"""
P&ID 연결 분석 v2 — 좌표 근접성 기반
======================================

원리: C-10111 분석에서 발견한 패턴
  - 설비(컬럼/탱크) = 긴 수직 LINE(≥200u) 또는 큰 원(r≥4.5)
  - 계기(instrument balloon)는 설비와 Y축 겹침 + 수평 200u 이내
  - LINE 연결이 없어도 좌표 근접성으로 연결 판단

순서:
1. DXF 로드, LINE/CIRCLE/TEXT 추출
2. instrument balloon 식별 (CIRCLE r≥1.5 + func+num)
3. equipment anchor 식별 (긴 수직 LINE + 큰 원)
4. 좌표 근접성 기반 연결 (Y겹침 + 수평거리 ≤ 200)
5. LINE endpoint 연결과 비교
"""

import ezdxf, re, time, math
from shapely.geometry import LineString, Point, MultiLineString, box
from shapely.strtree import STRtree
from shapely.ops import linemerge
from collections import defaultdict

DXF_PATH = "src/Web/uploads/pid/No-10_Plant_PID.dxf"
TAG_RE = re.compile(r'^[A-Z]{1,6}-\d{2,6}(-[A-Z0-9]+)*$', re.I)
FUNC_RE = re.compile(r'^[FPLTASHQWVXZBDCRK][ICTREVYSAQGZ]{1,3}$')
NUM_RE  = re.compile(r'^\d{3,6}[A-Z]?$')
PIPE_RE = re.compile(r'^\w+-\d{3,6}-\d+[A-Z]?-[A-Z][A-Z0-9]', re.I)
LOOP_RE = re.compile(r'\d{3,6}[A-Z]?')

# ── 1-2. DXF 로드 + LINE/CIRCLE/TEXT ──────────────────────────────────
t0 = time.time()
doc = ezdxf.readfile(DXF_PATH)
msp = doc.modelspace()
print(f"1. DXF 로드: {time.time()-t0:.2f}s", flush=True)

# LINE 추출
lines_raw = []
for e in msp:
    t = e.dxftype()
    if t == 'LINE':
        s = (e.dxf.start.x, e.dxf.start.y)
        e2 = (e.dxf.end.x, e.dxf.end.y)
        if s != e2:
            lines_raw.append(LineString([s, e2]))
    elif t == 'LWPOLYLINE':
        pts = [(p[0], p[1]) for p in e.vertices()]
        if len(pts) >= 2:
            lines_raw.append(LineString(pts))

# linemerge
merged = linemerge(MultiLineString(lines_raw))
merged_list = list(merged.geoms) if merged.geom_type == 'MultiLineString' else [merged]

# CIRCLE
circles = [(e.dxf.center.x, e.dxf.center.y, e.dxf.radius) for e in msp if e.dxftype() == 'CIRCLE']

# TEXT
text_entries = []
for e in msp:
    if e.dxftype() == 'TEXT':
        v = e.dxf.text.strip().replace('%%U', '').replace('%%C', 'Φ')
        if v:
            text_entries.append((e.dxf.insert.x, e.dxf.insert.y, v))

print(f"2. 추출: LINE={len(lines_raw)}, CIRCLE={len(circles)}, TEXT={len(text_entries)}", flush=True)

# ── 3. Instrument balloon 식별 ────────────────────────────────────────
t0 = time.time()

# TEXT → CIRCLE 포함 (r≥1.5)
text_in_circle = defaultdict(list)
for x, y, v in text_entries:
    for cx, cy, r in circles:
        if r >= 1.5 and (x-cx)**2 + (y-cy)**2 <= r*r:
            text_in_circle[v].append((cx, cy, r))

