mssql-python 驱动程序将 SQL Server 的 decimal、numeric 和 money 数据类型映射为 Python 的 decimal.Decimal,以便进行准确的财务和科学计算。 SQL Server 提供了以下精确的数值类型:
| SQL 类型 | Python 类型 | 精准率 | 用例 |
|---|---|---|---|
| decimal(p,s) | decimal.Decimal |
1到38位数字 | 精确计算 |
| numeric(p,s) | decimal.Decimal |
1到38位数字 | 和十进制一样 |
| money | decimal.Decimal |
19位数字,4位小数 | 货币 |
| smallmoney | decimal.Decimal |
10位数字,4位小数 | 小额货币 |
| float | float |
~15位数字 | 近似 |
| real | float |
~7位数字 | 近似 |
Python十进制类型
接收小数值
十进制列返回 Python decimal.Decimal 对象:
from decimal import Decimal
import mssql_python
conn = mssql_python.connect(connection_string)
cursor = conn.cursor()
cursor.execute(
"SELECT SubTotal, TaxAmt, TotalDue "
"FROM Sales.SalesOrderHeader WHERE SalesOrderID = 43659"
)
row = cursor.fetchone()
print(type(row.SubTotal)) # <class 'decimal.Decimal'>
print(row.SubTotal) # Decimal('20565.6206')
print(row.TaxAmt) # Decimal('1971.5149')
print(row.TotalDue) # Decimal('23153.2339')
发送十进制值
传递 Decimal 对象以实现精确插入:
from decimal import Decimal
cursor.execute("""
CREATE TABLE #ProductDemo (
Name NVARCHAR(50),
Price DECIMAL(10,2),
Cost DECIMAL(10,2)
)
""")
cursor.execute("""
INSERT INTO #ProductDemo (Name, Price, Cost)
VALUES (%(name)s, %(price)s, %(cost)s)
""", {
"name": "Widget",
"price": Decimal("199.99"),
"cost": Decimal("87.50")
})
conn.commit()
避免对财务数据使用浮点数
浮动型存在精度问题,不适合用于财务计算。 为了获得准确结果,务必使用 Decimal :
from decimal import Decimal
# Bad: float has precision issues
price = 0.1 + 0.2 # 0.30000000000000004
# Good: Decimal is exact
price = Decimal("0.1") + Decimal("0.2") # Decimal('0.3')
# Use string constructor for literal values
correct = Decimal("19.99") # Exact
avoid = Decimal(19.99) # Might introduce float imprecision
货币类型
与货币柱的工作
使用 Python 十进制对象检索和更新 SQL Server 的资金列:
from decimal import Decimal
# Create and populate a temp table with a money column
cursor.execute("""
CREATE TABLE #Accounts (
ID INT,
AccountID NVARCHAR(20),
Balance MONEY
)
""")
cursor.execute(
"INSERT INTO #Accounts (ID, AccountID, Balance) VALUES (1, %(acct)s, %(bal)s)",
{"acct": "ACCT-001", "bal": Decimal("1234.5678")}
)
conn.commit()
cursor.execute("SELECT AccountID, Balance FROM #Accounts WHERE ID = 1")
row = cursor.fetchone()
print(row.Balance) # Decimal('1234.5678')
print(type(row.Balance)) # <class 'decimal.Decimal'>
# Money arithmetic
cursor.execute("""
UPDATE #Accounts
SET Balance = Balance + %(amount)s
WHERE ID = %(id)s
""", {"amount": Decimal("100.00"), "id": 1})
conn.commit()
货币格式化
将十进制值格式化为带有符号和逗号分隔符的货币字符串:
from decimal import Decimal
def format_currency(value: Decimal, symbol: str = "$") -> str:
"""Format decimal as currency string."""
