sansenhoshi/ancient_date_parser
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

兼容性增强

Browse Source 
资源数据补全
依赖更新
This commit is contained in:
sansen
2026-08-07 16:12:59 +08:00
parent a8d6e6f870
commit a850c72ab5
4 changed files with 275 additions and 33 deletions
Download Patch File Download Diff File Expand all files Collapse all files
parser.py
+250 -25
View File
@@ -3,14 +3,21 @@ import re
import pandas as pd
from opencc import OpenCC
from era_data import dynasty_alias, KNOWN_ERAS
from era_data import dynasty_alias, KNOWN_ERAS, VARIANT_MAP, TEMPLE_ALIASES
cc = OpenCC('t2s')
def normalize_chars(text):
"""OpenCC 之外的异体字/俗字归一化。文本与 CSV 两侧必须走同一归一化，保证一致。"""
for variant, canonical in VARIANT_MAP.items():
text = text.replace(variant, canonical)
return text
# === 中文数字转阿拉伯数字 ===
def chinese_to_num(text):
text = text.replace("卅", "三十").replace("卌", "四十")
text = text.replace("卅", "三十").replace("卌", "四十").replace("拾", "十")
if text in ["元", "元年"]:
return 1
@@ -18,22 +25,36 @@ def chinese_to_num(text):
'零': 0, '〇': 0, '○': 0,
'一': 1, '二': 2, '三': 3, '四': 4, '五': 5, '六': 6, '七': 7, '八': 8, '九': 9,
'十': 10, '百': 100, '千': 1000,
'壹': 1, '貳': 2, '叁': 3, '肆': 4, '伍': 5, '陸': 6, '柒': 7, '捌': 8, '玖': 9, '拾': 10,
'壹': 1, '貳': 2, '贰': 2, '叁': 3, '肆': 4, '伍': 5, '陸': 6, '陆': 6, '柒': 7, '捌': 8, '玖': 9, '拾': 10,
'卄': 20, '廿': 20, '卅': 30, '卌': 40,
}
if text in chinese_numerals:
return chinese_numerals[text]
if '千' in text:
parts = text.split('千')
total = chinese_to_num(parts[0]) * 1000 if parts[0] else 1000
if len(parts) > 1 and parts[1]:
total += chinese_to_num(parts[1])
return total
if '百' in text:
parts = text.split('百')
total = chinese_to_num(parts[0]) * 100 if parts[0] else 100
if len(parts) > 1 and parts[1]:
total += chinese_to_num(parts[1])
return total
if '十' in text:
parts = text.split('十')
total = 0
if parts[0] == '':
total += 10
else:
total += chinese_numerals.get(parts[0], 0) * 10
total += chinese_to_num(parts[0]) * 10
if len(parts) > 1 and parts[1]:
total += chinese_numerals.get(parts[1], 0)
total += chinese_to_num(parts[1])
return total
if '廿' in text or '卄' in text:
@@ -68,14 +89,71 @@ def normalize_dynasty_name(dynasty):
# === 年号表 ===
def _cc(s):
"""CSV 单元格归一化：空值→''，其余走 OpenCC 繁转简 + 异体字归一"""
if pd.isna(s):
return ''
return normalize_chars(cc.convert(str(s)))
era_df = pd.read_csv("中国年号.csv")
era_df["年号"] = era_df["年号"].map(lambda x: cc.convert(str(x)))
era_df["所属朝代"] = era_df["所属朝代"].map(lambda x: cc.convert(str(x)))
era_df["年号"] = era_df["年号"].map(_cc)
era_df["所属朝代"] = era_df["所属朝代"].map(_cc)
era_df["名号"] = era_df["名号"].map(_cc)
def _safe_span(v):
"""CSV 使用年数列（可能为空或非数字）"""
try:
return int(float(v))
except (TypeError, ValueError):
return None
era_dict = {}
# 年号使用年数：用于干支/年数结果的使用期校验（如庆元仅6年，期内无庚子）
era_span = {}
