Naut/examples/anitomy.lua

-- anitomy.lua — a compact, dependency-free anime filename parser in pure Lua.
--
-- A focused reimplementation of the ideas in Anitomy (akatquas/Anitomy): it
-- pulls the anime title, season, and episode (plus release group, resolution,
-- and extension) out of typical scene/fansub filenames. It is intentionally
-- self-contained — no `require`, no io/os — so it can be embedded verbatim in a
-- sandboxed Naut-Torrent script (see anime_sort.lua).
--
-- Usage:
--   local anitomy = require("anitomy")   -- outside the sandbox
--   local p = anitomy.parse("[Group] Some Show - 12 [1080p].mkv")
--   --> p.title="Some Show", p.season=1, p.episode=12, p.extension="mkv"
--
-- It is a heuristic parser; the limitations are documented at the bottom.

local M = {}

----------------------------------------------------------------------
-- keyword / token tables
----------------------------------------------------------------------

local VIDEO_EXT = {
    mkv=true, mp4=true, avi=true, ogm=true, wmv=true, mov=true, flv=true,
    webm=true, m4v=true, mpg=true, mpeg=true, ts=true, m2ts=true, rmvb=true,
}

-- Technical/release tags that are never part of a title. Lowercased; matched
-- token-by-token after bracket removal and delimiter normalization.
local KEYWORD = {}
for _, w in ipairs({
    -- source
    "bd","bdrip","bdremux","bluray","blu-ray","dvd","dvdrip","dvd5","dvd9",
    "web","webrip","web-dl","webdl","hdtv","tvrip","hdrip","remux","hr",
    -- video codec / depth
    "x264","x265","h264","h265","h.264","h.265","hevc","avc","xvid","divx",
    "vp9","10bit","10bits","10-bit","8bit","hi10p","hi10",
    -- audio
    "aac","ac3","eac3","flac","dts","dts-hd","opus","mp3","truehd","aacx2",
    "flacx2","2.0","2.1","5.1","7.1","2ch","6ch",
    -- language / subs
    "dual","dualaudio","multi","multisubs","eng","engsub","jpn","jap","ita",
    "esp","vostfr","vf","raw","sub","subs","subbed","softsub","softsubs",
    "hardsub","hardsubs","dub","dubbed",
    -- release misc
    "uncensored","censored","batch","repack","proper","internal","final",
    "ncop","nced","menu","preview",
}) do KEYWORD[w] = true end

local ORDINAL = { st=true, nd=true, rd=true, th=true }

----------------------------------------------------------------------
-- small helpers
----------------------------------------------------------------------

local function splitext(name)
    local base, ext = name:match("^(.*)%.([%w]+)$")
    if base and #ext <= 4 then return base, ext:lower() end
    return name, nil
end

local function is_resolution(t)
    return t:match("^%d+[pi]$") ~= nil          -- 1080p, 480i
        or t:match("^%d+[xX]%d+$") ~= nil        -- 1280x720
        or t:match("^[24][kK]$") ~= nil          -- 4k, 2k
end

local function is_crc(t)
    -- 8 hex digits with at least one a-f letter (a CRC32, not a date)
    return #t == 8 and t:match("^%x+$") ~= nil and t:match("%a") ~= nil
end

local function is_junk(t)
    local low = t:lower()
    return KEYWORD[low] or is_resolution(t) or is_crc(t)
end

-- tokenize on spaces, keeping "-" as its own token (a common episode delimiter)
local function tokenize(s)
    local out = {}
    for tok in s:gmatch("%S+") do
        out[#out + 1] = tok
    end
    return out
end

----------------------------------------------------------------------
-- season + episode extraction over the cleaned token list
----------------------------------------------------------------------

-- Returns season, episode, episode_end, cut_index (first token that belongs to
-- the season/episode region, so title = tokens[1 .. cut_index-1]).
local function extract(toks)
    local n = #toks
    local season, episode, episode_end, e_idx, s_idx

    -- 1) combined SxxExx in a single token
    for i = 1, n do
        local s, e = toks[i]:match("^[Ss](%d+)[Ee](%d+)$")
        if s then
            local _, e2 = toks[i]:match("[Ee](%d+)[%-~][Ee]?(%d+)$")
            return tonumber(s), tonumber(e), e2 and tonumber(e2) or nil, i
        end
    end

    -- 2) season: "Season 2", "2nd Season", or a standalone "S2"
    for i = 1, n do
        local t = toks[i]
        if t:match("^[Ss]eason$") and toks[i + 1] and toks[i + 1]:match("^%d+$") then
            season, s_idx = tonumber(toks[i + 1]), i
            break
        end
        local num, ord = t:match("^(%d+)(%a%a)$")
        if num and ORDINAL[ord:lower()] and toks[i + 1]
           and toks[i + 1]:match("^[Ss]eason$") then
            season, s_idx = tonumber(num), i
            break
        end
        local sn = t:match("^[Ss](%d+)$")
        if sn and i > 1 then
            season, s_idx = tonumber(sn), i
            break
        end
    end