# circle center → func+num
circ_data = {}
for v, occ in text_in_circle.items():
    for cx, cy, r in occ:
        k = (round(cx, 1), round(cy, 1))
        if k not in circ_data:
            circ_data[k] = {'r': r, 'x': cx, 'y': cy}
        if FUNC_RE.match(v):
            circ_data[k]['func'] = v
        elif NUM_RE.match(v):
            circ_data[k]['num'] = v
        elif TAG_RE.match(v):
            circ_data[k]['tag'] = v

instruments = []
for k, d in circ_data.items():
    if 'func' in d:
        tag = f"{d['func']}-{d.get('num', '?')}"
        # prefer explicit tag if available
        if 'tag' in d:
            tag = d['tag']
        instruments.append({'x': d['x'], 'y': d['y'], 'tag': tag, 'r': d['r'],
                            'loop': re.search(r'\d{3,6}', tag).group() if re.search(r'\d{3,6}', tag) else None})

# loop number별 계기 그룹
loop_inst = defaultdict(list)
for inst in instruments:
    if inst['loop']:
        loop_inst[inst['loop']].append(inst)

print(f"3-A. Instrument: {len(instruments)}개")
print(f"   Loop 번호 그룹: {len(loop_inst)}개")

# ── 3-B. 방향표지판 검출 ──────────────────────────────────────────
t0 = time.time()

# 방향별 분류: h_seg는 raw LINE만 (O(n²) 회피), v/d_seg는 merged+raw
h_lines = []
v_lines = []
d_lines = []
for l in lines_raw:
    coords = list(l.coords)
    if len(coords) < 2:
        continue
    x1, y1 = coords[0]
    x2, y2 = coords[-1]
    dx = abs(x2 - x1)
    dy = abs(y2 - y1)
    if dy <= 1 and dx >= 3:
        h_lines.append({'xl': min(x1, x2), 'xr': max(x1, x2), 'y': y1})
    elif dx <= 1 and dy >= 2.5:
        v_lines.append({'x': x1, 'y1': min(y1, y2), 'y2': max(y1, y2)})
    elif dx >= 1 and dy >= 1:
        angle = math.degrees(math.atan2(dy, dx))
        if 20 <= angle <= 70 or 110 <= angle <= 160:
            d_lines.append({'x1': x1, 'y1': y1, 'x2': x2, 'y2': y2})
# merged LINE 보강 (v/d만) - merged LINE은 길어서 relaxed threshold
for l in merged_list:
    coords = list(l.coords)
    if len(coords) < 2:
        continue
    x1, y1 = coords[0]
    x2, y2 = coords[-1]
    dx = abs(x2 - x1)
    dy = abs(y2 - y1)
    if dx <= 1 and dy >= 2.5:
        v_lines.append({'x': x1, 'y1': min(y1, y2), 'y2': max(y1, y2)})
    elif dx >= 1 and dy >= 1:
        angle = math.degrees(math.atan2(dy, dx))
        if 20 <= angle <= 70 or 110 <= angle <= 160:
            d_lines.append({'x1': x1, 'y1': y1, 'x2': x2, 'y2': y2})

# 수평선쌍: 같은 xl/xr + y간격 2~8u (xr 기준 버킷으로 O(n²) 회피)
h_by_xr = defaultdict(list)
for h in h_lines:
    h_by_xr[round(h['xr'] / 5) * 5].append(h)
h_pairs = []
seen_pair = set()  # (xl_round, xr_round, y_mid_round) 중복 제거
for bucket in h_by_xr.values():
    for i, h1 in enumerate(bucket):
        for h2 in bucket[i+1:]:
            if abs(h1['y'] - h2['y']) < 2:
                continue
            if abs(h1['xl'] - h2['xl']) <= 3 and abs(h1['xr'] - h2['xr']) <= 3:
                ylo = min(h1['y'], h2['y'])
                yhi = max(h1['y'], h2['y'])
                gap = yhi - ylo
                if 2 <= gap <= 8:
                    xl = (h1['xl']+h2['xl'])/2
                    xr = (h1['xr']+h2['xr'])/2
                    k = (round(xl), round(xr), round((ylo+yhi)/2))
                    if k not in seen_pair:
                        seen_pair.add(k)
                        h_pairs.append({'xl': xl, 'xr': xr, 'ylo': ylo, 'yhi': yhi, 'y_mid': (ylo+yhi)/2})

def _has_left_vert(hp, vlist, tol=8):
    xl, ym = hp['xl'], hp['y_mid']
    return any(abs(v['x'] - xl) <= tol and v['y1'] <= ym <= v['y2'] for v in vlist)

