dave yarwood
Announcing Alda 2


June 30, 2021



I’m excited to announce the release of Alda 2.0.0! This new version of Alda is a from-the-ground-up rewrite that I’ve been working on since late 2018, and I couldn’t be happier to finally share it with the world.

Despite the shiny new version number, at a language level, Alda 2 is almost exactly the same as the Alda that you’re used to. I have sampled scores from the wild and taken great care to ensure that they are treated the same way by Alda 1 and Alda 2. I even wrote a comparison tool and ran it on a number of Alda scores that I found on GitHub, and it helped me to find and fix some bugs and inconsistencies. As a result, I can say with confidence that your Alda scores should still parse successfully and sound exactly the same as they did with Alda 1.

One notable breaking change is that, because Alda is no longer written in Clojure, Alda no longer supports inline Clojure code, and the attribute syntax is different in a couple of places. I’ve written an Alda 2 migration guide with details about what has changed and what to expect.

Here’s a quick summary:

What’s new?

  • alda is now a native application written in Go. Releases are available for Linux, Mac and Windows.

  • There is also a small background alda-player process, written in Kotlin. Alda starts these processes in the background to play your scores.

  • Because of the move to native and most of the work now being done in the alda client process (instead of in a worker process), Alda 2 is noticeably faster than Alda 1 when it comes to parsing and evaluating scores.

  • You no longer have to start a server in order to use Alda. (No more running alda up!) After installing Alda 2, you can use the alda play command right away and hear your text-based musical creations.

  • The new alda doctor command runs some basic health checks and looks for signs that your system might not be set up for Alda to work properly. If you run into any unexpected problems, alda doctor can help you troubleshoot.

  • The Alda REPL (Read-Eval-Play Loop) has been upgraded to work over a network connection. You can run alda repl in either --client or --server mode, or run alda repl with no flags for the experience that you’re used to from Alda 1.

    This new, network-enabled REPL has promising potential as a foundation for tooling and a platform for collaborative score-writing in the future!

  • You can now specify volume in terms of dynamic markings, which you may be familiar with from sheet music notation. For example, you can use the (mp) attribute for a quiet-ish mezzo piano, or the (ffffff) attribute for the loudest possible dynamic, fortississississimo! Under the hood, these dynamic attributes are equivalent to volume attributes like (vol 46) and (vol 100).

    For more information about the new dynamic attributes, see the docs.

What’s changed?

  • As I mentioned above, inline Clojure code is no longer supported. But if you’re a Clojure programmer and you’re still interested in writing Alda scores programmatically (e.g. to create algorithmic music compositions), have no fear! The alda-clj library can do everything that you were able to do with inline Clojure code in Alda 1, and much more.

  • Whereas Alda 1 had inline Clojure code, Alda 2 has a tiny, built-in Lisp implementation called “alda-lisp”. alda-lisp lacks a lot of Clojure’s syntax and features, but it provides just enough to support Alda’s attributes. Most attributes (e.g. (volume 50)) will work the same way they did before. Two of them have changed slightly:

    • (key-signature! [:a :major]) is now (key-signature! '(a major))

    • (octave :up) and (octave :down) are now (octave 'up) and (octave 'down).

      (Although you’d usually just write > and <, and those still work the same way.)

  • When we implemented the new dynamic attributes, we realized that the default volume level that we’ve been using all this time was a poor choice. The default volume in Alda 1 is (vol 100), which is 100% volume, or fortississississimo. That’s too loud!

    If you ask most classical musicians what a good “default” volume is, they’ll probably tell you that it’s mezzo forte, which is a “middle of the road” volume: neither quiet nor loud. So, in Alda 2, the default volume is (mf), which is equivalent to (vol 54).

    If you have existing scores that assume the starting volume is (vol 100), you may notice that they are quieter than you remember. If you run into this, an easy fix is to simply add a global (vol! 100) at the beginning of the score, preserving the Alda 1 behavior. Or, you can go through the volume attribute changes in your score and adjust them to be relative to the new starting volume, which is (mf) or (vol 54). (It would be a good excuse to try the new dynamic attributes!)

  • The alda parse output in Alda 2 is different from that of Alda 1 in a number of ways. If you happen to have built any tooling or workflows that rely on the Alda 1 alda parse output, you will likely need to make adjustments after upgrading to Alda 2.

Try it out!

If you haven’t installed Alda yet, you can get the latest version by following the instructions on the Alda website.

If you have Alda 1 installed, you can upgrade to Alda 2 by running alda update twice. (The first time will install the last release in the 1.x series, and the second time will install the latest 2.x release.)

After installing Alda 2, give these commands a try:

alda version
alda --help
alda doctor
alda play -c "piano: c8 d e f | d+2/f+/b32 > c+16. < b8 a+ g+ | c+2/e+/f+"

I hope you enjoy the next generation of Alda, and that you’re as excited as I am about the future. Please join our Slack group and let us know what you think of Alda 2. I’d love to hear your thoughts!

Finally, I should mention that if you enjoy my work on Alda, you can support me and sustain the ongoing development of Alda by becoming a sponsor.

Have fun!


Reply to this tweet with any comments, questions, etc.!