy amount of legwork to be done (which is eventually translated into 1’s and 0’s for the computer to read).  Early music making in digital devices had to be done using numbers, rather than symbols. Think of the early pagers and cellular phones that did not have the ability to record and playback music.  These would still play simple 16-bit melodies – but actually creating them required sort of a programming background as well as a musical background.  The fun part happens when the complexity is taken out of the experiment, and it becomes more of a trial and error process rather than a language learning process.  Today, ring tones aren’t the selling point when it comes to buying a phone – but it sure makes it more fun to play with. http://www.factmag.com/2013/08/15/from-hacks-to-beats-how-just-blaze-very-nearly-became-a-programmer/