Converting empty string to Null when inserting/updating

I am using the following query to calculate date differences:
select ........DATEDIFF(d, recruitment_advertising.advertising_date, career_details.RTS_Email AS Datetime) AS Ad_to_RTS_days FROM ....

I have stored all my dates as NVARCHAR because of the issues with localization.
If the value is an empty String my output is eg: -38700. which is way off and incorrect. Some of the values in my table areNULL and they produce the correct result.

Is there a T-SQL statement to replace empy Strings with the NULL value in my tables.
I'd like to use it as a trigger when inserting or updating to convert empty strings to NULL
before the values are inserted.

Thanks guys.

You REALLY should store your dates as a datetime. There is no localization "Problem" with datetimes if you use them correctly, and you can't sort and/or generate good indexes if they are stored in a nvarchar field (Unless you specifically use the YYYYMMDD or YYYY-MM-DD format).

That aside, yes, try NULLIF() like:

INSERT INTO MyTable(col1) VALUES (NULLIF(@.val1,''))

http://msdn.microsoft.com/library/default.asp?url=/library/en-us/tsqlref/ts_nos-nz_3uhy.asp

Converting data types

I've inherited a monster database that needs updating. It has a field called Birthdays that has been a text field and I want to convert it to datetime. The data entry people have been sloppy from time to time and some of the values generate errors when you try and change datatypes. Is there a way to change the datatype and have the oddball values left blank without bombing out the conversion process?

Thanksheh. how big is the database?

if it's not that big, then you can simply store them in a temp table, and move the values over one by one in a cursor style fashion. If it's absolutely enormous, I'd actually create another table just to store the data, and the same cursor fasion. Just make sure you put them in transactions so you can rollback accordingly based on any errors generated.

Since it's a one shot deal, you can even do it out of the database, and just manipulate all the values via code, then do the alter so it doesn't cause any problems.

Do you know specifically what the bad dates are? If so, via code, you can convert the dates based on how the errors were inputted, and update those records. :-\