microtonal-utils

A javascript library for doing precise calculations involving microtonal intervals and notes, inspired by fraction.js.

Interact with this code live at: yacavone.net/microtonal-utils.

The main datatype of this library is Interval, located in lib/interval.js. Internally, intervals are stored as prime factorizations with fractional exponents, which allows for both justly tuned intervals (e.g. 5/4 which can be written as 2^(-2) * 5^1) and intervals in equal temperaments (e.g. 4\12 which can be written as 2^(4/12)) to be represented by the same datatype.

Like fraction.js, you can perform calculations using Intervals without losing any precision. For example, notice that the representations of 5/4 * 4\12 / 1\12 and 5/4 * 3\12 are identical:

> Interval(5,4).mul(Interval(2).pow(4,12)).div(Interval(2).pow(1,12))
{ _fact: { '2': { s: -1n, n: 7n, d: 4n }, '5': { s: 1n, n: 1n, d: 1n } } }
> Interval(5,4).mul(Interval(2).pow(3,12))
{ _fact: { '2': { s: -1n, n: 7n, d: 4n }, '5': { s: 1n, n: 1n, d: 1n } } }

This is because both have the factorization 2^(-7/4) * 5^1. Trying to do this same calculation with javascript's number type, we get imprecision:

> 5/4 * Math.pow(2, 4/12) / Math.pow(2, 1/12)
1.4865088937534015
> 5/4 * Math.pow(2, 3/12)
1.4865088937534012

Another example of where using Interval is more precise, and more convenient, is when computing cents:

> Interval(2).pow(4,12).toCents()
400
> Math.log(Math.pow(2, 4/12)) / Math.log(2) * 1200
400.00000000000006

See the JSDoc comments above the functions in interval.js for more examples.

This repository also contains:

pythagorean.js: functions for constructing and getting properties of Pythagorean intervals, as well as functions for formatting Pythagorean interval and note symbols
edo.js: functions for approximating intervals in equal temperaments, as well as functions for formatting EDO-step and ups-and-downs notation interval and note symbols
fjs.js: functions related to the Functional Just System (FJS) and systems like it, in particular, functions for formatting FJS interval and note symbols
color.js: functions for formatting color notation for intervals, notes, and temperaments
parser.js: a parser for arbitrary expressions involving notes, intervals, and the all note/interval symbols mentioned above (used in xen-calc) as well as inverses of all the formatting functions mentioned above
sets.js: generators for sets of intervals, e.g. all intervals in an odd limit, all intervals less than some Tenney height
approx.js: functions for getting best rational and best EDO approximations of an interval
english.js: an experiment with programmatically assigning English names to arbitrary intervals

Building

To build this library, run:

$ npm install

If you've changed grammar.ne, run the following to update grammar.js:

$ npm run nearley

Using

After building on your local machine, to use this library in an interactive REPL run:

$ npm run repl

You can also visit yacavone.net/microtonal-utils or repl.it/@m_yac/microtonal-utils instead to run the REPL directly in your web browser.

Everything exported in the lib directory is made accessible. Here's an example of an interactive session:

> Interval([-2,0,1]).toCents()
386.3137138648344
> colorTemperament(32805,32768)
'layo'
> updnsSymb(22, edoApprox(22, parsePySymb("M3")))
[ 'vM3' ]
> parseCvt("sqrt(3/2)")
{
  type: 'interval',
  cents: 350.97750043269366,
  intv: { _fact: { '2': [Object], '3': [Object] } },
  ref: { hertz: 440, intvToA4: { _fact: {} } },
  symb: { NFJS: 'n3' },
  english: [ 'Pythagorean neutral third' ]
}

To use this library as part of a website, run:

$ npm run build:all

A single-file ("browserified") version of the entire library can then be found at dist/microtonal-utils.js. Copy that as well as microtonal-utils.js.map into your project and simply include the former as you would any other javascript file. All the functions of the library are included under microtonal_utils (e.g. microtonal_utils.parseCvt). A minified version of this file can also be found at dist/microtonal-utils.min.js.

Testing

This library includes a test suite of property-based tests, located in /test. The term "property-based" means that each test consists of some property (e.g. Interval(2).pow(fr).toCents() == fr.mul(1200)) which is then checked using many (usually 100) randomly generated values (e.g. a randomly generated value for fr would be Fraction(99/38)). Since each run of the test suite checks these properties with a totally new set of random values, the fact that the test suite consistently passes should give high confidence that these properties do hold in general. There are also a few regression tests, which are not property-based.

To run the test suite, use the command npm run test or npm run test:all. The latter also includes tests of the parser, which are often fairly slow. The output of npm run test:all should look like:

$ npm run test:all

  Interval constructors and conversions
    ✓ Interval(n).factors() is the prime factorization of n
    ✓ Interval(a/b).factors() is the prime factorization of a/b
    ✓ Interval(fr).toFrac() == fr
    ✓ Interval(monzo).toMonzo() == monzo
    ✓ Interval(2).pow(fr).toCents() == fr.mul(1200)

  Interval and Fraction operations
    ✓ mul: Interval(fr1).mul(fr2) == fr1.mul(fr2)
    ✓ div: Interval(fr1).div(fr2) == fr1.div(fr2)
    ✓ recip: Interval(fr).recip() == fr.inverse()
    ✓ pow: Interval(fr).pow(n) == fr.pow(n)
    ✓ equals: Interval(fr1).equals(f2) iff fr1.equals(fr2)
    ✓ compare: Interval(fr1).compare(f2) iff fr1.compare(fr2)
    ✓ valueOf: Interval(fr).valueOf() ~= fr.valueOf()