def _count_right_diag(hp, dlist, tol=8):
    xr, ym = hp['xr'], hp['y_mid']
    return sum(1 for d in dlist
               if abs(d['x1'] - xr) <= tol and abs(d['y1'] - ym) <= tol
               or abs(d['x2'] - xr) <= tol and abs(d['y2'] - ym) <= tol)

def _nearest_tag_to(hp, texts, limit=25):
    mx, my = (hp['xl']+hp['xr'])/2, hp['y_mid']
    best, best_d = None, 999
    for tx, ty, tv in texts:
        d = math.hypot(tx - mx, ty - my)
        if d < best_d and d <= limit:
            best_d, best = d, tv
    return best, best_d

print(f"   marker filter start... v={len(v_lines)} d={len(d_lines)} texts={len(text_entries)}")
markers = []
for i, hp in enumerate(h_pairs):
    if not _has_left_vert(hp, v_lines):
        continue
    cnt_diag = _count_right_diag(hp, d_lines)
    if cnt_diag < 2:
        continue
    tag, tag_dist = _nearest_tag_to(hp, text_entries)
    if tag is None:
        continue
    mx = round((hp['xl']+hp['xr'])/2, 1)
    my = round(hp['y_mid'], 1)
    markers.append({'tag': tag, 'mx': mx, 'my': my,
                    'xl': hp['xl'], 'xr': hp['xr'],
                    'ylo': hp['ylo'], 'yhi': hp['yhi'], 'dr': cnt_diag})

et = time.time()
# 위치 중복 제거 (동일태그 + 동일좌표 → 1개)
seen = set()
deduped = []
for m in markers:
    k = (m['tag'], round(m['mx'], 0), round(m['my'], 0))
    if k not in seen:
        seen.add(k)
        deduped.append(m)
markers = deduped

print(f"3-B. 방향표지판(마커): {len(markers)}개 (h_pairs={len(h_pairs)}, elapsed={et-t0:.2f}s)")
print(f"   ROI(y≥5100): {sum(1 for m in markers if m['my'] >= 5100)}개")
tag_groups = defaultdict(list)
for m in markers:
    tag_groups[m['tag']].append(m)
multi = {t: v for t, v in tag_groups.items() if len(v) >= 2}
print(f"   태그 있는 마커: {sum(1 for m in markers if m['tag'])}개")
print(f"   동일태그 그룹(≥2): {len(multi)}개")
for t, v in sorted(multi.items(), key=lambda x: -len(x[1]))[:10]:
    poss = ', '.join(f"({m['mx']:.0f},{m['my']:.0f})" for m in v)
    print(f"     {t}: {len(v)}개 [{poss}]")

# ── 4. Equipment anchor 식별 ──────────────────────────────────────────
t0_4 = time.time()

# (A) 긴 수직 LINE (≥200u, almost vertical)
vert_lines = []
for l in merged_list:
    c = list(l.coords)
    xs = [p[0] for p in c]
    ys = [p[1] for p in c]
    dx = max(xs) - min(xs)
    dy = max(ys) - min(ys)
    if dy >= 200 and dy > dx * 3:  # vertical-ish
        vert_lines.append({
            'x': round(sum(xs)/len(xs), 1),
            'y_min': min(ys), 'y_max': max(ys),
            'length': l.length, 'bounds': l.bounds
        })

# dedup by x
vert_grouped = defaultdict(list)
for v in vert_lines:
    vert_grouped[round(v['x'])].append(v)

vert_anchors = []
for x, vl in vert_grouped.items():
    y_min = min(v['y_min'] for v in vl)
    y_max = max(v['y_max'] for v in vl)
    vert_anchors.append({'x': x, 'y_min': y_min, 'y_max': y_max,
                         'count': len(vl), 'total_len': sum(v['length'] for v in vl)})