for _, row in era_df.iterrows():
dynasty, era, start = row["所属朝代"], row["年号"], int(row["公元纪年"])
if not era:
continue
era_dict.setdefault(dynasty, {})[era] = start
span = _safe_span(row["使用年数"])
if span is not None:
era_span.setdefault(dynasty, {}).setdefault(era, span)
# 有效年号白名单 = 手工精选 ∪ CSV 年号（自动补全：白名单不应小于数据表）。
# 排除：民国（有专门匹配段）、带括号的年号（如"(后元)"，防止与东汉"中元"等歧义）
_csv_eras = {
e for e in era_df["年号"]
if e and e != '民国' and '(' not in e and '（' not in e and 2 <= len(e) <= 8
}
KNOWN_ERAS_ALL = KNOWN_ERAS | _csv_eras
# === 庙号表 ===
# 名号列（如"高宗""太宗"）表示整个在位期，公元取该名号下最早的年份（即位首年）。
# "代宗(景帝)" 等带括号的写法拆成两种叫法都收录；只收祖/宗/帝/王/后/皇结尾（或帝开头）的名号，
# 排除"拖雷""孺子婴"等非庙号条目。
def _is_temple_name(name):
return len(name) >= 2 and (name.startswith('帝') or name.endswith(('祖', '宗', '帝', '王', '后', '皇')))
temple_dict = {}
# 庙号在位年数（该名号下各年号使用年数之和）：用于名号+年数结果的使用期校验
temple_span = {}
for _, row in era_df.iterrows():
name, dynasty, start = row["名号"], row["所属朝代"], int(row["公元纪年"])
if not name or not dynasty:
continue
for part in re.split(r'[()（）]', name):
part = part.strip()
if not part or not _is_temple_name(part):
continue
part = TEMPLE_ALIASES.get(part, part)
if not _is_temple_name(part):
continue
temple_dict.setdefault(dynasty, {}).setdefault(part, start)
span = _safe_span(row["使用年数"])
if span is not None:
d = temple_span.setdefault(dynasty, {})
d[part] = d.get(part, 0) + span
# === 干支纪年 ===
@@ -111,21 +189,24 @@ def arabic_to_chinese_year(num: int) -> str:
# === 古代年号正则 ===
# 从 era_dict 构建朝代匹配组（按长度降序排列，优先匹配长朝代名如"北宋"而非"北"）
# 额外添加常见但不在CSV中的朝代简称（如"汉"）
# 合并 temple_dict 的键：周/秦等年号列全空的朝代也必须有匹配能力
_extra_dynasty_names = ['汉', '晋', '宋', '周']
_dynasty_names = sorted(
list(era_dict.keys()) + [d for d in _extra_dynasty_names if d not in era_dict],
list(era_dict.keys()) + list(temple_dict.keys())
+ [d for d in _extra_dynasty_names if d not in era_dict and d not in temple_dict],
key=len, reverse=True
)
_dynasty_alt = '|'.join(re.escape(d) for d in _dynasty_names)
# 常见朝代前缀（大、皇等），用于匹配"大唐""大清"等
_dynasty_prefix = '大皇前后胡'
# 非元年数字字符（排除元，避免歧义：建元元年中第一个元是年号的一部分）
_year_digits = '〇零一二三四五六七八九十廿卄卅卌'
# 非元年数字字符（排除元，避免歧义：建元元年中第一个元是年号的一部分）。
# 含阿拉伯/全角数字（"康熙20年""乾隆２０年"）与大写数字（"道光贰拾年"）。
_year_digits = '〇零一二三四五六七八九十廿卄卅卌' + '0123456789０１２３４５６７８９' + '壹貳贰叁肆伍陸陆柒捌玖拾'
# 已知年号白名单见 era_data.py
# 预编译已知年号正则片段（按长度降序，优先匹配长年号如"太平兴国"而非"太平"）
_known_era_alt = '|'.join(sorted(KNOWN_ERAS, key=len, reverse=True))
_known_era_alt = '|'.join(sorted(KNOWN_ERAS_ALL, key=len, reverse=True))
_known_era_group = r'(?P<era>' + _known_era_alt + r')'
_yd = '[' + _year_digits + ']'
@@ -173,15 +254,18 @@ def extract_era_years(text):
# 去除标记用的方括号（用户标注用，非原文内容）
text = text.replace('[', '').replace(']', '')
original_text = text
simplified_text = cc.convert(text)
simplified_text = normalize_chars(cc.convert(text))
