#!/bin/sh
# qndl-artist — 아티스트 폴더/메타데이터 통일 헬퍼
export LC_ALL="${LC_ALL:-C.UTF-8}"

MUSIC="${XDG_MUSIC_DIR:-$HOME/Music}"
ALIASES="${QNDL_ALIASES:-${XDG_DOTFILES_DIR:-$HOME/.dotfiles}/global/Music/artist-aliases.tsv}"
PLAYLIST="${QNDL_MPD_PLAYLIST:-$HOME/.config/mpd/playlists/entire.m3u}"

# 이름 → 표준명. 맵 정확일치(연쇄 추적) → 기존 폴더 대소문자무시 매칭 → 원본.
cmd_normalize() {
  _name="$1"
  if [ -f "$ALIASES" ]; then
    _cur="$_name"; _hops=0
    while [ "$_hops" -lt 20 ]; do
      _next="$(awk -F'\t' -v n="$_cur" '/^#/ || NF < 2 { next } $1 == n { print $2; exit }' "$ALIASES")"
      [ -z "$_next" ] && break
      [ "$_next" = "$_cur" ] && break
      _cur="$_next"
      _hops=$((_hops + 1))
    done
    if [ "$_cur" != "$_name" ]; then
      printf '%s\n' "$_cur"
      return 0
    fi
  fi
  if [ -d "$MUSIC" ]; then
    _m="$(find "$MUSIC" -mindepth 1 -maxdepth 1 -type d -printf '%f\n' 2>/dev/null |
      awk -v n="$_name" 'tolower($0) == tolower(n) { print; exit }')"
    [ -n "$_m" ] && { printf '%s\n' "$_m"; return 0; }
  fi
  printf '%s\n' "$_name"
}

# 후크용: 파일의 현재 아티스트 폴더명을 표준명으로 해석해 apply.
cmd_apply_download() {
  _fp="$1"
  case "$_fp" in
  "$MUSIC"/*) : ;;
  *) return 0 ;;   # Music 밖이면 무시
  esac
  _rel="${_fp#"$MUSIC"/}"
  _seg="${_rel%%/*}"
  _canon="$(cmd_normalize "$_seg")"
  cmd_apply "$_fp" "$_canon"
}

# mp3 1개를 표준 폴더로 이동(필요시) + album_artist 설정.
cmd_apply() {
  _fp="$1"; _canon="$2"
  [ -f "$_fp" ] || { printf 'apply: no such file: %s\n' "$_fp" >&2; return 1; }
  case "$_fp" in
  "$MUSIC"/*) : ;;
  *) printf 'apply: not under %s: %s\n' "$MUSIC" "$_fp" >&2; return 1 ;;
  esac
  case "$_fp" in
  *.mp3) : ;;
  *) return 0 ;;   # mp3 아니면 무시
  esac
  _rel="${_fp#"$MUSIC"/}"
  _artist_seg="${_rel%%/*}"
  _subpath="${_rel#*/}"          # album/.../title.mp3
  if [ "$_artist_seg" != "$_canon" ]; then
    _destdir="$MUSIC/$_canon/$(dirname "$_subpath")"
    _dest="$_destdir/$(basename "$_fp")"
    if [ -e "$_dest" ]; then
      printf 'apply: skip (exists): %s\n' "$_dest" >&2
      return 0
    fi
    mkdir -p "$_destdir"
    mv "$_fp" "$_dest" || return 1
    # 빈 원본 앨범/아티스트 폴더 정리 (MUSIC 루트는 지우지 않음)
    rmdir "$MUSIC/$_artist_seg/$(dirname "$_subpath")" 2>/dev/null || true
    rmdir "$MUSIC/$_artist_seg" 2>/dev/null || true
    _fp="$_dest"
  fi
  # 태깅은 best-effort: 실패해도 이미 이동은 성공했으므로 전체 실패로 취급하지 않음.
  _tmp="$(dirname "$_fp")/.qndl-tag-$$.mp3"
  if ffmpeg -v error -y -i "$_fp" -map 0 -c copy -metadata album_artist="$_canon" "$_tmp" 2>/dev/null &&
     mv "$_tmp" "$_fp"; then
    :
  else
    rm -f "$_tmp"
    printf 'apply: tag failed (file kept): %s\n' "$_fp" >&2
  fi
  return 0
}