# (B) 큰 원 (r≥4.5, empty or with text)
large_circs = [(cx, cy, r) for cx, cy, r in circles if r >= 4.5]
# Find nearest TAG text
eqp_by_circle = []
for cx, cy, r in large_circs:
    best_tag = None
    best_d = 999
    for tx, ty, v in text_entries:
        if TAG_RE.match(v):
            d = math.hypot(tx-cx, ty-cy)
            if d < best_d and d > r and d < 200:
                best_d = d
                best_tag = v
    eqp_by_circle.append({'x': cx, 'y': cy, 'tag': best_tag or f'CIRCLE@{cx:.0f},{cy:.0f}',
                           'r': r, 'kind': 'large_circle'})

print(f"\n4. Equipment anchor:")
print(f"   수직 LINE(≥200u): {len(vert_anchors)}개")
for va in sorted(vert_anchors, key=lambda v: v['x']):
    print(f"     x={va['x']:6.1f}, y=[{va['y_min']:.0f}, {va['y_max']:.0f}], 높이={va['y_max']-va['y_min']:.0f}")
print(f"   큰 원(r≥4.5): {len(eqp_by_circle)}개")
for eq in eqp_by_circle:
    print(f"     ({eq['x']:6.1f}, {eq['y']:6.1f}) r={eq['r']:.1f} tag={eq['tag']}")

# (C) 작은 원 (r<4.5, r≥1.5, exclude instrument balloons)
# instrument balloon circles already identified as those containing TEXT
balloon_circles = set()
for v, occ in text_in_circle.items():
    for cx, cy, r in occ:
        balloon_circles.add((round(cx, 1), round(cy, 1)))
small_circs = [(cx, cy, r) for cx, cy, r in circles
               if 1.5 <= r < 4.5 and (round(cx,1), round(cy,1)) not in balloon_circles]
small_eqp = []
for cx, cy, r in small_circs:
    best_tag = None
    best_d = 999
    for tx, ty, v in text_entries:
        if TAG_RE.match(v):
            d = math.hypot(tx-cx, ty-cy)
            if d < best_d and d > r and d < 100:
                best_d = d
                best_tag = v
    small_eqp.append({'x': cx, 'y': cy, 'tag': best_tag or f'SMALL@{cx:.0f},{cy:.0f}',
                       'r': r, 'kind': 'small_circle'})
eqp_by_small = small_eqp

print(f"   작은 원(1.5≤r<4.5, excl. balloon): {len(eqp_by_small)}개")
for eq in eqp_by_small[:10]:
    print(f"     ({eq['x']:6.1f}, {eq['y']:6.1f}) r={eq['r']:.1f} tag={eq['tag']}")

# Merge small into _what_at equipment list
all_eqp = eqp_by_circle + eqp_by_small

# ── 3-C. 방향표지판 LINE endpoint 추적 (2-hop BFS, 직각굽힘 허용) ──
t0 = time.time()

# Precompute: endpoint spatial index (grid bucketing)
# 30u limit → grid cell size 15 so we only check 3×3 neighbors
GRID = 15
ep_grid = defaultdict(list)
line_ep = {}
for li, l in enumerate(merged_list):
    coords = list(l.coords)
    if not coords:
        continue
    s, e = coords[0], coords[-1]
    for ep, ei in [(s, 0), (e, -1)]:
        gx, gy = int(ep[0] // GRID), int(ep[1] // GRID)
        ep_grid[(gx, gy)].append((li, ei, ep))
    line_ep[li] = {'s': s, 'e': e}

def _grid_neighbors(pt):
    gx, gy = int(pt[0] // GRID), int(pt[1] // GRID)
    return [(gx+dx, gy+dy) for dx in (-1,0,1) for dy in (-1,0,1)]

def _find_ep_idx(pt, limit=30):
    best = None
    for gk in _grid_neighbors(pt):
        for li, ei, ep in ep_grid.get(gk, []):
            d = math.hypot(pt[0]-ep[0], pt[1]-ep[1])
            if d <= limit and (best is None or d < best[0]):
                best = (d, li, ei, ep)
    return best

def _follow_line(li, ei):
    ep = line_ep[li]
    return ep['e'] if ei == 0 else ep['s']

def _connected_at(pt, tol=5):
    result = []
    for gk in _grid_neighbors(pt):
        for li, ei, ep in ep_grid.get(gk, []):
            d = math.hypot(pt[0]-ep[0], pt[1]-ep[1])
            if d <= tol:
                result.append((li, ei, ep))
    return result

