3

Say I have three tables or four tables that I want to check for values present and I don't want any of them to be null. IF they are null, I will send an email notification per row. What would be the best way to go about this? Should I make a stored procedure with IF statements separately?

I say this because all these tables are not having any constraints, they just pull data out of an excel sheet. So making a table type variable and unioning or inner joining returns nothing as even the column names are different, but the information they store is the same kind of information. So for example if in one table, a column exists called 'DueDate', in the other table, this same sort of information is called 'PaymentDate'. How would I perform such a check on all these tables?

Update: Here's what I've thought of as of now: Say the two tables are A and B, then

DECLARE @messageBody NVARCHAR(MAX)
DECLARE @iterator int
DECLARE @Rowcount int
DECLARE @varTableA TABLE(tablecounter int IDENTITY(1,1), DueDate date, Account NVARCHAR(20), Retailer NVARCHAR(20), Interest float)


INSERT INTO @varTableA (DueDate, Account, Retailer, Interest)
SELECT DueDate, Account, Retailer, Interest
FROM TableA 
WHERE DueDate IS NULL OR Account IS NULL OR Retailer IS NULL OR Interest IS NULL

SET @Rowcount = @@ROWCOUNT
SET @iterator = 1


WHILE(@iterator <= @Rowcount)
BEGIN 

SET @messageBody = (SELECT 'No values in following columns' + 'Duedate: ' + ISNULL(Duedate, 'No value') + CHAR(10) + CHAR(13) 
+ 'Account: ' + ISNULL(Account, 'No value') + CHAR(10) + CHAR(13)
+ 'Retailer: ' + ISNULL(Retailer, 'No Value') + CHAR(10) + CHAR(13)
+ 'Interest: ' + ISNULL(Interest, 'No Value') + CHAR(10) + CHAR(13)

                    FROM @varTableA 
                    WHERE @tablecounter = @iterator)

EXEC msdb.dbo.sp_send_dbmail @profilename = [name], @recipients = [reclist], @subject = [sub], @body = @messageBody
SET @iterator = @iterator + 1
END

And to repeat this in either the same or a separate stored Procedure for TableB and Table C. Is there a better way of doing this or do you think I am going about this the right way?

4
  • 4
    Start with worrying about correctness before efficiency. Next, please better explain "I don't want any of them to be null" - do you mean in any row you don't want a single null, or in any column you don't want a single null, or something else? Finally, there is no code you could possibly write that would tell you that DueDate and PaymentDate contain the same kind of information, other than checking that they share the same data type. This is what source control and documentation are for. Oct 14, 2015 at 13:58
  • Sorry, yes, in any row I do not want a single null.
    – Ravi
    Oct 14, 2015 at 13:59
  • and you can uncheck AllowNull in DB design
    – wiretext
    Oct 14, 2015 at 13:59
  • 1
    Still not quite clear on the actual requirements. Do you want to send a single e-mail if any row in a table contains a single NULL? Do you want to send an e-mail per row where any column contains a single NULL? Do you want to send an e-mail per NULL instance (e.g. if row 1 has 2 NULLs and row 2 has 3 NULLs, send 5 e-mails)? Specific details about what you're actually trying to do will be useful. Oct 14, 2015 at 14:32

2 Answers 2

5

For part 1 of your question, I created these three tables in an empty database:

CREATE TABLE dbo.table1(a INT, b INT);
CREATE TABLE dbo.table2(a INT, b INT);
CREATE TABLE dbo.table3(a INT, b INT);

INSERT dbo.table1(a,b) VALUES(1,1),(1,2),(1,NULL);
INSERT dbo.table2(a,b) VALUES(1,1),(NULL,NULL),(1,NULL);
INSERT dbo.table3(a,b) VALUES(1,1),(1,2),(1,2);

Then I wrote this code to extract various bits of NULL information:

CREATE TABLE #x
(
  t           NVARCHAR(512),                     -- table name
  nullrows    INT NOT NULL DEFAULT 0,            -- number of rows with at least one NULL
  nullvalues  INT NOT NULL DEFAULT 0,            -- total NULL values in table
  nulldist    NVARCHAR(MAX) NOT NULL DEFAULT N'' -- how many NULLs in each column
);

INSERT #x(t) VALUES(N'dbo.table1'),(N'dbo.table2'),(N'dbo.table3');

Now some dynamic SQL fun derived from metadata:

DECLARE @sql NVARCHAR(MAX) = N'';


SELECT @sql = @sql + N'
  UPDATE #x SET nullrows = nullrows 
    + (SELECT COUNT(*) FROM ' + t.t 
    + N' WHERE (' + STUFF((SELECT N' OR ' 
    + QUOTENAME(c.name) + N' IS NULL'
  FROM sys.columns AS c
  WHERE OBJECT_ID(t.t) = c.[object_id]
  FOR XML PATH, TYPE).value(N'.[1]',N'nvarchar(max)'),
    1, 4, N'') + N')) WHERE t = N''' + t.t + N''';'
FROM #x AS t;

EXEC sys.sp_executesql @sql;
SET @sql = N'';

SELECT @sql = @sql + N'
  IF EXISTS (SELECT 1 FROM ' + t + N' WHERE ' 
    + QUOTENAME(c.name) + N' IS NULL)
  UPDATE #x SET nullvalues = nullvalues 
    + (SELECT COUNT(*) FROM ' + t + N' WHERE ' 
    + QUOTENAME(c.name) + N' IS NULL),
    nulldistribution = nulldistribution + '''
    + c.name + N':'' + RTRIM((SELECT COUNT(*) FROM ' 
    + t + N' WHERE ' + QUOTENAME(c.name) + N' IS NULL))
    + N'','' WHERE t = N''' + t + N''';'
  FROM #x AS t
  INNER JOIN sys.columns AS c
  ON OBJECT_ID(t.t) = c.[object_id];

EXEC sys.sp_executesql @sql;

SELECT * FROM #x;

Results:

t             nullrows   nullvalues   nulldistribution
dbo.table1    1          1            b:1,
dbo.table2    2          3            a:1,b:2,
dbo.table3    0          0  

I'll leave it as an exercise to the reader how to send an e-mail based on those results - that does not exactly seem like the hard part of this.

For part 2 of your question, as I said in a comment:

There is no code you could possibly write that would tell you that DueDate and PaymentDate contain the same kind of information, other than checking that they share the same data type. This is what source control and documentation are for.

I would argue that this is a different and unanswerable question.

1
  • Hi Aaron, I have updated the question, I thought instead of letting modifications sit in the comments, I should just update the main question. Thank you for this Dynamic SQL answer too.
    – Ravi
    Oct 14, 2015 at 17:39
1

Most operations on a NULL will return NULL. You can use this to collapse all column values into a single values which is NULL if one or more columns is NULL.

create table t1 (id int NOT NULL, a int, b varchar(10), c date);
go

insert t1(id, a, b, c)
values (1, 1, 'a', '2015-10-01'), 
       (2, NULL, 'a', '2015-10-01'), 
       (3, 1, NULL, '2015-10-01'), 
       (4, 1, 'a', NULL);
go

select * from t1;
go

;with x as
(
    select
        ID, convert(varchar(max), a)
          + convert(varchar(max), b) 
          + convert(varchar(max), c) as HasNull
    from t1
)
select
    ID
from x
where x.HasNull is null;

go
drop table t1;

This solution will perform a single table scan on each table once per execution. The output will be a list of the row IDs which have one or more NULL values. I assume you must have a non-nullable primary key column. How would you indicate to the email recipient which row was in trouble otherwise?

See Aaron's answer on how you can construct this dynamically at run time from the system tables.

Your sample email tells the recipient exactly which columns are NULL. I sacrifice this for a more compact query. Once the user opens the data in an editor and navigates to the indicated key it should be obvious which columns contain NULL.

0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.