# MUSIC의 모든 아티스트 폴더를 <name>\t<mp3수>로 출력.
_artist_counts() {
  find "$MUSIC" -mindepth 1 -maxdepth 1 -type d -printf '%f\n' 2>/dev/null |
    while IFS= read -r _d; do
      _n="$(find "$MUSIC/$_d" -type f -name '*.mp3' 2>/dev/null | wc -l)"
      printf '%s\t%s\n' "$_d" "$_n"
    done
}

# stdin: <name>\t<count>. stdout: 실제 중복 그룹만 <canonical>\t<m1>\t<m2>...
_group_awk() {
  awk -F'\t' '
    function norm(x,   y){ y=tolower(x); gsub(/[^[:alnum:]가-힣]/,"",y); return y }
    function find(a){ while(parent[a]!=a){ parent[a]=parent[parent[a]]; a=parent[a] } return a }
    function union(a,b,  ra,rb){ ra=find(a); rb=find(b); if(ra!=rb) parent[rb]=ra }
    function addtok(idx,chunk,  k){ k=norm(chunk); if(k=="") return; if(k in owner) union(owner[k],idx); else owner[k]=idx }
    function caserank(s,  u,l){ u=(s ~ /[A-Z]/); l=(s ~ /[a-z]/); return (u&&l)?2:1 }
    {
      name[NR]=$1; cnt[NR]=$2+0; parent[NR]=NR
      s=$1; gsub(/（/,"(",s); gsub(/）/,")",s); rest=s
      while (match(rest,/\([^)]*\)/)) {
        addtok(NR, substr(rest,RSTART+1,RLENGTH-2))
        rest=substr(rest,1,RSTART-1) " " substr(rest,RSTART+RLENGTH)
      }
      addtok(NR, rest)
    }
    END{
      for(i=1;i<=NR;i++){ r=find(i); members[r]=members[r] i " " }
      for(r in members){
        n=split(members[r], m, " ")
        real=0; for(j=1;j<=n;j++) if(m[j]!="") real++
        if(real<2) continue
        haveEng=0
        for(j=1;j<=n;j++){ if(m[j]=="") continue; if(name[m[j]] !~ /[가-힣]/) haveEng=1 }
        best=""; brank=-1; bcnt=-1
        for(j=1;j<=n;j++){
          if(m[j]=="") continue
          nm=name[m[j]]
          if(haveEng && nm ~ /[가-힣]/) continue
          cr=caserank(nm); c=cnt[m[j]]
          if(cr>brank || (cr==brank && c>bcnt) || (cr==brank && c==bcnt && (best=="" || nm<best))){
            best=nm; brank=cr; bcnt=c
          }
        }
        line=best
        for(j=1;j<=n;j++){ if(m[j]=="") continue; line=line "\t" name[m[j]] }
        print line
      }
    }
  '
}

# stdin(그룹 라인) → 각 그룹의 비표준 멤버 이동 파일 수 계산해 dry-run 한 줄 출력.
_merge_preview() {
  while IFS="$(printf '\t')" read -r _canon _rest; do
    [ -z "$_canon" ] && continue
    _others=""; _files=0
    _oldifs="$IFS"; IFS="$(printf '\t')"
    for _mem in $_rest; do
      [ "$_mem" = "$_canon" ] && continue
      _others="${_others:+$_others, }$_mem"
      _c="$(find "$MUSIC/$_mem" -type f -name '*.mp3' 2>/dev/null | wc -l)"
      _files=$((_files + _c))
    done
    IFS="$_oldifs"
    [ -z "$_others" ] && continue
    printf '%s → %s (move %s files)\n' "$_others" "$_canon" "$_files"
  done
}

