| 123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139 |
- """精确数值工具:Decimal 运算、缺失值传播、权重封顶。
- 设计原则(沿用第一阶段第三版代码):
- * 金额/数量一律用 ``Decimal``,不经过浮点,避免累加误差;
- * 任何无法识别的输入都向上传播为 ``None``(缺数),绝不当成 0;
- * 除数为 0 或缺失时返回 ``None``,由上层写入“计算说明”而不是填 0。
- """
- from __future__ import annotations
- import re
- import unicodedata
- from decimal import Decimal, DefaultContext, InvalidOperation, getcontext
- DefaultContext.prec = 50 # 新建的处理/接口线程也必须继承 50 位精度
- getcontext().prec = 50
- D = Decimal
- #: 指数基点(2025 年 1 月 = 1000)
- INDEX_BASE = D(1000)
- #: 单一价格样本权重上限
- WEIGHT_CAP = D("0.1")
- def number(raw: object) -> Decimal | None:
- """把任意原始单元格转换为 Decimal;无法识别时返回 None(不是 0)。"""
- if raw is None:
- return None
- if isinstance(raw, Decimal):
- return raw if raw.is_finite() else None
- if isinstance(raw, (int, float)):
- value = D(str(raw))
- return value if value.is_finite() else None
- text = unicodedata.normalize("NFKC", str(raw)).strip().strip('"\'').strip()
- text = text.replace("−", "-")
- if re.search(r"人民币|RMB|CNY|[¥¥€£]", text, re.I):
- return None
- text = re.sub(r"^(?:USD\s*|\$\s*)|(?:\s*USD|\s*美元)$", "", text, flags=re.I).strip()
- if text.startswith('(') and text.endswith(')'):
- text = '-' + text[1:-1]
- if re.fullmatch(r"[+-]?\d{1,3}(?:[, ]\d{3})+(?:\.\d+)?", text):
- text = text.replace(',', '').replace(' ', '')
- if text in {"", "-", "--", "/", "NaN", "nan", "NULL", "null"}:
- return None
- try:
- value = D(text)
- except (InvalidOperation, ValueError):
- return None
- return value if value.is_finite() else None
- def ratio(numerator: Decimal | None, denominator: Decimal | None, scale: object = 1) -> Decimal | None:
- """numerator / denominator * scale;任一缺失或分母为 0 时返回 None。"""
- if numerator is None or denominator is None or denominator == 0:
- return None
- return numerator / denominator * D(str(scale))
- def growth(current: Decimal | None, previous: Decimal | None) -> Decimal | None:
- """环比增长率 = 本期/上期 - 1(返回小数,写 Excel 时按百分比显示)。"""
- value = ratio(current, previous)
- return None if value is None else value - 1
- def index_of(current: Decimal | None, base: Decimal | None) -> Decimal | None:
- """固定基期指数 = 当期/基期 * 1000。"""
- return ratio(current, base, INDEX_BASE)
- def to_text(value: object) -> str:
- """把 Decimal 等写成不带科学计数法、不做舍入的文本。"""
- if value is None:
- return ""
- if isinstance(value, Decimal):
- if value == 0:
- return "0"
- text = format(value.normalize(), "f")
- return text
- return str(value)
- class StrictSum:
- """SQLite 聚合函数:任一成员无法识别则整体返回 NULL,而不是静默当 0。"""
- def __init__(self) -> None:
- self.total = D(0)
- self.bad = False
- def step(self, value: object) -> None:
- parsed = number(value)
- if parsed is None:
- self.bad = True
- else:
- self.total += parsed
- def finalize(self) -> str | None:
- return None if self.bad else str(self.total)
- def weights(amounts: list[Decimal]) -> list[Decimal]:
- """按年度进口金额计算价格指数样本权重。
- 规则:
- * 样本 < 10 个时取消 10% 上限,直接按金额占比(否则无解);
- * 样本 >= 10 个时,单样本不超过 10%,超额部分在剩余样本中按金额占比迭代再分配。
- """
- if not amounts or any(a is None or a <= 0 for a in amounts):
- return []
- total = sum(amounts)
- if total <= 0:
- return []
- if len(amounts) < 10:
- return [a / total for a in amounts]
- free = set(range(len(amounts)))
- result = [D(0)] * len(amounts)
- remaining = D(1)
- while free:
- subtotal = sum(amounts[i] for i in free)
- if subtotal <= 0:
- break
- over = {i for i in free if amounts[i] * remaining / subtotal > WEIGHT_CAP}
- if not over:
- for i in free:
- result[i] = amounts[i] * remaining / subtotal
- break
- for i in over:
- result[i] = WEIGHT_CAP
- remaining -= WEIGHT_CAP * len(over)
- free -= over
- return result
|