如何使用Python中的表值参数访问SQL Server 2008存储过程

Question

I'm looking for a way to take a result set and use it to find records in a table that resides in SQL Server 2008 – without spinning through the records one at a time. The result sets that will be used to find the records could number in the hundreds of thousands. So far I am pursuing creating a table in memory using sqlite3 and then trying to feed that table to a stored procedure that takes a table valued parameter. The work on the SQL Server side is done, the user defined type is created, the test procedure accepting a table valued parameter exists and I've tested it through TSQL and it appears to work just fine. In Python a simple in memory table was created through sqlite3. Now the catch, the only documentation I have found for accessing a stored procedure with a table valued parameter uses ADO.Net and VB, nothing in Python. Unfortunately, I'm not enough of a programmer to translate. Has anyone used a SQL Server stored procedure with a table valued parameter? Is there another approach I should look into?

Here are some links: Decent explanation of table valued parameters and how to set them up in SQL and using in .Net

http://www.sqlteam.com/article/sql-server-2008-table-valued-parameters

http://msdn.microsoft.com/en-us/library/bb675163.aspx#Y2142

Explanation of using ADO in Python – almost what I need, just need the structured parameter type. http://www.mayukhbose.com/python/ado/ado-command-3.php

My simple code

--TSQL to create type on SQL database
create Type PropIDList as Table
(Prop_Id BigInt primary key)
--TSQL to create stored procedure on SQL database.  Note reference to 
create procedure PropIDListTest @PIDList  PropIDList READONLY
as
SET NOCOUNT ON
select * from
@PIDList p 
SET NOCOUNT OFF
--TSQL to test objects.  
--Declare variable as user defined type (table that has prop_id)
declare @pidlist as propidlist
--Populate variable
insert into @pidlist(prop_id)
values(1000)
insert into @pidlist(prop_id)
values(2000)

--Pass table variable to stored procedure
exec PropIDListTest @pidlist

Now the tough part – Python.

Here is the code creating the in memory table

import getopt, sys, string, os, tempfile, shutil
import _winreg,win32api, win32con
from win32com.client import Dispatch
from adoconstants import *
import sqlite3

conn1 = sqlite3.connect(':memory:')
c = conn1.cursor()
# Create table
c.execute('''create table PropList
        (PropID bigint)''')

# Insert a row of data
c.execute("""insert into PropList
                  values (37921019)""")

# Save (commit) the changes
conn1.commit()
c.execute('select * from PropList order by propID')
# lets print out what we have to make sure it works
for row in c:
    print row

Ok, my attempt at connecting through Python

conn = Dispatch('ADODB.Connection')
conn.ConnectionString = "Provider=sqloledb.1; Data Source=nt38; Integrated Security = SSPI;database=pubs"   
conn.Open()
cmd = Dispatch('ADODB.Command')
cmd.ActiveConnection = conn
cmd.CommandType = adCmdStoredProc
cmd.CommandText = "PropIDListTest @pidlist = ?"
param1 = cmd.CreateParameter('@PIDList', adUserDefined)  # I “think” the parameter type is the key and yes it is most likely wrong here.
cmd.Parameters.Append(param1)
cmd.Parameters.Value = conn1 # Yeah, this is probably wrong as well

(rs, status) = cmd.Execute()
while not rs.EOF:
    OutputName = rs.Fields("Prop_ID").Value.strip().upper()
    print OutputName
    rs.MoveNext()
rs.Close()
rs = None
conn.Close()
conn = None
 # We can also close the cursor if we are done with it
c.close()
conn1.close()

Answer 1

I have coded TVPs from ADO.NET before. Here is a question on TVPs in classic ADO that I am interested in, sql server - Classic ADO and Table-Valued Parameters in Stored Procedure - Stack Overflow . It does not give a direct answer but alternatives.

• The option of XML is easier, you have probably already considered it; it would require more server side processing.
• Here is the msdn link for low level ODBC programming of TVPs. Table-Valued Parameters (ODBC) . This one is the closest answer if you can switch to ODBC.
• You could pass a csv string to nvarchar(max) and then pass it to a CLR SplitString function , that one is fast but has default behaviour I disagree with.

Please post back what works or does not here.