return f"{symbol}{value:,.2f}"
cursor.execute("SELECT TotalDue FROM Sales.SalesOrderHeader")
for row in cursor:
print(format_currency(row.TotalDue)) # $1,234.56
十进制精度与刻度
了解精度和标度
- 精度(p):总位数
- 标度 (s):小数点后的位数
例如,decimal(10, 2) 可以存储从 -99999999.99 到 99999999.99 的值,而 decimal(5, 4) 可以存储从 -9.9999 到 9.9999 的值。
在参数中指定精度
mssql-python驱动自动从十进制值推算精度和扩展性:
from decimal import Decimal
# Create a temp table for the measurement value
cursor.execute("CREATE TABLE #Measurements (Value DECIMAL(18,6))")
# The driver infers precision from the Decimal value
value = Decimal("123.456789")
cursor.execute("INSERT INTO #Measurements (Value) VALUES (%(v)s)", {"v": value})
conn.commit()
控制四舍五入
使用 quantize() 方法将 Decimal 值舍入到指定的小数位数:
from decimal import Decimal, ROUND_HALF_UP, ROUND_DOWN
price = Decimal("19.999")
# Round to 2 decimal places
rounded = price.quantize(Decimal("0.01"), rounding=ROUND_HALF_UP)
print(rounded) # Decimal('20.00')
# Round down (truncate)
truncated = price.quantize(Decimal("0.01"), rounding=ROUND_DOWN)
print(truncated) # Decimal('19.99')
财务计算
安全算术
使用 Decimal 并采用适当的舍入方式来实现财务计算:
from decimal import Decimal, ROUND_HALF_UP
def calculate_total(price: Decimal, quantity: int, tax_rate: Decimal) -> Decimal:
"""Calculate order total with tax."""
subtotal = price * quantity
tax = (subtotal * tax_rate).quantize(Decimal("0.01"), rounding=ROUND_HALF_UP)
return subtotal + tax
cursor.execute("SELECT ListPrice FROM Production.Product WHERE ProductID = 1")
price = cursor.fetchval()
total = calculate_total(
price=price,
quantity=3,
tax_rate=Decimal("0.0825") # 8.25% tax
)
cursor.execute("""
CREATE TABLE #OrderCalc (
ProductID INT, Quantity INT, Total DECIMAL(10,2), Tax DECIMAL(10,2)
)
""")
cursor.execute("""
INSERT INTO #OrderCalc (ProductID, Quantity, Total, Tax)
VALUES (%(prod)s, %(qty)s, %(total)s, %(tax)s)
""", {"prod": 1, "qty": 3, "total": total, "tax": total * Decimal("0.0825")})
百分比计算
计算基于百分比的折扣和调整:
from decimal import Decimal
def calculate_discount(price: Decimal, discount_percent: Decimal) -> Decimal:
"""Calculate discounted price."""
discount = (price * discount_percent / 100).quantize(Decimal("0.01"))
return price - discount
original = Decimal("99.99")
discounted = calculate_discount(original, Decimal("15")) # 15% off
print(f"Original: ${original}, After discount: ${discounted}")
分摊金额
将总金额平均分配给多个方,处理剩余部分:
from decimal import Decimal, ROUND_DOWN
def split_amount(total: Decimal, ways: int) -> list[Decimal]:
"""Split amount evenly, handling remainder."""
