SQL Query Delete rows from a table - sql-server

Question

I have a table #temp. The data in #temp are table names in a database. I wish to only show the table names of which the table has data. How can I do this without using dynamic SQL?
My sample data is as below:
create TABLE #temp (Table_Name VARCHAR(50))
insert into #temp values ('#temp1')
,('#temp2')
,('#temp3')
,('#temp4')
create TABLE #temp1 (Col1 int)
insert into #temp1 values (1)
,(3)
,(4)
create TABLE #temp2 (Col1 int)
insert into #temp2 values (7)
,(9)
,(6)
create TABLE #temp3 (Col1 int)
create TABLE #temp4 (Col1 int)
I manually delete the blank tables, How to do this using a query for numerous blank tables?
DELETE FROM #temp
WHERE Table_Name = '#temp3'
or Table_Name = '#temp4'
This is the result I want
select * from #temp
-- It only shows the two table names which are not blank
DROP TABLE #temp
DROP TABLE #temp1
DROP TABLE #temp2
DROP TABLE #temp3
DROP TABLE #temp4
This is my old query for this question:
DECLARE #TABLE_NAME VARCHAR(50), #COMMAND VARCHAR(500), #COUNT INT, #COUNTT INT
DECLARE #CountResults TABLE (CountReturned INT)
create TABLE #TABLE_NAME (TABLE_NAME VARCHAR(50))
SELECT #COUNTT= COUNT(*) FROM #temp
WHILE #COUNTT > 0
BEGIN
SELECT TOP 1 #TABLE_NAME = Table_Name FROM #temp
SET #COMMAND = 'SELECT COUNT(*) FROM ' + #TABLE_NAME
INSERT #CountResults EXEC (#COMMAND)
SET #Count = (SELECT * FROM #CountResults)
BEGIN TRANSACTION
DELETE #CountResults
ROLLBACK TRANSACTION
IF(#Count > 0)
BEGIN
INSERT INTO #TABLE_NAME VALUES (#TABLE_NAME)
END
DELETE FROM #temp WHERE Table_Name = #TABLE_NAME
SELECT #COUNTT= COUNT(*) FROM #temp
END
SELECT * FROM #TABLE_NAME

Answer 1

I don't know of any way to determine whether or not a table is empty without querying that table, which in your case means dynamic SQL. Your comments make it sound like you're okay with this but are looking for a way to do this more concisely than using a loop. Here's a (limited) possibility:
declare #sql nvarchar(max);
select #sql =
-- coalesce() ensures that UNION ALL is inserted before every SELECT but the first.
coalesce(#sql + N' union all ', N'') +
-- Select each table name. Note that SQL Server allows table names that contain
-- single quotes. In this case (or in the case of plain old bad/malicious data in
-- #temp), we need to make sure those characters are enclosed within the string
-- literal we're building.
N'select ''' + replace(table_name, N'''', N'''''') +
-- Use EXISTS to make sure there are one or more records in the table.
N''' where exists (select 1 from ' + quotename(table_name) + N')'
from #temp;
exec sp_executesql #sql;
This will build and execute a query that looks like this:
select '#temp1' where exists (select 1 from [#temp1])
union all
select '#temp2' where exists (select 1 from [#temp2])
union all
select '#temp3' where exists (select 1 from [#temp3])
union all
select '#temp4' where exists (select 1 from [#temp4])
This approach has a few limitations that you should be aware of:
The query will fail if #temp contains any string which is not the name of a table or view. Normally I'd suggest mitigating this by using object_id() or querying INFORMATION_SCHEMA.TABLES, but the fact that you've loaded #temp with the names of other temp tables complicates matters.
The query will also fail if #temp contains a table name that explicitly names the table schema, e.g. dbo.Stuff, because quotename() will render it as [dbo.Stuff] rather than [dbo].[Stuff]. But if you omit quotename(), you run the risk of incorrect and/or damaging behavior if a table_name contains spaces or other problematic characters.
In short, if you just want something for personal use and are okay with making certain assumptions about the data in #temp, then something like the above ought to work. But if you want something that will work correctly and safely under any circumstances, then it's going to take some doing, enough so that even if you could avoid using some kind of a loop, doing so is unlikely to make things any less complicated.

Answer 2

I have a method that does not use dynamic sql. It uses the sysindexes table, which according to Microsoft is subject to change at their whim. So this may not be a good candidate for a production system. But it could be a good place to start. This is also a bit easier if your source table is not a temp table, since temp tables have actual names that do not match the name used to create them.
This script worked for me on SQL Server 2008 r2.
-- drop table #MyTempTable;
Create table #MyTempTable(Table_Name varchar(50));
insert #MyTempTable values ('#MyTempTable2');
insert #MyTempTable values ('#MyTempTable3');
insert #MyTempTable values ('#MyTempTable4');
Create table #MyTempTable2 (Col1 int);
insert #MyTempTable2 values (1);
Create table #MyTempTable4 (Col1 int);
Create table #MyTempTable3 (Col1 int);
SELECT *
FROM #MyTempTable M1
JOIN tempdb.sys.tables T ON T.name LIKE (M1.Table_Name + '%')
JOIN [tempdb].[dbo].[sysindexes] S ON S.id = T.object_id
WHERE S.rowcnt > 0
It's not an ideal solution, but it satisfies your requirements. If you play around with it in your environment, it might give you some insight into a better way to achieve your larger goals. good luck.
EDIT: sysindexes will have one entry per index on the table. Or in the case of my example, for the heap (with no index.) So if your base tables have multiple indexes, you will need to modify the query a bit. Maybe change the JOIN and WHERE clause to a WHERE EXISTS SELECT * FROM [tempdb].[dbo].[sysindexes] S WHERE S.id = T.object_id AND S.rowcnt > 0 Play with it and you should be able to get where you were asking.
EDIT 2: Replacing sys.tables with sysobjects.
SELECT *
FROM #MyTempTable M1
JOIN [tempdb].[dbo].[sysobjects] O ON O.name LIKE (M1.Table_Name + '%')
JOIN [tempdb].[dbo].[sysindexes] S ON S.id = O.id
WHERE S.rowcnt > 0

Answer 3

Based on DeadZone's Query, the following works for non temp tables:
SELECT DISTINCT Table_Name
INTO #TABLE_NAME
FROM #Temp M1
JOIN [dbo].[sysobjects] O ON O.name LIKE (M1.Table_Name + '%')
JOIN [dbo].[sysindexes] S ON S.id = O.id
WHERE S.rowcnt > 0