absolute shapers for slope library

[trentgill]

'absolute' shapers from readme-development

[trentgill]

-- the absolute shapers are interesting though as a set of divisions
-- thinking primarily in terms of absolutely dividing up the output
-- space.
-- user provides:
--  zero point to base off (could be a 'key')
--  fold point (1V becomes 1 octave, but could be eg 1v2 for buchla)
--  either | divisions (divides fold-size into n equal steps)
--         | list (divides fold-size unequally according a list)
--
-- the natural use is to provide scales for quantizing. by using the
-- list approach, one can send arbitrary pitches, so this could be
-- tonal scales directly, rather than semitones of 'divisions'.
--
-- list can also take fractions to natively handle just intonation,
-- readily accepting a scale as a list of fractions.
--
-- a weighting mechanism could be added to give a wider window to
-- preferred values.
--
-- setting divisions to a large number, then manipulating fold will
-- essentially become a bit crusher with continuously variable bits
--

local shaper = { zero = 0
               , fold = 1  --volts
               , divs = 12 --semitones
               , list = {} --empty means nil
               }

function S_absolute( samp, divs, fold )
    div_sz  = fold / divs
    bracket = math.floor(samp / fold)
    shelf   = samp % fold
    shelf1  = shelf / fold
    div     = math.floor(0.5 + (shelf1 * divs))
    return ( div * div_sz
             + (bracket * fold)
           )
end

function S_absolute( samp, divs, fold )
    local bracket = math.floor(samp / fold)
    local shelf   = (samp % fold) / fold
    local plateau = 0
    if type(divs) == 'table' then
        if divs[1] == 'ji' then
            --TODO needs log shaping i believe
            plateau = divs[2][math.floor(1.5+(shelf * #divs[2]))] - 1
        else
            plateau = divs[2][math.floor(1.5+(shelf * #divs))] / divs[1]
        end
    else
        plateau = fold / divs * math.floor(0.5 + (shelf * divs))
    end

    --local plateau = type(divs) == 'table'
    --    and divs[math.floor(0.5 + (shelf * #divs))]
    --    or fold / divs * math.floor(0.5 + (shelf * divs))
    return plateau + bracket * fold
end

lydian = {12,{0,2,4,6,7,9,11}}
pentatonic = {12,{0,2,4,7,9}}
just_lydian = {'ji',{ 1/1, 9/8, 4/3, 11/8, 3/2, 10/7, 15/8}}
S_absolute( 0.1
          , lydian
          , 1
          )

[trentgill]

float S_absolute( float samp
                , int   divs // 12 // must be >0
                , float fold // 1.0
                //, float zero // assume 0.0 for now
                ){
    // TODO optimize more elements into k-rate
    // k-rate
    float div_sz = fold / (float)divs; // abs size of a div
    // a-rate
    int bracket  = (int)floorf(samp / fold); // confirm this rounds to -ve
    float shelf  = samp % fold; // 0-fold range
    float shelf1 = shelf / fold; // 0-1 range within fold
    int div      = (int)(0.5 + (shelf1 * (float)divs)); // select the div
    return ( (float)div * div_sz
             + ((float)bracket * fold) // add #folds back
           );
}

[whimsical-sam]

https://github.com/whimsical-sam/crow/blob/shapers/lua/shaper.lua

Check out this implementation for user-facing customizable shapers in lua

basic idea: four different domains

• input signal --> convert to phase space based off a foldover value (e.g. 1V, or 1.2V etc.)
• phase space (# of folds + 0-1 phase within a single cycle/octave etc. )
• lookup table (lookup table divides 0-1 into a series of windows; window size can be affected by different weightings, boundaries may be related to the values within the lookup table, e.g. accurately spacing notes in a scale rather than equally dividing up 0-1)
• output space (converting look up table values to output values, e.g. ratios to volts, semitones to volts, or any other custom mapping; interpolation between lookup table values is also availabile; also includes a function mode for using lookup-tables made up of functions instead of numbers)

here is a script which shows setting a few of them up - including running markov models with one shaper!

The library is mostly tested but it got too big for testing it all as a single user script upload, so I am working on finally getting a build environment set up :)