DB_JOB/작업파일/공공기관3/_gnb_extract.py
hehihoho3 b0acd1e705 백업: 셀병합 정상화(D~J)+KDHC 옵션2/행높이17, 오늘(06-21) 제출 17곳 검수·압축
- 공공기관2/3 작업본 + 오늘 제출 17곳 D~J 카테고리 셀병합 정상화
- 한국지역난방공사 옵션2(고아셀 F98 수정)+전행 높이17
- 제출_프리랜서2_2026-06-21.zip 생성(17개 xlsx, 2,468행)

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-21 13:21:26 +09:00

142 lines
7.1 KiB
Python

# -*- coding: utf-8 -*-
"""공공기관3 GNB 권위트리 추출 (메가메뉴 펼침 + 자동 컨테이너 선택) → _gnb/{num}.json
사용: python _gnb_extract.py [num ...]"""
import sys, io, os, json, re, warnings
sys.stdout = io.TextIOWrapper(sys.stdout.buffer, encoding='utf-8')
warnings.filterwarnings('ignore')
from playwright.sync_api import sync_playwright
GNB_DIR = r"D:\01.프로젝트\DB수집\작업파일\공공기관3\_gnb"
os.makedirs(GNB_DIR, exist_ok=True)
# num: (name, base, entry_url, container_sel(None=auto), toggle_sel(None))
SITES = {
1: ("한국원자력환경공단", "https://www.korad.or.kr", "https://www.korad.or.kr/", None, 'a:has-text("전체 항목 보기")'),
2: ("한국의료분쟁조정중재원", "https://www.k-medi.or.kr", "https://www.k-medi.or.kr/", ".gnb", None),
3: ("한국의약품안전관리원", "https://www.drugsafe.or.kr", "https://www.drugsafe.or.kr/", ".allmenu", None),
4: ("한국인터넷진흥원", "https://www.kisa.or.kr", "https://www.kisa.or.kr/", "#gnb", None),
5: ("한국임업진흥원", "https://www.kofpi.or.kr", "https://www.kofpi.or.kr/", None, None),
6: ("한국자산관리공사", "https://www.kamco.or.kr", "https://www.kamco.or.kr/main.do", "header nav", None),
7: ("한국잡월드", "https://www.koreajobworld.or.kr", "https://www.koreajobworld.or.kr/", None, None),
8: ("한국장기조직기증원", "https://www.koda1458.kr", "https://www.koda1458.kr/", ".header-menu", 'button:has-text("전체메뉴")'),
9: ("한국장애인개발원", "https://www.koddi.or.kr", "https://www.koddi.or.kr/", ".gnb", None),
10: ("한국장애인고용공단", "https://www.kead.or.kr", "https://www.kead.or.kr/", None, None),
11: ("한국장학재단", "https://www.kosaf.go.kr", "https://www.kosaf.go.kr/ko/main.do", None, None),
12: ("한국재정정보원", "https://www.fis.kr", "https://www.fis.kr/", None, None),
13: ("한국저작권보호원", "https://www.kcopa.or.kr", "https://www.kcopa.or.kr/", None, None),
}
JUNK = re.compile(r'^(바로가기|닫기|열기|검색|전체\s*메뉴|전체메뉴|메뉴\s*보기|메뉴|메인메뉴|서브메뉴|본문\s*바로|주\s*메뉴|TOP|위로|맨위|로그인|로그아웃|회원가입|마이페이지|언어선택|언어|english|eng|日本|中文|sitemap|사이트맵|즐겨찾기|새창|팝업|로고|이동|home|홈)\s*$', re.I)
# 자동 컨테이너 후보
CAND_SELS = ["#gnb","ul#gnb","#gnbmenu",".gnb",".gnb_area",".gnb_wrap","nav .gnb",
".allmenu",".all_menu","#allMenu",".total_menu",".totalmenu",".menu_all",
"#header nav ul","header nav","#nav",".nav","ul.depth1","ul.dep1",".gnb ul",
".lnb_all",".navi",".navbar","#snb",".menu_wrap",".header_menu","#headerWrap nav"]
EXTRACT_JS = r"""
(args) => {
const {sels, candSels} = args;
function topUls(el){
return [...el.querySelectorAll('ul')].filter(ul=>{
let p=ul.parentElement;
while(p&&p!==el){ if(p.tagName==='UL') return false; p=p.parentElement; }
return true; });
}
function labelA(li){
for(const a of li.querySelectorAll('a, button')){
let p=a.parentElement, inside=false;
while(p&&p!==li){ if(p.tagName==='UL'){inside=true;break;} p=p.parentElement; }
