This query runs, but it produces the name of a table as a result, rather than actually selecting from that table.
SELECT T.*
FROM (SELECT tablename
FROM ListOfTables
WHERE id = 0) AS T
where ListOfTables contains id=0, tablename='some_table', I want to return the same result set as if I had written this directly:
SELECT * FROM some_table
Is there a native way to do this in MySQL 5, or do I have to do in in the application?
To do this in MySQL, you need to create a prepared statement which you can only create from a user variable:
SELECT #tn := tablename FROM ListOfTables WHERE id = 0;
SET #qs = CONCAT('SELECT * FROM ', #tn);
PREPARE ps FROM #qs;
EXECUTE ps;
You need to use dynamic SQL to get this result (the below code assumes SQL Server, I can't speak for other RDBMS').
declare #tableName varchar(100)
declare #query varchar(500)
select #tableName = tablename
from ListOfTables
where id = 0
select #query = 'select * from ' + #tableName
exec (#query)
Almost the same as #Shark's answer, except you also quote the name of the table to avoid syntax errors.
-- Using variables just for better readability.
DECLARE #Name NVARCHAR(4000)
DECLARE #Query NVARCHAR(4000)
-- Get the relevant data
SET #Name = QUOTENAME(SELECT tablename FROM ListOfTables WHERE id=0)
-- Build query
SET #Query = 'SELECT * FROM ' + #Schema + '.' + #Name + ''
-- execute it.
EXEC(#Query)
Related
when I do:
SELECT *
FROM SOMETABLE
I get all the columns from SOMETABLE, but I DON'T want the columns which are NULL (for all records). How do I do this?
Reason: this table has 20 columns, 10 of these are set but 10 of them are null for certain queries. And it is time consuming to type the columnnames....
Thanks,
Voodoo
SQL supports the * wildcard which means all columns. There is no wildcard for all columns except the ones you don't want.
Type out the column names. It can't be more work than asking questions on Stack Overflow. Also, copy & paste is your friend.
Another suggestion is to define a view that selects the columns you want, and then subsequently you can select * from the view any time you want.
It's possible to do, but kind of complicated. You can retrieve the list of columns in a table from INFORMATION_SCHEMA.COLUMNS. For each column, you can run a query to see if any non-null row exists. Finally, you can run a query based on the resulting column list.
Here's one way to do that, with a cursor:
declare #table_name varchar(256)
set #table_name = 'Airports'
declare #rc int
declare #query nvarchar(max)
declare #column_list varchar(256)
declare columns cursor local for select column_name
from INFORMATION_SCHEMA.COLUMNS where TABLE_NAME = #table_name
open columns
declare #column_name varchar(256)
fetch next from columns into #column_name
while ##FETCH_STATUS = 0
begin
set #query = 'select #rc = count(*) from ' + #table_name + ' where ' +
#column_name + ' is not null'
exec sp_executesql #query = #query, #params = N'#rc int output',
#rc = #rc output
if #rc > 0
set #column_list = case when #column_list is null then '' else
#column_list + ', ' end + #column_name
fetch next from columns into #column_name
end
close columns
deallocate columns
set #query = 'select ' + #column_list + ' from ' + #table_name
exec sp_executesql #query = #query
This runs on SQL Server. It might be close enough for Sybase. Hopefully, this demonstrates that typing out a column list isn't that bad :-)
New to MySQL, need help in translating the next query to MySQL
If exists (select * from INFORMATION_SCHEMA.COLUMNS where table_name= 'MyTable' and column_name = 'MyColumn')
begin
update MyTable set MyColumn='' where Code=21
end;
Based on the comments posted on your question, here is a code snippet that should answer your need. It works by first checking if the column exists in INFORMATION_SCHEMA, and then dynamically building a SQL query that is prepared, then executed. It the column does not exists, a dummy query is executed instead of the UPDATE. I tested it in this db fiddlde.
SET #dbname = DATABASE();
SET #tablename = "my_table";
SET #columnname = "my_column";
-- check if the column exists
SELECT COUNT(*) INTO #cnt
FROM INFORMATION_SCHEMA.COLUMNS
WHERE
(table_name = #tablename)
AND (table_schema = #dbname)
AND (column_name = #columnname)
;
-- build a dynamic SQL statement
SET #preparedStatement = (SELECT IF(
#cnt > 0,
CONCAT("UPDATE ", #tablename, " SET ", #columnname, " = '' WHERE my_code = 21;"),
"SELECT 1"
));
-- run the statement
PREPARE updateIfExists FROM #preparedStatement;
EXECUTE updateIfExists;
DEALLOCATE PREPARE updateIfExists;
i have the following problem. I inherited a software that uses a database prefix for every customer.
All tables have the same structure and columns. For a data migration to new version i want to union all these tables
and set a customer foreign key instead and get rid of all the subtables. i'm looking for a way to create
a view for this task because i also want to stay backwards compatible for now.
I found this dynamic query which seems to do what i want
but i can't execute on my mysql server. I assume it was written for another sql server.
The table name structure is (about 80 customer tables):
customer1_faxe
customer2_faxe
customer3_faxe
customer4_faxe
...
How would you approach this problem?