def _is_origin(kind, tag, pt, origin_tag, origin_pos, span=40):
    """추적 결과가 출발 마커 자기자신인지 (cycle) 판정."""
    if origin_tag is None or kind != 'marker' or tag != origin_tag:
        return False
    return math.hypot(pt[0] - origin_pos[0], pt[1] - origin_pos[1]) <= span


def _trace_2hop(pt, markers, vert_anchors, all_eqp, limit=30,
                origin_tag=None, origin_pos=None):
    """마커 endpoint → 2-hop 추적. 출발 마커로 되돌아오는 사이클 차단 + 중복 제거."""
    hop0 = _find_ep_idx(pt, limit)
    if not hop0:
        return []
    _, li1, ei1, ep1 = hop0
    op1 = _follow_line(li1, ei1)
    best = {}  # (kind, tag) → (total_len, path)  최단 1개만 유지

    def _offer(kind, tag, total_len, path, at_pt):
        if kind == 'empty':
            return
        if _is_origin(kind, tag, at_pt, origin_tag, origin_pos):
            return  # self-cycle: 출발 마커로 회귀
        k = (kind, tag)
        if k not in best or total_len < best[k][0]:
            best[k] = (total_len, path)

    # hop1: 같은 endpoint를 공유하는 세그먼트(중복 li2 제거)
    seen_li2 = set()
    for li2, ei2, ep2 in _connected_at(op1):
        if li2 == li1 or li2 in seen_li2:
            continue
        seen_li2.add(li2)
        op2 = _follow_line(li2, ei2)
        kind, tag = _what_at_pt(op2, markers, vert_anchors, all_eqp, limit)
        _offer(kind, tag, merged_list[li1].length + merged_list[li2].length,
                f"seg#{li1}→seg#{li2}", op2)

    # hop0 자체의 반대쪽 끝
    kind1, tag1 = _what_at_pt(op1, markers, vert_anchors, all_eqp, limit)
    _offer(kind1, tag1, merged_list[li1].length, f"seg#{li1}", op1)

    return [(k[0], k[1], v[0], v[1]) for k, v in best.items()]

# Precompute marker grid for fast what_at lookup
marker_grid = defaultdict(list)
for mi, m in enumerate(markers):
    gx, gy = int(m['mx'] // GRID), int(m['my'] // GRID)
    marker_grid[(gx, gy)].append(mi)

def _what_at_pt(pt, markers, vert_anchors, all_eqp, limit=30):
    px, py = pt
    for gk in _grid_neighbors(pt):
        for mi in marker_grid.get(gk, []):
            m = markers[mi]
            if math.hypot(px-m['mx'], py-m['my']) <= limit:
                return 'marker', m['tag']
    for va in vert_anchors:
        if va['y_min'] <= py <= va['y_max'] and abs(px-va['x']) <= limit:
            return 'equipment', f"VLINE@{va['x']:.0f}"
    for eq in all_eqp:
        if math.hypot(px-eq['x'], py-eq['y']) <= limit:
            return 'equipment', eq['tag']
    return 'empty', ''

marker_traces = []
trace_log = []
for m in markers:
    mx, my, xl, xr, tag = m['mx'], m['my'], m['xl'], m['xr'], m['tag']
    for side, pt in [('left', (xl, my)), ('right', (xr, my))]:
        hops = _trace_2hop(pt, markers, vert_anchors, all_eqp,
                           origin_tag=tag, origin_pos=(mx, my))
        for kind, to_tag, total_len, path in hops:
            marker_traces.append({
                'from': tag, 'from_pos': f"({mx:.0f},{my:.0f})",
                'side': side, 'to_kind': kind, 'to_tag': to_tag,
                'total_len': total_len, 'path': path
            })
            if tag == 'P-10101':
                trace_log.append(f"  {side} ({pt[0]:.1f},{pt[1]:.1f}) → {path}: {kind}={to_tag} ({total_len:.1f}u)")

et = time.time()
print(f"3-C. 마커 LINE endpoint 추적 (2-hop BFS, elapsed={et-t0:.2f}s):")
print(f"   추적 엣지: {len(marker_traces)}개")
lk = defaultdict(int)
for e in marker_traces:
    lk[e['to_kind']] += 1
for k, c in sorted(lk.items()):
    print(f"     → {k}: {c}개")
if trace_log:
    print(f"   P-10101 상세 추적:")
    for line in trace_log:
        print(line)