results = []
# 覆盖集：先匹配的段落占据位置，后续段落跳过重叠（防止同一文本被多段重复摘出）
covered = set()
# === 1. 民国 ===
minguo_pattern = re.compile(r'民[国國]\s*([〇零一二三四五六七八九十百廿卄卅卌元\d]+)\s*年')
minguo_pattern = re.compile(r'民[国國]\s*([〇零一二三四五六七八九十百廿卄卅卌元\d０-９壹貳贰叁肆伍陸陆柒捌玖拾]+)\s*年')
for m in minguo_pattern.finditer(simplified_text):
year_text = m.group(1)
year_num = chinese_to_num(year_text)
start, end = m.start(), m.end()
covered.update(range(start, end))
raw_fan = original_text[start:end]
results.append({
"原文": raw_fan,
@@ -207,6 +291,7 @@ def extract_era_years(text):
year_in_era = ((gz_offset - base_gz) % 60) + 1
gregorian = 1912 + year_in_era - 1
start, end = m.start(), m.end()
covered.update(range(start, end))
raw_fan = original_text[start:end]
results.append({
"原文": raw_fan,
@@ -221,16 +306,82 @@ def extract_era_years(text):
"类型": "民国纪年（干支）",
})
# === 2b. 年号+公元（如"清康熙公元一六八七年丁卯"） ===
# 年号后直接给出公元年份：年数=公元-年号起始+1，与"年号+年数"公式自洽；
# 紧跟的干支用于校验年份一致性。置于公元段之前，合并结果优先、避免重复摘出。
_era_gongyuan = re.compile(
r'(?P<dynasty>(?:(?<![一-鿿])[' + _dynasty_prefix + r'](?:' + _dynasty_alt + r'))|(?:(?<![一-鿿])' + _dynasty_alt + r'))?'
+ r'(?P<era>' + _known_era_alt + r')'
+ r'公元\s*(?P<gy_year>[〇零○O一二三四五六七八九\d０-９]{3,4})\s*年?'
+ r'(?P<gy_ganzhi>' + _gz_pair + r')?'
)
for m in _era_gongyuan.finditer(simplified_text):
start, end = m.start(), m.end()
if any(pos in covered for pos in range(start, end)):
continue
era = cc.convert(m.group("era"))
year_num = chinese_to_num(m.group("gy_year"))
if not (1000 <= year_num <= 2100):
continue
dynasty = normalize_dynasty_name(m.group("dynasty") or "")
base = None
multi = []
if dynasty:
for d, eras in era_dict.items():
if dynasty in d or d in dynasty:
if era in eras:
base = eras[era]
dynasty = d
break
if base is None:
for d, eras in era_dict.items():
if era in eras:
multi.append((d, eras[era]))
if multi:
dynasty, base = multi[0]
if base is None:
continue
covered.update(range(start, end))
year_in_era = year_num - base + 1
ganzhi = m.group("gy_ganzhi")
result = {
"原文": original_text[start:end],
"原文(简体)": m.group(0),
"朝代": dynasty,
"年号": era,
"年数": year_in_era,
"公元": year_num,
"公元中文": f"{arabic_to_chinese_year(year_num)}年",
"位置": {"起始": start, "结束": end},
"类型": "年号+公元",
}
if len(multi) > 1:
result["多候选"] = True
result["候选"] = [list(c) for c in multi]
# 公元超出年号使用期（如"康熙公元1950年"）→ 标注，值保留
span = era_span.get(dynasty, {}).get(era)
if span is not None and year_num > base + span - 1:
result["超出使用期"] = True
# 紧跟的干支与公元年份不符（如"一六八九年丁卯"，1689应为己巳）→ 标注
if ganzhi:
result["干支"] = ganzhi
if gregorian_to_ganzhi(year_num) != ganzhi:
result["干支不符"] = True
results.append(result)
# === 2. 公元（允许无"年"字）===
gongyuan_pattern = re.compile(
r'(公元)?\s*([〇零○O一二三四五六七八九\d]{3,4})\s*(?:（.*?）)?\s*(年)?'
r'(公元)?\s*([〇零○O一二三四五六七八九\d０-９]{3,4})\s*(?:（.*?）)?\s*(年)?'
