#!/usr/bin/env python # -*- coding: utf-8 -*- """ 导出择校决策数据层:国家线 + 院校专业历年复试线(线差)+ 院校索引 口径要点 -------- 1. 分数线来源 core_hist_lines(2020-2024,源:考研365 各校分数线页), 一条记录 = 某校某专业某年的复试基本线。 2. 国家线来源 src_national_lines(2022-2026,研招网原文),按 codes 字段 精确匹配专业类(如 1253 会计 = 201 分),未命中时退回门类「其他学科专业」。 3. 线差 gap = 院校复试线 - 同年同专业类国家线(A 区)。这是跨年份可比的 难度指标,直接消除国家线波动。 4. 同校同专业同年可能存在多个学院多条线,取 MIN(total) 作为该校该专业的 最低门槛,并记录学院数 colleges。 5. 计划数 plan 直接取源表字段,填充率 78.2%,为空则置 None,不臆造。 输出 ---- data/national.json 国家线(按年) data/rank_index.json 专业码索引(供首页热榜/搜索) data/rank/{code}.json 该专业下院校难度明细(按需加载) data/school_lines.json 院校索引(全校线差、有线专业数) """ import json import os import sqlite3 import collections DB = r'D:/择校/数据库/zexiao_2027.db' OUT = r'D:/择校/website/data' RANK_DIR = os.path.join(OUT, 'rank') # 专业码前两位 -> 门类名(国家线 category_code 体系) CAT = { '01': '哲学', '02': '经济学', '03': '法学', '04': '教育学', '05': '文学', '06': '历史学', '07': '理学', '08': '工学', '09': '农学', '10': '医学', '11': '军事学', '12': '管理学', '13': '艺术学', '14': '交叉学科', } os.makedirs(RANK_DIR, exist_ok=True) # 2022/2023 年国家线源缺失「专硕单列线」,只有门类合并线。 # 这些类别的专硕当年实际另划线(如 2022 会计 193 vs 管理学 353), # 用门类线算线差会严重失真,故这两年一律不给线差,宁缺勿错。 SPLIT_2022_2023 = ('1251', '1252', '1253', '1254', '1255', '1256', '1257', '0452', '0451', '0453', '1057') def clean_name(s): """去掉 "(专业学位)" 之类前缀修饰,保留可读名。""" if not s: return '' s = s.replace('(专业学位)', '').replace('(专业学位)', '') return s.strip() def main(): con = sqlite3.connect(DB) con.row_factory = sqlite3.Row # ---------- 1. 国家线 ---------- nat_rows = [dict(r) for r in con.execute( 'select year,category,category_code,subject,codes,a_total,a_single100,a_single101,policy ' 'from src_national_lines where a_total is not null')] national = collections.defaultdict(list) for r in nat_rows: codes = [] if r['codes']: try: codes = json.loads(r['codes']) except Exception: codes = [] national[str(r['year'])].append({ 'cat': r['category_code'], 'name': r['category'], 'subj': r['subject'] or '', 'codes': codes, 'total': r['a_total'], 's100': r['a_single100'], 's101': r['a_single101'], }) def nat_for(code, year): """按 专业类码(codes 精确) > 门类「其他/各学科专业」> 该门类首条 的顺序取国家线。 2022/2023 年对已单列但源缺失的专硕类别返回 None。""" pool = national.get(str(year)) or [] if not pool: return None if str(year) in ('2022', '2023') and code[:4] in SPLIT_2022_2023: return None c4 = code[:4] for it in pool: if it['codes'] and c4 in it['codes']: return it cat = code[:2] loose = [it for it in pool if it['cat'] == cat and not it['codes']] for it in loose: if '其他学科' in it['subj'] or '各学科' in it['subj']: return it if loose: return loose[0] return None with open(os.path.join(OUT, 'national.json'), 'w', encoding='utf-8') as f: json.dump({'years': sorted(national.keys()), 'by': national}, f, ensure_ascii=False, separators=(',', ':')) # ---------- 2. 院校库(去曾用名重复,按 id 建索引) ---------- schools = json.load(open(os.path.join(OUT, 'schools.json'), encoding='utf-8')) sid_info = {} for s in schools: sid = s.get('id') if sid is None or sid in sid_info: continue sid_info[sid] = { 'name': s['name'], 'prov': s.get('prov') or '', 'city': s.get('city') or '', 'level': s.get('level') or '双非', 'd27': s.get('d27') or 0, 'm27': s.get('m27') or 0, 'enroll27': s.get('enroll27') or 0, } majors = json.load(open(os.path.join(OUT, 'majors.json'), encoding='utf-8')) maj_by_code = {m['code']: m for m in majors} # ---------- 3. 分数线聚合 ---------- raw = con.execute(''' select school_id_ggyx as sid, school, code, name, year, total, plan, college from core_hist_lines where code is not null and length(code) >= 4 and total is not null and total > 0 and (data_flag is null) ''').fetchall() # (code, sid, year) -> {line, colleges:set, plan} agg = collections.defaultdict(lambda: {'line': None, 'colleges': set(), 'plan': None}) code_names = {} for r in raw: code = str(r['code']) sid = r['sid'] y = int(r['year']) k = (code, sid, y) a = agg[k] t = r['total'] a['line'] = t if a['line'] is None else min(a['line'], t) a['colleges'].add(r['college'] or '') if r['plan'] and r['plan'] > 0: # 同校同专业同年多行时,plan 常是专业级数值重复出现,取 MAX 避免虚高 a['plan'] = r['plan'] if a['plan'] is None else max(a['plan'], r['plan']) if code not in code_names and r['name']: code_names[code] = clean_name(r['name']) # code -> sid -> [year rows] tree = collections.defaultdict(lambda: collections.defaultdict(list)) for (code, sid, y), a in agg.items(): nat = nat_for(code, y) gap = None nat_total = None if nat and nat.get('total'): nat_total = nat['total'] gap = a['line'] - nat_total tree[code][sid].append({ 'y': y, 'l': a['line'], 'n': nat_total, 'g': gap, 'p': a['plan'], 'c': len([c for c in a['colleges'] if c]), }) # ---------- 4. 