per_person = (total / ways).quantize(Decimal("0.01"), rounding=ROUND_DOWN)
remainder = total - (per_person * ways)
amounts = [per_person] * ways
# Add remainder to first person
amounts[0] += remainder
return amounts
total = Decimal("100.00")
split = split_amount(total, 3)
print(split) # [Decimal('33.34'), Decimal('33.33'), Decimal('33.33')]
print(sum(split)) # Decimal('100.00') - always equals original
聚合操作
SUM和AVG带小数点
SQL Server 聚合函数返回十进制类型以保证精度:
# SUM preserves decimal type
cursor.execute("SELECT SUM(TotalDue) AS TotalSales FROM Sales.SalesOrderHeader")
total_sales = cursor.fetchval()
print(type(total_sales)) # <class 'decimal.Decimal'>
# AVG might return more decimal places
cursor.execute("SELECT AVG(ListPrice) AS AvgPrice FROM Production.Product")
avg_price = cursor.fetchval()
# Round to desired precision
avg_price = avg_price.quantize(Decimal("0.01"))
在聚合中处理NULL
当没有匹配的行时,SQL Server 聚合函数可以返回None:
cursor.execute("SELECT SUM(TotalDue) FROM Sales.SalesOrderHeader WHERE CustomerID = 0")
total_discount = cursor.fetchval()
# SUM returns NULL if no rows match
if total_discount is None:
total_discount = Decimal("0.00")
用于十进制的输出转换器
输出转换器使用中cursor.description所示的Python类型作为键,而非ODBC SQL整数类型。 将转换器注册为 decimal.Decimal ,使其对 、 decimal、 numeric和 money 列有效,smallmoney这些列都映射到 decimal.Decimal。
转换成浮点(当精度不关键时)
为了兼容期望浮点值的库,定义输出转换器:
import mssql_python
from decimal import Decimal
def decimal_to_float(value):
"""Convert Decimal to float."""
if value is None:
return None
return float(value) # value is already a Decimal object
conn = mssql_python.connect(connection_string)
# Convert decimals to float for compatibility with libraries that expect float
conn.add_output_converter(Decimal, decimal_to_float)
cursor = conn.cursor()
cursor.execute("SELECT ListPrice FROM Production.Product WHERE ProductID = 1")
row = cursor.fetchone()
print(type(row.ListPrice)) # <class 'float'>
自定义货币格式转换器
定义一个自定义输出转换器,自动将十进制值格式化为货币字符串:
from decimal import Decimal
def money_to_string(value):
"""Convert Decimal to formatted string."""
if value is None:
return None
return f"${value:,.2f}" # value is already a Decimal object
conn.add_output_converter(Decimal, money_to_string)
cursor = conn.cursor()
cursor.execute("CREATE TABLE #Accounts (Balance DECIMAL(19,4))")
cursor.execute(
"INSERT INTO #Accounts (Balance) VALUES (%(bal)s)",
{"bal": Decimal("1234.56")}
)
conn.commit()
cursor.execute("SELECT Balance FROM #Accounts")
for row in cursor:
print(row.Balance) # "$1,234.56"
常见模式
币种转换
利用汇率在不同货币之间转换货币金额:
from decimal import Decimal
def convert_currency(amount: Decimal, rate: Decimal) -> Decimal:
"""Convert currency at given exchange rate."""
return (amount * rate).quantize(Decimal("0.01"))
usd_amount = Decimal("100.00")
eur_rate = Decimal("0.92")
eur_amount = convert_currency(usd_amount, eur_rate)
print(f"${usd_amount} = €{eur_amount}")
余额更新和交易
使用适当的锁定提示交易,确保资金转移并防止更新丢失:
from decimal import Decimal
def transfer_funds(conn, from_account: int, to_account: int, amount: Decimal):
"""Transfer funds between accounts atomically."""