)
for m in gongyuan_pattern.finditer(simplified_text):
start, end = m.start(), m.end()
if any(pos in covered for pos in range(start, end)):
continue
year_text = m.group(2)
year_num = chinese_to_num(year_text)
if not (1000 <= year_num <= 2100):
continue
start, end = m.start(), m.end()
covered.update(range(start, end))
raw_fan = original_text[start:end]
era_type = "公元纪年（中华人民共和国）" if year_num >= 1949 else "公元纪年（近代/其他）"
results.append({
@@ -243,7 +394,6 @@ def extract_era_years(text):
})
# === 3. 干支纪年（如"清雍正乙巳十一月十二午时生"）===
covered = set()
def _process_ganzhi_match(m):
"""处理干支纪年匹配"""
raw_dynasty = m.group("dynasty") if "dynasty" in m.groupdict() else ""
@@ -255,6 +405,7 @@ def extract_era_years(text):
# 查找朝代和基准年
base = None
multi = []
if dynasty:
for d, eras in era_dict.items():
if dynasty in d or d in dynasty:
@@ -263,11 +414,12 @@ def extract_era_years(text):
dynasty = d
break
if not base:
# 无朝代前缀：收集全部候选（如"大定"=北周581/金1161），取首个但标注歧义
for d, eras in era_dict.items():
if era in eras:
base = eras[era]
dynasty = d
break
multi.append((d, eras[era]))
if multi:
dynasty, base = multi[0]
if not base:
return
@@ -281,7 +433,7 @@ def extract_era_years(text):
return
covered.update(range(start, end))
raw_fan = original_text[start:end]
results.append({
result = {
"原文": raw_fan,
"原文(简体)": m.group(0),
"朝代": dynasty,
@@ -292,13 +444,22 @@ def extract_era_years(text):
"公元中文": f"{arabic_to_chinese_year(gregorian)}年",
"位置": {"起始": start, "结束": end},
"类型": "干支纪年",
})
}
if len(multi) > 1:
result["多候选"] = True
result["候选"] = [list(c) for c in multi]
# 使用期校验：换算结果超出年号实际使用年数（如"庆元庚子"→1240，庆元仅1195–1200）。
# 值保留，仅标注，供人工判断（碑刻原数据常与年号实际期间不符）。
span = era_span.get(dynasty, {}).get(era)
if span is not None and gregorian > base + span - 1:
result["超出使用期"] = True
results.append(result)
# 模式A：有朝代前缀（如"清雍正乙巳"）
# 单字年号加前瞻(?![地支])，防止单字年号"丙"吞掉天干"丙午"中的"丙"
_gz_era_group = r'(?P<era>' + '|'.join(
(re.escape(e) + (r'(?![子丑寅卯辰巳午未申酉戌亥])' if len(e) == 1 else ''))
for e in sorted(KNOWN_ERAS, key=len, reverse=True)
for e in sorted(KNOWN_ERAS_ALL, key=len, reverse=True)
) + r')'
ganzhi_dynasty = re.compile(
r'(?P<dynasty>(?:(?<![一-鿿])[' + _dynasty_prefix + r'](?:' + _dynasty_alt + r'))|(?:(?<![一-鿿])' + _dynasty_alt + r'))'
@@ -309,7 +470,7 @@ def extract_era_years(text):
_process_ganzhi_match(m)
# 模式B：无朝代前缀（如"嘉庆癸亥"），年号≥2字符避免天干误匹配
_long_era_names = [e for e in KNOWN_ERAS if len(e) >= 2]
_long_era_names = [e for e in KNOWN_ERAS_ALL if len(e) >= 2]
_long_era_alt = '|'.join(sorted(_long_era_names, key=len, reverse=True))
ganzhi_no_dynasty = re.compile(
r'(?P<dynasty>)'
@@ -319,6 +480,49 @@ def extract_era_years(text):
for m in ganzhi_no_dynasty.finditer(simplified_text):
_process_ganzhi_match(m)
# === 3b. 庙号纪年（如"宋高宗""清圣祖"，指整个在位期，公元=即位首年）===
# 庙号必须带朝代前缀（"高宗"在唐/宋/清均有，单独出现有歧义）
_temple_names = sorted({t for d in temple_dict for t in temple_dict[d]}, key=len, reverse=True)