if(inside) continue;
const t=(a.textContent||'').trim();
if(t) return a;
}
return null;
}
function extractFrom(cont){
const rows=[];
function walk(ul, depth){
[...ul.children].filter(li=>li.tagName==='LI').forEach(li=>{
const a=labelA(li);
if(a){
const label=(a.textContent||'').trim().replace(/\s+/g,' ');
const href=a.tagName==='A'?(a.href||''):'';
if(label) rows.push({depth, label, href});
}
topUls(li).forEach(sub=>walk(sub, depth+1));
});
}
(cont.tagName==='UL'?[cont]:topUls(cont)).forEach(u=>walk(u,1));
return rows;
}
// 컨테이너 선택: 지정 sels 우선, 없으면 후보 중 행수 최다
let chosen=null, chosenSel=null, best=-1;
const trySels = sels && sels.length ? sels : candSels;
for(const sel of trySels){
document.querySelectorAll(sel).forEach(el=>{
const r=extractFrom(el);
const score=r.length*10 + Math.max(...r.map(x=>x.depth),0);
if(r.length>=3 && score>best){ best=score; chosen=el; chosenSel=sel; }
});
if(chosen && sels && sels.length) break;
}
if(!chosen) return {err:'no container', rows:[], sel:null};
return {sel: chosenSel, rows: extractFrom(chosen)};
}
"""
def clean_rows(rows):
out=[]
for r in rows:
lab=r['label']
if JUNK.search(lab): continue
if len(lab)>60: continue
out.append({'depth':min(r['depth'],5),'label':lab,'href':r['href'] or ''})
if out:
md=min(r['depth'] for r in out)
if md>1:
for r in out: r['depth']-=(md-1)
return out
def run(nums):
with sync_playwright() as p:
br=p.chromium.launch(headless=True)
ctx=br.new_context(ignore_https_errors=True, viewport={'width':1920,'height':1080},
user_agent='Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 Chrome/120 Safari/537.36')
for num in nums:
name, base, url, sel, toggle = SITES[num]
pg=ctx.new_page()
try:
pg.goto(url, wait_until='domcontentloaded', timeout=40000)
pg.wait_for_timeout(2500)
# 전체메뉴 토글 시도 (있으면 펼침)
for tsel in ([toggle] if toggle else
['button:has-text("전체메뉴")','a:has-text("전체메뉴")','.btn-allmenu','.allmenu_btn','#allmenuBtn','.gnb_all','.btn_total','button[class*=all]','button[class*=menu]']):
try:
el=pg.query_selector(tsel)
if el:
el.click(timeout=2000); pg.wait_for_timeout(1200); break
except Exception:
pass
# 메가메뉴 hover로 펼침(top li들에 마우스)
try:
tops=pg.query_selector_all((sel or 'nav')+' > ul > li, '+(sel or 'nav')+' li')
for el in tops[:12]:
try: el.hover(timeout=500); pg.wait_for_timeout(120)
except Exception: pass
except Exception: pass
res=pg.evaluate(EXTRACT_JS, {'sels':[sel] if sel else [], 'candSels':CAND_SELS})
rows=clean_rows(res.get('rows',[]))
d1=[r['label'] for r in rows if r['depth']==1]
data={'num':num,'name':name,'base':base,'sel':res.get('sel'),'gnb':rows}
json.dump(data, open(os.path.join(GNB_DIR,f'{num}.json'),'w',encoding='utf-8'), ensure_ascii=False, indent=1)
print(f"[{num:2}.{name}] sel={res.get('sel')} rows={len(rows)} d1={len(d1)} :: {d1[:10]}")
except Exception as e:
print(f"[{num:2}.{name}] 실패: {type(e).__name__}: {e}")
pg.close()
br.close()
if __name__=='__main__':
nums=[int(x) for x in sys.argv[1:]] or list(SITES.keys())
run(nums)