# ── 3-D. 태그 매칭 연결 ─────────────────────────────────────────
t0 = time.time()

tag_match_edges = []
for tag, ml in multi.items():
    for i in range(len(ml)):
        for j in range(i+1, len(ml)):
            tag_match_edges.append({
                'tag': tag,
                'from_pos': f"({ml[i]['mx']:.0f},{ml[i]['my']:.0f})",
                'to_pos': f"({ml[j]['mx']:.0f},{ml[j]['my']:.0f})",
                'span': abs(ml[i]['mx'] - ml[j]['mx'])
            })

et = time.time()
print(f"3-D. 태그 매칭 연결: {len(tag_match_edges)}쌍 (elapsed={et-t0:.2f}s)")
for t, ml in sorted(multi.items(), key=lambda x: -len(x[1]))[:10]:
    xs = [m['mx'] for m in ml]
    span = max(xs) - min(xs)
    print(f"     {t}: {len(ml)}개 위치, 스팬={span:.0f}u")

# ── 5. 좌표 근접성 기반 연결 ──────────────────────────────────────────
t0 = time.time()

# 각 instrument → 가장 가까운 equipment anchor 찾기
# 기준: Y범위 겹침 + 수평거리 ≤ 200
HORIZONTAL_LIMIT = 200

inst_eqp_connections = []  # (inst, eqp, dist, method)
unconnected_inst = []

for inst in instruments:
    ix, iy = inst['x'], inst['y']
    best_eqp = None
    best_d = 999
    
    # 수직 LINE anchor
    for va in vert_anchors:
        if va['y_min'] <= iy <= va['y_max']:
            d = abs(ix - va['x'])
            if d < best_d and d <= HORIZONTAL_LIMIT:
                best_d = d
                best_eqp = (f"VLINE@{va['x']:.0f}", va, 'vline')
    
    # 큰 원 anchor
    for eq in eqp_by_circle:
        d = math.hypot(ix - eq['x'], iy - eq['y'])
        if d < best_d and d <= HORIZONTAL_LIMIT:
            best_d = d
            best_eqp = (eq['tag'], eq, 'circle')
    
    if best_eqp:
        inst_eqp_connections.append((inst, best_eqp[0], best_d, best_eqp[2]))
    else:
        unconnected_inst.append(inst)

print(f"\n5. 좌표 근접 연결 (수평≤{HORIZONTAL_LIMIT}):")
print(f"   계기-설비 연결: {len(inst_eqp_connections)}개")
print(f"   미연결 계기: {len(unconnected_inst)}개")

# 연결 분류
by_method = defaultdict(int)
for *_, method in inst_eqp_connections:
    by_method[method] += 1
for m, c in by_method.items():
    print(f"     {m}: {c}개")

# loop별 연결 분포
loop_connected = defaultdict(list)
loop_unconnected = defaultdict(list)
for inst, eqp_tag, d, method in inst_eqp_connections:
    loop_connected[inst['loop']].append((inst, eqp_tag, d))
for inst in unconnected_inst:
    loop_unconnected[inst['loop']].append(inst)

print(f"\n   Loop별 연결/미연결:")
for loop in sorted([k for k in loop_connected if k is not None], key=int)[:20]:
    conn = len(loop_connected[loop])
    unconn = len(loop_unconnected.get(loop, []))
    total = conn + unconn
    if total >= 3:
        eqps = set(e for _, e, _ in loop_connected[loop])
        print(f"     Loop {loop}: {conn}/{total} 연결 → 설비: {', '.join(eqps)}")

