SQL Server TRY CATCH FINALLY

Question

I have a scenario where I need something similar to .NET's try-catch-finally block.

On my try, I will CREATE a #temp table , INSERT data to it & process other data sets based on #temp .

On CATCH then RAISERROR . Is it possible to have a FINALLY block to DROP #temp ? Below is the pseudo code:

BEGIN TRY

  CREATE TABLE #temp
  (
     --columns
  )
  --Process data with other data sets

END TRY
BEGIN CATCH

  EXECUTE usp_getErrorMessage

END CATCH
BEGIN FINALLY

  DROP TABLE #temp

END FINALLY

Answer 1

While not exactly the same as FINALLY, the T-SQL version of Try-Catch does allow that code that needs execute after both the Try and Catch blocks can occur after the end of the END CATCH statement. Using the question code as an example:

    BEGIN TRY
      CREATE TABLE #temp
       (
         --columns
       )
      --Process data with other data sets
    END TRY
    BEGIN CATCH
    EXECUTE usp_getErrorMessage
    END CATCH;

IF OBJECT_ID('tempdb..#temp') IS NOT NULL -- Check for table existence
    DROP TABLE #temp;

The DROP TABLE command will execute whether the Try or Catch execute. See: BOL Try...Catch

Answer 2

Instead of creating a table you could just declare a table variable (which will automatically go away when the query ends).

BEGIN TRY
DECLARE @temp TABLE
(
    --columns
)
--do stuff
END TRY
BEGIN CATCH
--do other stuff
END CATCH

Answer 3

there is no FINALLY equivalent.
an alternative may be table variables but is not exactly the same and must be evaluated on a case by case basis.
there is a SO question with details very useful to make an informed choice.
with table variables you don't need to clean up like you do with temp tables

Answer 4

"FINALLY" is often, but not always, functionally identical to having the "final" code follow the TRY/CATCH (without a formal "FINALLY" block). Where it is different is the case where something in the TRY/CATCH blocks could cause execution to end, such as a return statement.

For example, a pattern I've used is to open a cursor, then have the cursor-using code in the TRY block, with the cursor close/deallocate following the TRY/CATCH block. This works fine if the blocks won't exit the code being executed. However, if the TRY CATCH block does, for example, a RETURN (which sounds like a bad idea), if there were a FINALLY block, it would get executed, but with the "final" code placed after the TRY / CATCH, as T-SQL requires, should those code blocks cause the execution to end, that final code won't be called, potentially leaving an inconsistent state.

So, while very often you can just put the code after the TRY/CATCH, it will be a problem if anything in those blocks could terminate without falling through to the cleanup code.

Answer 5

Local temp tables (eg, "#Temp") are automatically dropped when the SQL connection ends. It's good practice to include an explicit DROP command anyway, but if it doesn't execute, the table will still be dropped.

If you must ensure that a DROP executes as soon as possible, you'll have to repeat the DROP command in a CATCH clause, since there's no FINALLY :

-- create temp table;
BEGIN TRY
    -- use temp table;
    -- drop temp table;
END TRY
BEGIN CATCH
    -- drop temp table;
    THROW;  -- rethrow the error
END CATCH

Table variables are an alternative: they're dropped when the variable goes out of scope. However, table variables do not support statistics, so if the table variable is large and used in multiple queries, it may not perform as well as a temp table.

Answer 6

using custom error number to indicate there no real error, just final code?

-- create temp table;
BEGIN TRY
    -- use temp table;
    THROW 50555;
END TRY
BEGIN CATCH
    -- drop temp table;
    IF ERROR_NUMBER() <> 50555
        THROW;  -- rethrow the error
END CATCH

Answer 7

The correct answer in this case is the one proposed by @Dave Bennett; after the TRY/CATCH block check for the existence of the table and drop it.

But what if you are raising an exception out of your CATCH and you need to do some "FINALLY" type processing?

Could it be as simple as setting a variable in the CATCH and checking it after you fall out of the CATCH?

DECLARE @is_error BIT = 0;

BEGIN TRY

  --Process data with other data sets

END TRY
BEGIN CATCH
  -- Your exception handling code here
  
  SET @is_error = 1;

END CATCH

-- Your "FINALLY" code here.

-- Then Check if you need to RAISERROR
IF @is_error = 0
BEGIN
    -- Your success code 
END
ELSE
BEGIN
    -- Your fail code
    -- RAISERROR
END;

Answer 8

With T-SQL, code placed after the TRY-CATCH block will get executed. So the answer to the question is simply to add a DROP TABLE IF EXISTS #temp right after the END CATCH like so:

BEGIN TRY
  CREATE TABLE #temp(
     --columns
  )
  --Process data with other data sets
END TRY
BEGIN CATCH
  EXECUTE usp_getErrorMessage
END CATCH
-- Anything under this line will execute regardless of if the code reached the CATCH block or not.
DROP TABLE IF EXISTS #temp;  -- Works in SQL Server 2016+, for older versions see here: https://stackoverflow.com/a/31171650/15596537