views:

900

answers:

2

I have some code like this that I use to do a BULK INSERT of a data file into a table, where the data file and table name are variables:

DECLARE @sql AS NVARCHAR(1000)
SET @sql = 'BULK INSERT ' + @tableName + ' FROM ''' + @filename + ''' WITH (CODEPAGE=''ACP'', FIELDTERMINATOR=''|'')'

EXEC (@sql)

The works fine for standard tables, but now I need to do the same sort of thing to load data into a temporary table (for example, #MyTable). But when I try this, I get the error:

Invalid Object Name: #MyTable

I think the problem is due to the fact that the BULK INSERT statement is constructed on the fly and then executed using EXEC, and that #MyTable is not accessible in the context of the EXEC call.

The reason that I need to construct the BULK INSERT statement like this is that I need to insert the filename into the statement, and this seems to be the only way to do that. So, it seems that I can either have a variable filename, or use a temporary table, but not both.

Is there another way of achieving this - perhaps by using OPENROWSET(BULK...)?


UPDATE: OK, so what I'm hearing is that BULK INSERT & temporary tables are not going to work for me. Thanks for the suggestions, but moving more of my code into the dynamic SQL part is not practical in my case.

Having tried OPENROWSET(BULK...), it seems that that suffers from the same problem, i.e. it cannot deal with a variable filename, and I'd need to construct the SQL statement dynamically as before (and thus not be able to access the temp table).

So, that leaves me with only one option which is to use a non-temp table and achieve process isolation in a different way (by ensuring that only one process can be using the tables at any one time - I can think of several ways to do that).

It's annoying. It would have been much more convenient to do it the way I originally intended. Just one of those things that should be trivial, but ends up eating a whole day of your time...

A: 

http://msdn.microsoft.com/en-us/library/ms191503.aspx

i would advice to create table with unique name before bulk inserting.

Andrey
Thanks, but I'm not sure what you're suggesting? If I create the table with a unique name that therefore isn't known till runtime, then I still have to dynamically construct the BULK INSERT statement?Wait - do you mean I should create a *non-temporary* table (which I can then drop later on)? Or a global temp table, I suppose.Hmmm... I guess that *might* work, but that would mean that all the *other* code that processes the table once the data has been inserted would *also* have to be dynamically-constructed in order to use the correct table name. That sounds like a nightmare...
Gary McGill
you answered your questions :) "other code" also have to be aware about temp table. you can do this: create nontemp table, bulk-insert, copy from nontemp to temp, drop nontemp, have fun
Andrey
No, do not use a global temp table. All this will do is ensure that two different users running at roughly the same time will mix their results together (or get an error when they try to create a ##table that already exists). I have yet to find a real, practical use case for a global temp table.
Aaron Bertrand
@Andrey: I think there's a flaw in your plan... in order to "copy from nontemp to temp", I'd need to use dynamic SQL (so that I could insert the name of the nontemp table with the generated unique name) - and once again, the dynamic SQL would not be able to access the temp table.
Gary McGill
@Aaron: I only meant that if I was going to create a non-temp table with a unique name, then it could equally well be a global temp table *with a unique name*. The only benefit being that the table would be automatically dropped at some point, so there would be no risk of the database getting cluttered up with tables that I'd created but failed to drop explicitly. I agree that - on their own - global temp tables do not actually help solve the problem at hand.
Gary McGill
+1  A: 

You could always construct the #temp table in dynamic SQL. For example, right now I guess you have been trying:

CREATE TABLE #tmp(a INT, b INT, c INT);

DECLARE @sql NVARCHAR(1000);

SET @sql = N'BULK INSERT #tmp ...' + @variables;

EXEC(@sql);

SELECT * FROM #tmp;

This makes it tougher to maintain (readability) but gets by the scoping issue:

DECLARE @sql NVARCHAR(MAX);

SET @sql = N'CREATE TABLE #tmp(a INT, b INT, c INT);

BULK INSERT #tmp ...' + @variables + ';

SELECT * FROM #tmp;';

EXEC master..sp_ExecuteSQL @sql;
Aaron Bertrand
@Aaron: unfortunately, I have tons of work to do with the data before I'm finished with it, so making all that code 'dynamic' would be impractical.
Gary McGill