# ── 6. LINE endpoint 연결과 비교 ──────────────────────────────────────
t0 = time.time()

# long lines (≥30)
long_lines = [l for l in merged_list if l.length >= 30]

# build anchor index
inst_tree = STRtree([Point(inst['x'], inst['y']) for inst in instruments])
inst_list = instruments

BUFFER = 50

def nearest_inst(pt):
    idxs = inst_tree.query(box(pt.x-BUFFER, pt.y-BUFFER, pt.x+BUFFER, pt.y+BUFFER), predicate='intersects')
    if len(idxs) == 0:
        return None
    best_d = 999
    best = None
    for idx in idxs:
        p = inst_tree.geometries[idx]
        d = p.distance(pt)
        if d < best_d:
            best_d = d
            best = (inst_list[idx], best_d)
    if best and best[1] <= BUFFER:
        return best
    return None

line_connections = []
inst_seen = set()
for l in long_lines:
    ep1 = Point(l.coords[0])
    ep2 = Point(l.coords[-1])
    n1 = nearest_inst(ep1)
    n2 = nearest_inst(ep2)
    if n1 and n2:
        i1, d1 = n1
        i2, d2 = n2
        if i1['tag'] != i2['tag']:
            line_connections.append((i1, i2, l.length, d1, d2))

print(f"\n6. LINE endpoint 연결 (≥30u, buffer={BUFFER}):")
print(f"   계기-계기 연결: {len(line_connections)}개")

# 같은 loop 내 연결 vs 다른 loop 연결
same_loop = 0
diff_loop = 0
for i1, i2, ll, d1, d2 in line_connections:
    if i1['loop'] and i2['loop'] and i1['loop'] == i2['loop']:
        same_loop += 1
    else:
        diff_loop += 1
print(f"   같은 loop: {same_loop}개")
print(f"   다른 loop: {diff_loop}개")

# ── 7. Loop 기반 통합 연결 ────────────────────────────────────────────
# loop → 가장 가까운 equipment anchor
print(f"\n7. Loop 기반 연결 분석:")
for loop in sorted(loop_inst, key=int):
    insts = loop_inst[loop]
    # Y범위
    ys = [i['y'] for i in insts]
    y_min, y_max = min(ys), max(ys)
    xs = [i['x'] for i in insts]
    x_min, x_max = min(xs), max(xs)
    
    # 가장 가까운 equipment 찾기
    cx, cy = sum(xs)/len(xs), sum(ys)/len(ys)
    best_eqp = None
    best_d = 999
    
    for va in vert_anchors:
        if va['y_min'] <= cy <= va['y_max']:
            d = abs(cx - va['x'])
            if d < best_d and d <= HORIZONTAL_LIMIT:
                best_d = d
                best_eqp = f"VLINE@{va['x']:.0f}"
    
    for eq in eqp_by_circle:
        d = math.hypot(cx - eq['x'], cy - eq['y'])
        if d < best_d and d <= HORIZONTAL_LIMIT:
            best_d = d
            best_eqp = eq['tag']
    
    if best_eqp:
        connected = sum(1 for inst in insts if any(inst['tag'] == ii['tag'] for ii, *_ in loop_connected.get(loop, [])))
        print(f"  Loop {loop}: {len(insts)}개 계기 → {best_eqp} (거리={best_d:.0f}, Y=[{y_min:.0f},{y_max:.0f}], X=[{x_min:.0f},{x_max:.0f}])")

# ── 8. 방향성 부여 + 유향 그래프 + JSON 출력 ──────────────────────────────
# ▶ 마커(V자=우측)는 국소 흐름이 좌→우.
#   side='left'  추적대상 = 상류(upstream)  → edge: target → marker
#   side='right' 추적대상 = 하류(downstream)→ edge: marker → target
# 태그매칭: 같은 태그가 여러 x위치 = off-page connector 연속. x오름차순 체인(좌→우).
import json
try:
    import networkx as _nx
    _HAVE_NX = True
except ImportError:
    _HAVE_NX = False

directed_edges = []  # {from, to, type, basis, weight}
_seen_de = set()
def _add_de(frm, to, typ, basis, w):
    if not frm or not to or frm == to:
        return
    k = (frm, to, typ)
    if k in _seen_de:
        return
    _seen_de.add(k)
    directed_edges.append({'from': frm, 'to': to, 'type': typ,
                           'basis': basis, 'weight': round(w, 1)})

