Whenever I learn a new song, I found that the easiest way for me to write down the chords, is to just open up a simple text editor on my computer and write it using a super simple syntax.

TL;DR:

I built txt2musicxml, a cli tool that turns this:

Cmaj7 | A7 | Dm9 | G7b9 |

into this:

The “rules” of the syntax are:

  • Roots of chords are capital letters (A, C G)
  • Accidentals are # b
  • All suffixes (or chord types) are be lowercase ascii letters, symbols and numbers (maj7, -7b5, 7b9,13)
  • Bass notes use a slash /

Full example:

Cmaj7 | A7 | Dm9 | G7b9 |

Although quite simple and fast to write, it is ugly and not comfortable to read while playing it back. So I built a CLI tool that turns this into something more meaningful: MusicXML.

MusicXML is a pretty good standard, mainly because it is widely accepted. I don’t know of any music software that doesn’t support it. It’s not an easy standard, but the end user doesn’t need to know the details. The important thing is that with my new tool I can turn my simple chord syntax into MusicXML, and then import it into my favorite music notation software to render it.

Antlr

Well, I had no idea how to do that. Luckily, I found Antlr. It is an open source software to create lexers and parsers (and more) in your favorite programming language. You basically define your new language’s grammar in a single file with a simple syntax, very similar to regex. The hard part was learning the theory behind it. What is a language, what is a lexer, what is a parser, etc. So here are some of the important concepts I learned:

Tokens

A token is the “atom” of a language. The smallest significant part of it. I already defined it above! For example a chord root is a token (D), an accidental is a token (#), a suffix is a token (maj7). But also a barline is a token | and a newline is a token \n.

Lexer

A lexer is the part of the program that turns raw text (an input stream) into tokens. So basically it turns:

Cmaj7 | A7 | Dm9 | G7b9 |

into:

['C', 'maj7', '|', 'A', '7', '|', 'D', 'm9', '|', 'G', '7b9', '|']

Parser

A parser gives structural meaning to tokens. It turns the stream of tokens into something called a parse-tree. In our examples, it would be something like this:

We can start seeing patterns and meaning in our language :)

Visitor (AST)

A parse-tree still contains all the original text, but a lot of it is redundant. We need to simplify our tree to a model that makes more sense. This is called an AST (Abstract Syntax Tree). In our case I will call it a “Sheet”. Antlr has a concept of a Visitor, which walks over all the nodes of the parse tree, and does something with them. In my case, I turned the parse-tree into python classes that have meaningful relationships. A Sheet contains lines, lines contain bars, bars contain chords, a chord has a root, and possibly a suffix and a bass, etc. In our example, it would be something like this (simplified):

Sheet(lines=[
    Line(bars=[
        Bar(chords=[Chord(root='C', suffix='maj7',)]),
        Bar(chords=[Chord(root='A', suffix='7',)]),
        Bar(chords=[Chord(root='D', suffix='m9',)]),
        Bar(chords=[Chord(root='G', suffix='7b9',)]),
    ])
])

XML Generator

Now that we are in python-land, it’s quite easy (yet very tedious), to create a MusicXML generator, which is basically a class that accepts a sheet, walks down the nodes and generates the correct xml according to the MusicXML specifications.

The result is this:

<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE score-partwise PUBLIC "-//Recordare//DTD MusicXML 4 Partwise//EN" "http://www.musicxml.org/dtds/partwise.dtd">
<score-partwise version="4"><part-list><score-part id="P1"><part-name>Chords</part-name></score-part></part-list><part id="P1"><measure number="1"><attributes><divisions>1</divisions><key><fifths>0</fifths><mode>major</mode></key><time><beats>4</beats><beat-type>4</beat-type></time><clef><sign>G</sign><line>2</line></clef></attributes><harmony><root><root-step>C</root-step></root><kind use-symbols="yes">major-seventh</kind></harmony><note><pitch><step>B</step><octave>4</octave></pitch><duration>1</duration><type>quarter</type><stem>none</stem><notehead>slash</notehead></note><note><pitch><step>B</step><octave>4</octave></pitch><duration>1</duration><type>quarter</type><stem>none</stem><notehead>slash</notehead></note><harmony><root><root-step>A</root-step></root><kind use-symbols="yes">dominant</kind></harmony><note><pitch><step>B</step><octave>4</octave></pitch><duration>1</duration><type>quarter</type><stem>none</stem><notehead>slash</notehead></note><note><pitch><step>B</step><octave>4</octave></pitch><duration>1</duration><type>quarter</type><stem>none</stem><notehead>slash</notehead></note></measure><measure number="2"><attributes><divisions>1</divisions></attributes><harmony><root><root-step>D</root-step></root><kind use-symbols="yes">minor-ninth</kind></harmony><note><pitch><step>B</step><octave>4</octave></pitch><duration>1</duration><type>quarter</type><stem>none</stem><notehead>slash</notehead></note><note><pitch><step>B</step><octave>4</octave></pitch><duration>1</duration><type>quarter</type><stem>none</stem><notehead>slash</notehead></note><harmony><root><root-step>G</root-step></root><kind use-symbols="yes">major</kind></harmony><note><pitch><step>B</step><octave>4</octave></pitch><duration>1</duration><type>quarter</type><stem>none</stem><notehead>slash</notehead></note><note><pitch><step>B</step><octave>4</octave></pitch><duration>1</duration><type>quarter</type><stem>none</stem><notehead>slash</notehead></note></measure></part></score-partwise>

Yikes!

What to do with this XML?

The CLI I wrote accepts chord text in stdin, and spits XML in stdout, it’s fully pipe-able and redirectable. So you can do whatever you want in one line. Save it as a file and you can open it with your favorite music notation software! MuseScore is an open source software, and it’s pretty feature rich. They have a CLI tool that let’s you import/export files. Here are some examples:

Pipe a string of chords into the cli and get the xml:

echo -n 'Cmaj7 A7 | Dm9 G7b913 |' | txt2musicxml

or redirect input/output from/to a file:

txt2musicxml < path/to/Thriller.crd > path/to/Thriller.musicxml

or convert it in 1 line to pdf using Musescore:

TMPSUFFIX=.musicxml; mscore -o path/to/output.pdf =(txt2musicxml < path/to/input.crd)