逐专业输出 ---------- index = [] for code, by_sid in tree.items(): rows = [] gaps_all = [] for sid, yrs in by_sid.items(): info = sid_info.get(sid) if not info: continue yrs.sort(key=lambda x: x['y']) # 近三年线差(越近权重越高:1,2,3) gs = [x for x in yrs if x['g'] is not None] g3 = None if gs: recent = gs[-3:] w = list(range(1, len(recent) + 1)) g3 = round(sum(x['g'] * wi for x, wi in zip(recent, w)) / sum(w), 1) gaps_all.append(g3) last = ys = None for x in yrs: if x['l']: last = x plan = None for x in yrs: if x['p']: plan = x['p'] rows.append({ 'id': sid, 'name': info['name'], 'prov': info['prov'], 'lv': info['level'], 'ys': yrs, 'g3': g3, 'gy': len(gs), 'last': last['l'] if last else None, 'ly': last['y'] if last else None, 'plan': plan, 'd27': info['d27'], 'm27': info['m27'], 'en27': info['enroll27'], }) if not rows: continue rows.sort(key=lambda r: (r['g3'] is None, -(r['g3'] or 0))) m = maj_by_code.get(code) name = (m['name'] if m else code_names.get(code)) or code ml = m['ml'] if m else CAT.get(code[:2], '') payload = { 'code': code, 'name': name, 'ml': ml, 'n': len(rows), 'rows': rows, } with open(os.path.join(RANK_DIR, code + '.json'), 'w', encoding='utf-8') as f: json.dump(payload, f, ensure_ascii=False, separators=(',', ':')) index.append({ 'code': code, 'name': name, 'ml': ml, 'n': len(rows), 'ns': (m['n_school'] if m else 0), 'med': sorted(gaps_all)[len(gaps_all) // 2] if gaps_all else None, 'min': min(gaps_all) if gaps_all else None, 'max': max(gaps_all) if gaps_all else None, }) index.sort(key=lambda x: -x['n']) with open(os.path.join(OUT, 'rank_index.json'), 'w', encoding='utf-8') as f: json.dump({'years': [2020, 2021, 2022, 2023, 2024], 'n_schools': len({r['sid'] for r in raw}), 'n_codes': len(index), 'majors': index}, f, ensure_ascii=False, separators=(',', ':')) # ---------- 5. 院校索引 ---------- sch = {} for code, by_sid in tree.items(): for sid, yrs in by_sid.items(): info = sid_info.get(sid) if not info: continue d = sch.setdefault(sid, {'id': sid, **info, 'gaps': [], 'majors': []}) gs = [x['g'] for x in yrs if x['g'] is not None] if gs: d['gaps'].append(sum(gs) / len(gs)) d['majors'].append({'code': code, 'name': (maj_by_code.get(code) or {}).get('name') or code_names.get(code) or code, 'g': round(sum(gs) / len(gs), 1) if gs else None, 'last': [x['l'] for x in yrs if x['l']][-1] if gs else None}) out_sch = [] for sid, d in sch.items(): g = d['gaps'] d['majors'].sort(key=lambda x: (x['g'] is None, -(x['g'] or 0))) out_sch.append({ 'id': sid, 'name': d['name'], 'prov': d['prov'], 'level': d['level'], 'nm': len(d['majors']), 'g': round(sum(g) / len(g), 1) if g else None, 'd27': d['d27'], 'm27': d['m27'], 'en27': d['enroll27'], }) out_sch.sort(key=lambda x: -x['nm']) with open(os.path.join(OUT, 'school_lines.json'), 'w', encoding='utf-8') as f: json.dump({'n': len(out_sch), 'schools': out_sch}, f, ensure_ascii=False, separators=(',', ':')) # ---------- 诊断 ---------- print('专业码输出:', len(index)) print('院校有分数线:', len({r['sid'] for r in raw})) print('有国家线的年份:', sorted(national.keys())) print('TOP10 专业(校数):', [(x['name'], x['n'], x['med']) for x in index[:10]]) print('院校索引:', len(out_sch)) big = [x for x in index if x['n'] >= 20] print('>=20 校的专业:', len(big)) total = sum(os.path.getsize(os.path.join(RANK_DIR, f)) for f in os.listdir(RANK_DIR)) print('rank 目录体积: %.1f MB / %d 文件' % (total / 1e6, len(os.listdir(RANK_DIR)))) if __name__ == '__main__': main()