review.lua (27d1c32) - memoire

review.lua

--[[
    stuff to help with the review mechanics
]]

--[[
    determines the delay until next review (in seconds)
    when `Prompt` `prompt` is reviewed at timestamp `revtime`
    with review success `revsucc`
]]
local function schedule(prompt, revtime, revsucc, conf)
    if revsucc and revsucc > 0 then
        local ad = revtime - (prompt.time or revtime)
        local adm = ad + math.max(40000 - 0.5 * ad, 0)
        -- SM2, as explained by Bjornstad
        -- https://controlaltbackspace.org/memory/spaced-repetition-from-the-ground-up/
        if revsucc < 0.8 then
            return (conf:get("HardRatio") or 1.2) * revsucc * 2 * adm
        else
            return prompt.ease * (1 + 2 * ((conf:get("EasyRatio") or 1.3) - 1) * (revsucc - 1)) * adm
        end
    else
        return 15
    end
end

--[[
    manipulates `Prompt` `prompt` in-place as it should be
    when reviewed at timestamp `revtime` with review success `revsucc`
]]
local function update(prompt, revtime, revsucc, conf)
    if revsucc and revsucc > 0 then
        prompt.succ = (prompt.succ or 0) + 1
    else
        prompt.fail = (prompt.fail or 0) + 1
    end
    -- allowed random variation, as a fraction of the "optimal" duration
    local ARV = conf:get("AllowedRandomVariation") or 0.1
    prompt.delay = schedule(prompt, revtime, revsucc, conf) * ((1 - ARV) + 2 * ARV * math.random())
    prompt.ease = prompt.ease + ({ -0.2, -0.15, 0, 0.15 })[2 * revsucc + 1]
    prompt.time = revtime
end

--[[
    checks (returns boolean) if `Prompt` `prompt` should be reviewed now
]]
local function should(prompt)
    local ct = os.time()
    return prompt.delay >= 0 and ct - (prompt.time or 0) >= (prompt.delay or 0) and (ct - (prompt.time or 0)) / prompt.delay
end

--[[
    shuffles a table in-place,
    intentionally biased to keep the list
    in a roughly-similar order to how it was originally
]]
local function biasedshuffle(t)
    for i = 1, (#t-1) do
        if math.random() < 0.5 then
            local s = t[i]
            t[i] = t[i + 1]
            t[i + 1] = s
        end
    end
end

local function alwaystrue(x)
    return true
end

--[[
    returns a table of pairs of integers to index the deck
    to get prompts to be reviewed now;
    prompts are shuffled in order
    if a filter function is provided, filters the list accordingly,
    calling the filter function on each `Prompt`
]]
local function toreviewlist(deck, filter)
    local ctrt = {}
    local ff = filter or alwaystrue
    for i, cg in ipairs(deck) do
        for j, pr in ipairs(cg) do
            if should(pr) and ff(pr) then
                table.insert(ctrt, { i, j })
            end
        end
    end
    table.sort(ctrt, function(a, b)
        return should(deck[a[1]][a[2]]) > should(deck[b[1]][b[2]])
    end)
    biasedshuffle(ctrt)
    return ctrt
end

return { schedule = schedule, update = update, should = should, list = toreviewlist }

memoire

Control and bias review order