Options
Issue with declaring @pv for BLOB data in 10.4.8.62
paschott
Posts: 71 Bronze 4
I generated scripts for our databases that disable triggers and constraints before modifying data in the database. I noticed that for several of these scripts, there is a DECLARE @pv command at the beginning of the script, followed by a DISABLE TRIGGER command, followed by a "GO". I recorded the output of my script and noticed that there were several errors because the variable "@pv" was not declared.
Looking through the rest of the script, it seemed that the @pv variable was declared within each transaction/batch, but for the very first data transaction/batch, it was declared outside of that section.
Shouldn't the DECLARE @pv command be done after all of the constraint and trigger disables, just before we actually start manipulating data?
Looking through the rest of the script, it seemed that the @pv variable was declared within each transaction/batch, but for the very first data transaction/batch, it was declared outside of that section.
Shouldn't the DECLARE @pv command be done after all of the constraint and trigger disables, just before we actually start manipulating data?
Comments
It doesn't seem to use or need the pointer for nvarchar(max) on SQL 2008, but it does declare @pv again in every block of split transactions.
Pretty sure it's with XML or Text data. This is coming from SQL Server 2005 going to SQL 2012 (though that ideally shouldn't matter). When I pulled up one of the smaller scripts that failed, I noticed that the only DECLARE @pv prior to using it was not in the same transaction block. When I ran the Data Compare directly, it worked and I figured the script would work as well, but I can't seem to run the saved script.
I'll try another data compare to see if that makes a difference. I was just hoping to avoid that as the generated scripts are hundreds of MB and if I can't reuse the scripts, that makes the process a little less useful.
I'm using SQLCMD to run the scripts. Otherwise, what is the point of saving out the data compare to a script. I want the ability to re-create an empty set of databases and run just the scripts to populate them with a base set of data to use for testing and development work.
Obviously that data will change in small ways over time, but populating a working set is a good start. I'm almost positive that this worked properly in a prior version of Data Compare.
I'll regenerate the scripts soon, but SQLCMD should work as a standard SQL Script. I can't declare a variable in one batch and then use it in a different batch. If Data Compare is not appropriately declaring the variables within their batches, that's a problem that should be addressed.
As for the file length, SQLCMD keeps going through the file after the failures. It just doesn't have a variable @pv to process the BLOB data in those particular batches.
Today I plan to try the scripts again without the option to break the transactions apart. Hopefully that will help. However, there still seems to be something off when choosing to split the script into multiple batches and declaring the @pv variable correctly. I'm pretty sure this worked in an earlier release, but don't recall which and I tweaked my scripts since then to disable triggers as well as PK/FK constraints when generating them.
I really think there is an issue with Data Compare not declaring the @pv variable in the correct place if the options are chosen to disable FK constraints and triggers.