    -- 3) episode: rightmost strong marker wins
    for i = 1, n do
        local t = toks[i]
        local ep, epend
        local m = t:match("^[Ee](%d+)$") or t:match("^[Ee][Pp]%.?(%d+)$")
               or t:match("^[Ee]pisode(%d+)$") or t:match("^#(%d+)$")
        if m then ep = tonumber(m) end
        if not ep then
            local v = t:match("^(%d+)[vV]%d+$")     -- 01v2
            if v then ep = tonumber(v) end
        end
        if not ep then
            local a, b = t:match("^(%d+)[%-~](%d+)$")  -- 01-12 range token
            if a then ep, epend = tonumber(a), tonumber(b) end
        end
        if not ep and (t:lower() == "ep" or t:lower() == "episode"
                       or t == "#") and toks[i + 1]
           and toks[i + 1]:match("^%d+$") then
            ep = tonumber(toks[i + 1])
        end
        if not ep and t:match("^%d+$") and i > 1 and toks[i - 1] == "-" then
            ep = tonumber(t)                         -- dash-delimited "- 12"
        end
        if ep ~= nil then
            episode, episode_end, e_idx = ep, epend, i
        end
    end

    -- 4) fallback: a bare trailing number (e.g. "Show 12"), ignoring a 4-digit
    --    year and refusing 0 so titles like "Steins;Gate 0" stay intact.
    if episode == nil then
        for i = n, 1, -1 do
            local t = toks[i]
            if t == "-" then
                -- skip trailing dashes
            elseif t:match("^%d+$") then
                local num = tonumber(t)
                local year = (#t == 4 and num >= 1900 and num <= 2099)
                if i > 1 and not year and num > 0 then
                    episode, e_idx = num, i
                end
                break
            else
                break
            end
        end
    end

    local cut
    if e_idx then cut = e_idx end
    if s_idx and (not cut or s_idx < cut) then cut = s_idx end
    if episode ~= nil and season == nil then season = 1 end
    return season, episode, episode_end, cut
end

----------------------------------------------------------------------
-- title assembly
----------------------------------------------------------------------

local function build_title(toks, cut)
    local last = (cut and cut - 1) or #toks
    local parts = {}
    for i = 1, last do
        local t = toks[i]
        if not is_junk(t) then parts[#parts + 1] = t end
    end
    -- drop trailing connective punctuation ("-", "~")
    while #parts > 0 and parts[#parts]:match("^[%-~:_]+$") do
        parts[#parts] = nil
    end
    -- drop a trailing release year (a movie tag), but keep numbers that are part
    -- of the title (e.g. "Mob Psycho 100", "Steins;Gate 0", "86").
    local year
    if #parts > 1 then
        local y = parts[#parts]:match("^([12]%d%d%d)$")
        if y and tonumber(y) >= 1900 and tonumber(y) <= 2099 then
            year = tonumber(y)
            parts[#parts] = nil
            while #parts > 0 and parts[#parts]:match("^[%-~:_]+$") do
                parts[#parts] = nil
            end
        end
    end
    local title = table.concat(parts, " ")
    title = title:gsub("%s+", " "):gsub("^%s+", ""):gsub("%s+$", "")
    return title, year
end

----------------------------------------------------------------------
-- public API
----------------------------------------------------------------------

function M.is_video(ext)
    return ext ~= nil and VIDEO_EXT[ext:lower()] == true
end

function M.parse(filename)
    local base, ext = splitext(filename)

    -- release group: a bracketed run at the very start
    local release = base:match("^%s*%[(.-)%]") or base:match("^%s*%((.-)%)")

    -- resolution: scan bracket/whole-string for an explicit token
    local resolution = base:match("(%d+[pi])%f[%A]")
                     or base:match("(%d+[xX]%d+)")

    -- strip balanced bracket groups; what remains is title + season/episode
    local work = base:gsub("%b[]", " "):gsub("%b()", " "):gsub("%b{}", " ")
    work = work:gsub("[_]", " ")
    if not work:find(" ") and work:find("%.") then
        work = work:gsub("%.", " ")              -- dotted delimiter style
    end
    work = work:gsub("%s+", " "):gsub("^%s+", ""):gsub("%s+$", "")

    local toks = tokenize(work)
    local season, episode, episode_end, cut = extract(toks)
    local title, year = build_title(toks, cut)

    if title == "" then title = release or "Unknown" end

    return {
        title         = title,
        season        = season,        -- nil only when no episode at all
        episode       = episode,       -- nil for movies / unparsed
        episode_end   = episode_end,   -- set for ranges like 01-12
        year          = year,          -- release year, when present
        release_group = release,
        resolution    = resolution,
        extension     = ext,
        file_name     = filename,
    }
end

return M

-- Known limitations (heuristic parser, documented honestly):
--  * A title that ends in a number with no episode marker (e.g.
--    "Mob Psycho 100.mkv") treats the number as the episode, matching Anitomy.
--  * Absolute episode numbering is reported with season 1 (no season marker).
--  * Episode *titles* after the number ("- 05 - The Battle") are discarded.
--  * Multi-token audio tags like "2.0" survive only if spelled as one token.