# (1) trace 기반 유향 엣지 (사이클 차단된 marker_traces 사용)
for e in marker_traces:
    if e['to_kind'] == 'empty' or not e['to_tag'] or e['to_tag'] == e['from']:
        continue
    if e['side'] == 'left':
        _add_de(e['to_tag'], e['from'], 'trace', 'marker-left(upstream)', e['total_len'])
    else:
        _add_de(e['from'], e['to_tag'], 'trace', 'marker-right(downstream)', e['total_len'])

# (2) 태그매칭 유향 엣지 (x오름차순 체인)
for tag, ml in multi.items():
    ordered = sorted(ml, key=lambda m: m['mx'])
    for a, b in zip(ordered, ordered[1:]):
        _add_de(f"{tag}@{a['mx']:.0f},{a['my']:.0f}",
                f"{tag}@{b['mx']:.0f},{b['my']:.0f}",
                'tagmatch', 'offpage-connector(+x)',
                abs(b['mx'] - a['mx']))

# 유향 그래프 → 약연결 성분
components_out = []
if _HAVE_NX:
    DG = _nx.DiGraph()
    for de in directed_edges:
        DG.add_edge(de['from'], de['to'])
    for comp in sorted(_nx.weakly_connected_components(DG), key=len, reverse=True):
        if len(comp) >= 2:
            components_out.append(sorted(comp))

out = {
    'drawing': DXF_PATH,
    'stats': {
        'instruments': len(instruments),
        'markers': len(markers),
        'marker_tag_groups': len(multi),
        'trace_edges_raw': len(marker_traces),
        'directed_edges': len(directed_edges),
        'tagmatch_directed': sum(1 for d in directed_edges if d['type'] == 'tagmatch'),
        'trace_directed': sum(1 for d in directed_edges if d['type'] == 'trace'),
        'components': len(components_out),
    },
    'markers': [{'tag': m['tag'], 'x': m['mx'], 'y': m['my'],
                 'dir': 'right', 'diag': m['dr']} for m in markers],
    'equipment': [{'tag': eq['tag'], 'kind': eq['kind'],
                   'x': round(eq['x'], 1), 'y': round(eq['y'], 1)}
                  for eq in all_eqp],
    'edges': directed_edges,
    'components': components_out,
}
OUT_JSON = "mcp-server/storage/No-10_connections.json"
import os as _os
_os.makedirs(_os.path.dirname(OUT_JSON), exist_ok=True)
with open(OUT_JSON, 'w', encoding='utf-8') as f:
    json.dump(out, f, ensure_ascii=False, indent=1)
print(f"\n8. 방향성 + JSON 출력:")
print(f"   유향 엣지: {len(directed_edges)}개 "
      f"(태그매칭 {out['stats']['tagmatch_directed']}, trace {out['stats']['trace_directed']})")
print(f"   유향 약연결 성분(≥2): {len(components_out)}개")
print(f"   → {OUT_JSON} ({_os.path.getsize(OUT_JSON)/1024:.0f} KB)")

# ── 요약 ──────────────────────────────────────────────────────────────────
print(f"\n{'='*60}")
print(f"요약")
print(f"{'='*60}")
print(f"Instruments (balloon): {len(instruments)}개")
print(f"방향표지판 (marker): {len(markers)}개 (태그있음={sum(1 for m in markers if m['tag'])}개)")
print(f"동일태그 그룹(≥2): {len(multi)}개, 태그매칭 엣지: {len(tag_match_edges)}쌍")
print(f"Equipment anchors: 수직LINE {len(vert_anchors)}개 + 큰원 {len(eqp_by_circle)}개 + 작은원 {len(eqp_by_small)}개")
print(f"좌표 근접 연결 (수평≤{HORIZONTAL_LIMIT}): {len(inst_eqp_connections)}개 / {len(instruments)}개 ({len(inst_eqp_connections)/max(len(instruments),1)*100:.1f}%)")
print(f"LINE endpoint 연결 (balloon-balloon): {len(line_connections)}쌍 (같은 loop {same_loop}개)")
print(f"Marker trace 엣지: {len(marker_traces)}개")