  Other Interval operations
    ✓ pow/mul: i.pow(n) == i.mul(i).mul(i) ... .mul(i)
    ✓ pow/div: i.pow(-n) == i.div(i).div(i) ... .div(i)
    ✓ pow: i.pow(fr).pow(fr.inverse()) == i
    ✓ root/toNthRoot: Interval(fr).root(n).toNthRoot() == {k: fr, n: n}
    ✓ root/valueOf: Interval(fr).root(n).valueOf() ~= Math.pow(fr, 1/n)
    ✓ factorOut: i1 == i2.pow(i1.factorOut(i2)[0]).mul(i1.factorOut(i2)[1])
    ✓ pow/valueOf_log: i.pow(fr).valueOf_log(i) == fr.valueOf()
    ✓ valueOf_log: i.valueOf_log() ~= Math.log2(i)
    ✓ valueOf_log: i1.valueOf_log(i2) ~= i1.valueOf_log() / i2.valueOf_log()
    ✓ toCents: i.toCents() ~= 1200 * i.valueOf_log()
    ✓ isPrimeLimit: i.inPrimeLimit(k) for all k >= i.primeLimit()
    ✓ isPrimeLimit: !i.inPrimeLimit(k) for all k < i.primeLimit()
    ✓ isOddLimit: Interval(fr).inOddLimit(k) for all k >= Interval(fr).oddLimit()
    ✓ isOddLimit: !Interval(fr).inOddLimit(k) for k < Interval(fr).oddLimit()
    ✓ oddLimit: Interval(odd,even).oddLimit() == Interval(even,odd).oddLimit() == odd
    ✓ iso1: i.iso1(fr1).iso1(fr2) == i.iso1(fr1.mul(fr2)) (58ms)

  Best approximations
    ✓ regression: bestRationalApproxsByNo2sHeight({2: 400/1200})
    ✓ regression: bestRationalApproxsByNo2sHeight({2: 400/1200}, {primeLimit: 13})
    ✓ regression: bestRationalApproxsByNo2sHeight({2: 300/1200}, {numIterations: 3})
    ✓ regression: bestRationalApproxsByNo2sHeight({2: 300/1200}, {primeLimit: 13, numIterations: 2})
    ✓ regression: bestRationalApproxsByNo2sHeight({2: 600/1200})
    ✓ regression: bestRationalApproxsByNo2sHeight({2: 600/1200}, {primeLimit: 13, oddLimit: 81})
    ✓ regression: bestRationalApproxsByNo2sHeight(81,64, {primeLimit: 13, oddLimit: 81})
    ✓ regression: bestRationalApproxsByHeight({2: 400/1200}, {primeLimit: 19})
    ✓ regression: bestRationalApproxsByHeight({2: 300/1200}, {primeLimit: 13})
    ✓ regression: bestRationalApproxsByHeight({2: 600/1200}, {primeLimit: 13, oddLimit: 81}) (38ms)
    ✓ regression: bestRationalApproxsByHeight(81,64, {primeLimit: 19})
    ✓ regression: bestRationalApproxsByDenom({2: 350/1200})
    ✓ regression: bestRationalApproxsByDiff({2: 350/1200}, {oddLimit: 9, cutoff: {2: 1/2}})
    ✓ regression: bestEDOApproxsByEDO(5,4)
    ✓ regression: bestEDOApproxsByEDO({2: 100/1200})
    ✓ regression: bestEDOApproxsByDiff(5,4)

  Pythagorean intervals
    ✓ pyDegree(pyInterval(d,o/4)) == d (if d != -1)
    ✓ pyOffset(pyInterval(d,o/4)) == o/4 (if d != -1)
    ✓ pyInterval(d,o/4).recip() == pyInterval(-d,o/4)
    ✓ pyZDegree(pyi1.mul(pyi2)) == pyZDegree(pyi1) + pyZDegree(pyi2)
    ✓ pyInterval(±d,o) == pyInterval(±d,0).mul(pyA1.pow(±o))

  Color notation
    ✓ colorSymb(63,40) == zg6
    ✓ colorSymb(63,40,{verbosity:1}) == zogu 6th
    ✓ colorFromSymb(0, -1, [0,0,-2], 2) == 2048/2025
    ✓ colorFromSymb(0, 0, 5, 1) == 80/81
    ✓ colorTemperament(135,128) == layobi
    ✓ colorTemperament([24,-21,4]) == sasa-quadyo

  55 passing (448ms)


  Intervals and the parser
    ✓ toNthRootString: i == parseCvt(i.toNthRootString()).intv (297ms)
    ✓ fr1.mul(fr2) == parseCvt(`${fr1} * ${fr2}`).ratio (370ms)

  Pythagorean intervals and the parser
    ✓ pyi == parseCvt(pySymb(pyi)).intv (498ms)

  Color notation intervals and the parser
    ✓ fr == parseCvt(colorSymb(fr)).ratio (1203ms)
    ✓ fr == parseCvt(colorSymb(fr, {useExps:1})).ratio (1198ms)
    ✓ fr == parseCvt(colorSymb(fr, {verbosity:1})).ratio (913ms)

  EDO-step intervals and the parser
    ✓ [n,edo] == parseCvt(n + '\\' + edo).edoSteps (776ms)

  Ups-and-downs intervals and the parser
    ✓ [n,edo] == parseCvt(updnsSymb(edo,n)[0]+'\\'+edo).edoSteps (774ms)
    ✓ [n,edo] == parseCvt(updnsSymb(edo,n,{verbosity:1})[0]+'\\'+edo).edoSteps (676ms)


  9 passing (7s)

m-yac/microtonal-utils

microtonal-utils

Building

Using

Testing