_temple_alt = '|'.join(re.escape(t) for t in _temple_names)
_temple_pattern = re.compile(
r'(?P<dynasty>(?:(?<![一-鿿])[' + _dynasty_prefix + r'](?:' + _dynasty_alt + r'))|(?:(?<![一-鿿])' + _dynasty_alt + r'))'
+ r'(?P<temple>' + _temple_alt + r')'
# 后面紧跟"X年/元年"时不按庙号匹配，交给古代纪年按年号解析（如"元泰定三年"）
+ r'(?!(?:' + _yd + r'+年|元年))'
)
for m in _temple_pattern.finditer(simplified_text):
start, end = m.start(), m.end()
if any(pos in covered for pos in range(start, end)):
continue
dynasty = normalize_dynasty_name(m.group("dynasty") or "")
temple = m.group("temple")
base = None
if dynasty:
# 先精确匹配朝代，再模糊匹配（如"宋"→"南宋"）
for d in temple_dict:
if d == dynasty and temple in temple_dict[d]:
dynasty, base = d, temple_dict[d][temple]
break
if base is None:
for d in temple_dict:
if (dynasty in d or d in dynasty) and temple in temple_dict[d]:
dynasty, base = d, temple_dict[d][temple]
break
if base is None:
continue
covered.update(range(start, end))
raw_fan = original_text[start:end]
results.append({
"原文": raw_fan,
"原文(简体)": m.group(0),
"朝代": dynasty,
"庙号": temple,
"公元": base,
"公元中文": f"{arabic_to_chinese_year(base)}年",
"位置": {"起始": start, "结束": end},
"类型": "庙号纪年",
})
# === 4. 古代年号 ===
def _process_era_match(m, year_override=None):
@@ -336,6 +540,7 @@ def extract_era_years(text):
year_text = year_override if year_override is not None else m.group("year")
year_num = chinese_to_num(year_text)
base = None
multi = []
if dynasty:
for d, eras in era_dict.items():
@@ -345,6 +550,7 @@ def extract_era_years(text):
dynasty = d
break
else:
# 无朝代前缀：收集全部候选（如"建元"=西汉-140/东晋343/南齐479），取首个但标注歧义
candidates = []
for d, eras in era_dict.items():
if era in eras:
@@ -353,13 +559,22 @@ def extract_era_years(text):
dynasty, base = candidates[0]
elif len(candidates) > 1:
dynasty, base = candidates[0]
multi = candidates
if base is None:
# 名号+年数兜底（如"汉惠帝三年"→即位首年-194 + 3 - 1，指皇帝在位第N年）
if dynasty:
for d, temples in temple_dict.items():
if (dynasty in d or d in dynasty) and era in temples:
base = temples[era]
dynasty = d
break
start, end = m.start(), m.end()
raw_fan = original_text[start:end]
covered.update(range(start, end))
if base:
gregorian = base + year_num - 1
results.append({
result = {
"原文": raw_fan,
"原文(简体)": m.group(0),
"朝代": dynasty,
@@ -369,7 +584,17 @@ def extract_era_years(text):
"公元中文": f"{arabic_to_chinese_year(gregorian)}年",
"位置": {"起始": start, "结束": end},
"类型": "古代纪年",
})
}
if len(multi) > 1:
result["多候选"] = True
result["候选"] = [list(c) for c in multi]
# 使用期校验：年数结果超出年号使用期（或名号在位年数）时标注，值保留
span = era_span.get(dynasty, {}).get(era)
if span is None:
span = temple_span.get(dynasty, {}).get(era)
if span is not None and gregorian > base + span - 1:
result["超出使用期"] = True
results.append(result)
else:
results.append({
"原文": raw_fan,