# 맵에 변형→표준 append (이미 있으면 생략).
_map_add() {
  _v="$1"; _c="$2"
  [ -f "$ALIASES" ] || { mkdir -p "$(dirname "$ALIASES")"; : > "$ALIASES"; }
  awk -F'\t' -v v="$_v" '$1==v{f=1} END{exit f?0:1}' "$ALIASES" && return 0
  # 파일이 개행으로 끝나지 않으면 마지막 줄과 이어붙는 손상을 방지
  if [ -s "$ALIASES" ] && [ -n "$(tail -c1 "$ALIASES")" ]; then
    printf '\n' >> "$ALIASES"
  fi
  printf '%s\t%s\n' "$_v" "$_c" >> "$ALIASES"
}

cmd_merge_apply() {
  _groups="$1"
  _failtmp="$(mktemp)" || return 1
  printf '%s\n' "$_groups" | while IFS="$(printf '\t')" read -r _canon _rest; do
    [ -z "$_canon" ] && continue
    _oldifs="$IFS"; IFS="$(printf '\t')"
    for _mem in $_rest; do
      [ "$_mem" = "$_canon" ] && continue
      IFS="$_oldifs"
      find "$MUSIC/$_mem" -type f -name '*.mp3' 2>/dev/null | while IFS= read -r _f; do
        cmd_apply "$_f" "$_canon" || printf 'x\n' >> "$_failtmp"
      done
      # mp3 외 잔여 파일(cover.jpg 등)도 정경 폴더로 이동해 variant 폴더가 완전히 비워지게 함.
      # (cmd_apply를 거치지 않는 단순 mv: 재태깅 대상이 아님)
      find "$MUSIC/$_mem" -type f ! -name '*.mp3' 2>/dev/null | while IFS= read -r _f; do
        _relsub="${_f#"$MUSIC/$_mem"/}"
        _destf="$MUSIC/$_canon/$_relsub"
        if [ -e "$_destf" ]; then
          printf 'merge: skip residue (exists): %s\n' "$_destf" >&2
          continue
        fi
        mkdir -p "$(dirname "$_destf")"
        mv "$_f" "$_destf" 2>/dev/null || true
      done
      find "$MUSIC/$_mem" -type d -empty -delete 2>/dev/null || true
      _map_add "$_mem" "$_canon"
      IFS="$(printf '\t')"
    done
    IFS="$_oldifs"
  done
  _fails="$(wc -l < "$_failtmp" | tr -d ' ')"
  rm -f "$_failtmp"

  if command -v mpc >/dev/null 2>&1; then
    mpc update --wait >/dev/null 2>&1 &&
      find "$MUSIC" -type f -name '*.mp3' | sed "s|$MUSIC/||" | sort \
        > "$PLAYLIST" 2>/dev/null || true
  fi
  if [ "${_fails:-0}" -gt 0 ]; then
    printf 'merge --apply 완료 (%s개 실패/스킵).\n' "$_fails"
  else
    printf 'merge --apply 완료.\n'
  fi
}

cmd_merge() {
  _apply=0
  [ "${1:-}" = "--apply" ] && _apply=1
  _groups="$(_artist_counts | _group_awk)"
  if [ -z "$_groups" ]; then
    printf 'No case/paren duplicate groups found.\n'
    return 0
  fi
  if [ "$_apply" -eq 0 ]; then
    printf '%s\n' "$_groups" | _merge_preview
    printf '\n(dry-run) 실제 병합하려면: qndl-artist merge --apply\n'
    return 0
  fi
  cmd_merge_apply "$_groups"
}

_sub="${1:-}"
[ $# -gt 0 ] && shift
case "$_sub" in
normalize) cmd_normalize "$@" ;;
apply) cmd_apply "$@" ;;
apply-download) cmd_apply_download "$@" ;;
merge) cmd_merge "$@" ;;
*) printf 'usage: qndl-artist {normalize|apply|apply-download|merge} ...\n' >&2; exit 2 ;;
esac