cursor = conn.cursor()
conn.autocommit = False
try:
balances = {}
for account_id in sorted((from_account, to_account)):
cursor.execute(
"SELECT Balance FROM #Accounts WITH (UPDLOCK) WHERE ID = %(id)s",
{"id": account_id}
)
row = cursor.fetchone()
if row is None:
raise ValueError(f"Account {account_id} not found")
balances[account_id] = row.Balance
if balances[from_account] < amount:
raise ValueError("Insufficient funds")
# Debit source
cursor.execute("""
UPDATE #Accounts SET Balance = Balance - %(amount)s
WHERE ID = %(id)s
""", {"amount": amount, "id": from_account})
# Credit destination
cursor.execute("""
UPDATE #Accounts SET Balance = Balance + %(amount)s
WHERE ID = %(id)s
""", {"amount": amount, "id": to_account})
conn.commit()
except Exception:
conn.rollback()
raise
finally:
conn.autocommit = True
# Set up demo accounts
cursor = conn.cursor()
cursor.execute("CREATE TABLE #Accounts (ID INT, Balance MONEY)")
cursor.executemany(
"INSERT INTO #Accounts (ID, Balance) VALUES (%(id)s, %(bal)s)",
[{"id": 1, "bal": Decimal("1000.00")}, {"id": 2, "bal": Decimal("500.00")}]
)
conn.commit()
# Usage
transfer_funds(conn, from_account=1, to_account=2, amount=Decimal("500.00"))
转账会预先锁定这两行,并按 ID 从小到大的顺序,分别对每个账户发出一个单独的 SELECT ... WITH (UPDLOCK)。
UPDLOCK 可防止出现丢失更新模式:两笔事务读取同一余额,并基于陈旧数据进行更新。 按排序后的顺序分别发出加锁语句,才能保证锁的获取顺序,防止同一两个账户之间方向相反的转账以相反顺序获取锁而发生死锁。
带小数的批量插入
使用 executemany() 插入包含 Decimal 值的多行:
from decimal import Decimal
cursor.execute("""
CREATE TABLE #BulkProducts (
Name NVARCHAR(50),
Price DECIMAL(10,2),
Cost DECIMAL(10,2)
)
""")
products = [
{"name": "Widget A", "price": Decimal("19.99"), "cost": Decimal("8.50")},
{"name": "Widget B", "price": Decimal("29.99"), "cost": Decimal("12.75")},
{"name": "Widget C", "price": Decimal("39.99"), "cost": Decimal("18.00")},
]
cursor.executemany("""
INSERT INTO #BulkProducts (Name, Price, Cost)
VALUES (%(name)s, %(price)s, %(cost)s)
""", products)
conn.commit()
最佳做法
财务数学一定要用十进制
绝不要用浮动指数进行财务计算;为了精确,始终使用十进制类型:
# Wrong: float precision issues
price = 0.10
quantity = 3
total = price * quantity # 0.30000000000000004
# Correct: Decimal is exact
price = Decimal("0.10")
quantity = 3
total = price * quantity # Decimal('0.30')
使用字符串构造器
通过字符串构造十进制数值,以避免浮点数精度误差:
# Good: exact value
amount = Decimal("123.45")
# Risky: might inherit float imprecision
amount = Decimal(123.45) # Could be 123.4500000000000028...
展示或储存前总是圆的
在显示或保存到数据库之前,将十进制值四舍五入到适当的小数位数:
from decimal import Decimal, ROUND_HALF_UP
calculated = Decimal("123.456789")
display = calculated.quantize(Decimal("0.01"), rounding=ROUND_HALF_UP)
如有需要,设置十进制上下文
十进制上下文精度影响 计算结果,而非从字符串解析或数据库读取的值。 查询始终返回一个 money 或 decimal 值,并保留其存储的完整小数位数,而不受 getcontext().prec 的影响。 上下文只有在你用它计算时才会生效。 要观察精度的效果,可以执行除法等操作。
from decimal import Decimal, getcontext
# A money value read from SQL Server arrives at its full stored scale,
# regardless of the context precision.
cursor.execute(
"SELECT SubTotal FROM Sales.SalesOrderHeader WHERE SalesOrderID = 43659"
)
subtotal = cursor.fetchone().SubTotal
print(subtotal) # Decimal('20565.6206')
# The context precision governs the calculation, not the fetched value.
# Split the subtotal into three equal installments.
getcontext().prec = 10
print(subtotal / 3) # 6855.206867 (10 significant digits)
getcontext().prec = 28 # Reset to default
print(subtotal / 3) # 6855.206866666666666666666667 (28 significant digits)