DECLARE #SelectClause VARCHAR(100) = 'SELECT *'
,#Query VARCHAR(1000) = ''
SELECT #Query = #Query + #SelectClause + ' FROM ' + TABLE_NAME + ' UNION ALL '
FROM INFORMATION_SCHEMA.TABLES
WHERE TABLE_NAME LIKE '%_faxe'
SELECT #Query = LEFT(#Query, LEN(#Query) - LEN(' UNION ALL '))
EXEC (#Query)
This query is using SQL Server syntax. You need something like this:
declare #SelectClause varchar(8000);
declare #Query varchar(8000);
set #SelectClause = 'SELECT *';
SELECT #Query := group_concat(#SelectClause, ' FROM ', TABLE_NAME SEPARATOR ' UNION ALL ')
FROM INFORMATION_SCHEMA.TABLES
WHERE TABLE_NAME LIKE '%_faxe';
prepare stmt from #Query;
execute stmt;
Note that the group_concat() with separator simplifies the logic.
Using SQL Server 2008, I would like to duplicate one row of a table, without knowing the field names. My key issue: as the table grows and mutates over time, I would like this copy-script to keep working, without me having to write out 30+ ever-changing fields, ugh.
Also at issue, of course, is IDENTITY fields cannot be copied.
My code below does work, but I wonder if there's a more appropriate method than my thrown-together text string SQL statement?
So thank you in advance. Here's my (yes, working) code - I welcome suggestions on improving it.
Todd
alter procedure spEventCopy
#EventID int
as
begin
-- VARS...
declare #SQL varchar(8000)
-- LIST ALL FIELDS (*EXCLUDE* IDENTITY FIELDS).
-- USE [BRACKETS] FOR ANY SILLY FIELD-NAMES WITH SPACES, OR RESERVED WORDS...
select #SQL = coalesce(#SQL + ', ', '') + '[' + column_name + ']'
from INFORMATION_SCHEMA.COLUMNS
where TABLE_NAME = 'EventsTable'
and COLUMNPROPERTY(OBJECT_ID('EventsTable'), COLUMN_NAME, 'IsIdentity') = 0
-- FINISH SQL COPY STATEMENT...
set #SQL = 'insert into EventsTable '
+ ' select ' + #SQL
+ ' from EventsTable '
+ ' where EventID = ' + ltrim(str(#EventID))
-- COPY ROW...
exec(#SQL)
-- REMEMBER NEW ID...
set #EventID = ##IDENTITY
-- (do other stuff here)
-- DONE...
-- JUST FOR KICKS, RETURN THE SQL STATEMENT SO I CAN REVIEW IT IF I WISH...
select EventID = #EventID, SQL = #SQL
end
No, there isn't any magic way to say "SELECT all columns except <foo>" - the way you're doing it is how you'll have to do it (the hack in the other answer aside).
Here is how I would alter your code, with these changes (some are hyperlinked so you can read my opinion about why):
use sys.columns over INFORMATION_SCHEMA.COLUMNS
use nvarchar instead of varchar
use scope_identity instead of ##identity
use sp_executesql instead of exec
use stuff instead of coalesce
use SET NOCOUNT ON
add semi-colons
use the schema prefix
use QUOTENAME since it's safer than '[' + ... + ']'
ALTER PROCEDURE dbo.spEventCopy
#EventID INT
AS
BEGIN
SET NOCOUNT ON;
DECLARE #sql NVARCHAR(MAX) = N'';
SELECT #sql += ',' + QUOTENAME(name)
FROM sys.columns
WHERE [object_id] = OBJECT_ID('dbo.EventsTable')
AND is_identity = 0;
SET #sql = STUFF(#sql, 1, 1, '');
SET #sql = N'INSERT dbo.EventsTable(' + #sql + ')
SELECT ' + #sql + ' FROM dbo.EventsTable
WHERE EventID = ' + CONVERT(VARCHAR(12), #EventID) + ';';
EXEC sp_executesql #sql;
SELECT #EventID = SCOPE_IDENTITY();
-- do stuff with the new row here
SELECT EventID = #EventID, SQL = #SQL;
END
If you know the what your identity column is called (and it won't be the column changing), you could do this:
SELECT * INTO #dummy FROM EventsTable where EventID = #EventID;
ALTER TABLE #dummy
DROP COLUMN MyIdentityColumn
INSERT EventsTable SELECT * FROM #dummy
DROP TABLE #dummy
Since a table can only every have one identity column, specifying that in the query shouldn't limit you too much.
As Aaron Bertrand points out, there are risks associated with this approach. Please read the discussion in the comments below.
I'm trying my first dynamic sql stored procedure. I need to append the exact same records into multiple tables with the same column names. What I have compiles, but when it runs I get 'invalid column name 'TradeDate. The driver sproc is first below, then the sproc containing the dynamic statement. If anyone could help, that'd be great..
ALTER PROCEDURE dbo.StoredProcedure2
AS
DECLARE #tableName varchar(120)
SET #tableName = 'tblDailyATR'
EXEC sprocAddDatesAndSymbolsToAggregatedStudy #tableName
RETURN
ALTER PROCEDURE dbo.sprocAddDatesAndSymbolsToAggregatedStudy
#table varchar(120)
AS
DECLARE #tableName varchar(120)
SET #tableName = #table
EXEC(
'INSERT INTO ' + #tableName + '(Symbol, TradeDate)
SELECT Symbol, TradingDate
FROM (SELECT tblSymbolsMain.Symbol, tblTradingDays.TradingDate
FROM tblSymbolsMain CROSS JOIN tblTradingDays
WHERE (tblTradingDays.TradingDate <= dbo.NextAvailableDataDownloadDate())) AS T1
WHERE (NOT EXISTS (SELECT TradeDate, Symbol
FROM' + #tableName +
' WHERE (TradeDate = T1.TradingDate) AND (Symbol = T1.Symbol)))')
RETURN
You're missing a space after the "FROM" in this line:
FROM' + #tableName +
Should be
FROM ' + #tableName +
Otherwise it's going to try running SELECT FROMTABLE.