mssql-python 驱动没有实现 DB-API 2.0 规范中的该 callproc() 方法。 调用 callproc() 会引发 NotSupportedError。 相反,使用ODBC {CALL ...} 转义序列配合标准查询执行方法。
存储过程的基本执行
无参数
使用 {CALL} 转义序列来执行存储过程。 本例调用 sp_databases 系统存储过程,该过程不取参数,每个数据库返回一行:
import mssql_python
conn = mssql_python.connect(connection_string)
cursor = conn.cursor()
cursor.execute("{CALL sp_databases}")
for row in cursor:
print(row.DATABASE_NAME, row.DATABASE_SIZE)
输入参数
使用位置或命名占位符传递参数:
# Single parameter
cursor.execute(
"{CALL dbo.uspGetManagerEmployees(?)}", (16,)
)
for row in cursor:
print(row.FirstName, row.LastName)
输出参数
声明并检索输出参数
SQL Server 存储过程可以通过输出参数返回值。 使用 Transact-SQL(T-SQL)变量捕获输出值,然后用以下 SELECT 语句检索:
cursor.execute("""
DECLARE @total_out MONEY;
SELECT @total_out = SUM(TotalDue)
FROM Sales.SalesOrderHeader
WHERE CustomerID = %(customer_id)s;
SELECT @total_out AS TotalAmount;
""", {"customer_id": 29825})
row = cursor.fetchone()
total = row.TotalAmount
print(f"Customer total: ${total}")
多个输出参数
通过声明独立变量来捕捉多个输出值。 同样的模式适用于任何具有 OUTPUT 参数的存储过程:
cursor.execute("""
DECLARE @total_orders INT, @total_spent MONEY;
SELECT @total_orders = COUNT(*), @total_spent = SUM(TotalDue)
FROM Sales.SalesOrderHeader
WHERE CustomerID = %(cust_id)s;
SELECT @total_orders AS OrderCount, @total_spent AS TotalSpent;
""", {"cust_id": 29825})
stats = cursor.fetchone()
print(f"Orders: {stats.OrderCount}, Total spent: ${stats.TotalSpent}")
返回值
捕获存储过程返回值
执行返回结果的存储过程:
cursor.execute("EXECUTE dbo.uspGetEmployeeManagers @BusinessEntityID = %(emp_id)s", {"emp_id": 5})
rows = cursor.fetchall()
if rows:
print(f"Found {len(rows)} managers in chain")
for row in rows:
print(f" Manager: {row.FirstName} {row.LastName}")
else:
print("No managers found")
带输出参数的返回值
cursor.execute("""
DECLARE @return_value INT, @message NVARCHAR(500);
SELECT @return_value = CASE WHEN COUNT(*) > 0 THEN 0 ELSE 1 END,
@message = CASE WHEN COUNT(*) > 0 THEN N'Customer found' ELSE N'Customer not found' END
FROM Sales.Customer WHERE CustomerID = %(cust_id)s;
SELECT @return_value AS ReturnCode, @message AS Message;
""", {"cust_id": 29825})
result = cursor.fetchone()
print(f"Return code: {result.ReturnCode}, Message: {result.Message}")
结果集
单一结果集
执行返回单个结果集的存储过程:
cursor.execute(
"{CALL dbo.uspGetBillOfMaterials(?, ?)}",
(800, "2026-01-01")
)
customers = cursor.fetchall()
for row in customers:
print(f"{row.ProductAssemblyID}: {row.ComponentDesc}")
多个结果集
某些存储过程返回多个结果集。 使用 nextset() 在它们之间导航:
cursor.execute("""
SELECT TOP 1 SalesOrderID, OrderDate, TotalDue
FROM Sales.SalesOrderHeader WHERE CustomerID = 29825;
SELECT TOP 3 Name, ListPrice
FROM Production.Product WHERE ListPrice > 0
ORDER BY ListPrice DESC;
""")
# First result set: order header
order = cursor.fetchone()
print(f"Order: {order.SalesOrderID}, Date: {order.OrderDate}")
cursor.nextset()
# Second result set: products
print("Products:")
for item in cursor:
print(f" {item.Name}: ${item.ListPrice}")
查看更多结果集
使用 nextset() 遍历存储过程返回的所有结果集:
cursor.execute("SELECT TOP 3 ProductID, Name FROM Production.Product; SELECT TOP 3 FirstName, LastName FROM Person.Person")
result_set_num = 1
while True:
print(f"--- Result Set {result_set_num} ---")
for row in cursor:
print(row)
if not cursor.nextset():
break
result_set_num += 1
涉及存储过程的事务
显式交易控制
在事务中封装多个存储过程调用以确保原子性:
conn.autocommit = False
try:
cursor.execute("{CALL dbo.DebitAccount(?, ?)}", (1001, 100.00))
cursor.execute("{CALL dbo.CreditAccount(?, ?)}", (1002, 100.00))
conn.commit()
print("Transfer completed")
except mssql_python.DatabaseError as e:
conn.rollback()
print(f"Transfer failed: {e}")
让存储过程管理事务
如果存储过程处理自己的事务:
conn.autocommit = True # Let SP manage transactions
cursor.execute("""
DECLARE @result INT;
EXECUTE @result = dbo.TransferFunds
@FromAccount = %(from_acc)s,
@ToAccount = %(to_acc)s,
@Amount = %(amount)s;
SELECT @result AS TransferResult;
""", {"from_acc": 1001, "to_acc": 1002, "amount": 100.00})
result = cursor.fetchone()
if result.TransferResult == 0:
print("Transfer successful")
错误处理
捕捉存储过程错误
处理由存储过程或 Transact-SQL 语句引发的异常:
try:
cursor.execute("{CALL dbo.DangerousProcedure}")
except mssql_python.ProgrammingError as e:
# Handle SQL errors raised by RAISERROR or THROW
print(f"Stored procedure error: {e}")
except mssql_python.DatabaseError as e:
# Handle other database errors
print(f"Database error: {e}")
捕获PRINT语句和信息消息
SQL ServerPRINT语句和严重级别低于 11 的RAISERROR会在执行后捕获到cursor.messages中。 每个条目均为一个 (message_type, message_text) 元组。 当 PRINT 在结果集之前运行时,它会先占用一个不包含任何行的结果集,因此应先读取 cursor.messages,然后调用 nextset() 以访问这些行:
cursor.execute("PRINT 'Operation complete'; SELECT 1 AS Status")
for msg_type, msg_text in cursor.messages:
print(f"Server message: {msg_text}")
cursor.nextset()
row = cursor.fetchone()
print(f"Status: {row.Status}")
当存储过程跨多个结果集发出 PRINT 消息时,请在 execute() 之后读取 cursor.messages,并在每次调用 nextset() 后再次读取,以捕获每个结果集中的消息:
cursor.execute("""
PRINT 'Starting first result set';
SELECT TOP 3 ProductID, Name FROM Production.Product;
PRINT 'Starting second result set';
SELECT TOP 3 FirstName, LastName FROM Person.Person;
""")
all_messages = []
while True:
all_messages.extend(cursor.messages)
if cursor.description:
for row in cursor:
print(row)
if not cursor.nextset():
break
for _, text in all_messages:
print(f"Server: {text}")
完整 cursor.messages API 请参见 光标管理。
最佳做法
使用命名参数
命名参数更清晰,且保持顺序独立性:
# Recommended: {CALL} with positional parameters
cursor.execute(
"{CALL dbo.uspGetBillOfMaterials(?, ?)}",
(800, "2026-01-01")
)
# Also valid: EXECUTE with named T-SQL parameters
cursor.execute("""
EXECUTE dbo.uspGetBillOfMaterials
@StartProductID = ?,
@CheckDate = ?
""", (800, "2026-01-01"))
处理可空的输出参数
从存储过程获取输出参数时,检查是否为 NULL:
cursor.execute("""
DECLARE @optional_value NVARCHAR(100);
SELECT @optional_value = Color FROM Production.Product WHERE ProductID = %(id)s;
SELECT @optional_value AS OutputValue;
""", {"id": 1})
result = cursor.fetchone()
if result.OutputValue is not None:
print(f"Value: {result.OutputValue}")
else:
print("No value returned")
在存储过程中使用 SET NOCOUNT ON
为了提升性能和更干净的结果处理,确保你的存储过程包括:
CREATE PROCEDURE dbo.MyProcedure
AS
BEGIN
SET NOCOUNT ON; -- Prevents "n rows affected" messages
-- procedure logic
END
示例:完整工作流程
这里有一个实际的例子,可以调用存储过程,检索输出值并处理错误:
import mssql_python
def get_employee_report(manager_id: int) -> dict:
"""Look up a manager's employees and compute average vacation hours."""
conn = mssql_python.connect(connection_string)
conn.autocommit = False
cursor = conn.cursor()
try:
# Get manager info
cursor.execute("""
SELECT BusinessEntityID, JobTitle
FROM HumanResources.Employee
WHERE BusinessEntityID = %(mgr)s
""", {"mgr": manager_id})
mgr = cursor.fetchone()
print(f"Manager {mgr.BusinessEntityID}: {mgr.JobTitle}")
# Get direct reports via stored procedure
cursor.execute("{CALL dbo.uspGetManagerEmployees(?)}", (manager_id,))
employees = cursor.fetchall()
print(f"Found {len(employees)} employee(s)")
# Compute average vacation hours
cursor.execute("""
DECLARE @avg_hours INT;
SELECT @avg_hours = AVG(VacationHours)
FROM HumanResources.Employee;
SELECT @avg_hours AS AvgVacation;
""")
avg = cursor.fetchone().AvgVacation
print(f"Avg vacation hours: {avg}")
conn.commit()
return {"manager": mgr.JobTitle, "reports": len(employees), "avg_vacation": avg}
except mssql_python.DatabaseError as e:
conn.rollback()
raise
finally:
cursor.close()
conn.close()
# Usage
result = get_employee_report(manager_id=16)
使用 OUTPUT INSERTED 获取生成的键值
要从 INSERT (有无存储过程)中检索身份值,可以用 OUTPUT INSERTED 代替 SCOPE_IDENTITY()。 这种方法在同一结果集中返回该值:
cursor.execute("""
INSERT INTO Production.ProductCategory (Name)
OUTPUT INSERTED.ProductCategoryID
VALUES (%(name)s)
""", {"name": "Custom Parts"})
new_id = cursor.fetchval()
print(f"New category ID: {new_id}")
这种模式适用于任何带有身份列的表,且不需要存储过程。
不支持的功能
callproc()
如果调用 cursor.callproc(),mssql-python 驱动程序会引发 NotSupportedError。 使用 cursor.execute("{CALL ...}") 或 cursor.execute("EXECUTE ...") 替代,正如本文所展示的。
表值参数(TVP)
当前版本(1.11.0) mssql-python不支持表值参数。 如果你需要将一组行传递给存储过程,可以使用以下替代方案:
- 先插入临时表,再让存储过程从中读取数据。
- 使用
bulkcopy()将数据加载到暂存表中。 - 传递一个 JSON 字符串,然后在该过程中使用
OPENJSON解析它。