Generating DDL/DML scripts for all tables/columns in a database - sql

Question

This is mostly an experiment I've made to learn some SQL Server concepts. Suppose the following scenario:
I have a production database, a development database and a test database;
The development database is more updated than the test database, containing several new tables and columns that were recently developed;
I would like to update the test database as well (with these new tables and columns), but would rather not drop and recreate that DB (it contains useful test data)
The script I wrote below is executed against the "development" database so that it will generate a script with conditions for every column of the database. The script should then be used against the other database to update it, and the conditions should add whatever column or table that the test database doesn't already have:
DECLARE #CURRENT_COLUMN nvarchar(100)
DECLARE #COLUMN_LITERAL nvarchar(100)
DECLARE #CURRENT_DEFAULT nvarchar(20)
DECLARE #CURRENT_DATATYPE nvarchar(100)
DECLARE #CURRENT_SCHEMA nvarchar(100)
DECLARE #SQLA nvarchar(max)
DECLARE #SQLB nvarchar(max)
DECLARE #CURRENT_TABLE nvarchar(100)
DECLARE #COMPUTED smallint
SET #COMPUTED = 0
PRINT '
DECLARE #SQL nvarchar(max)
'
DECLARE CUR_SCHEMA CURSOR FOR
SELECT TABLE_SCHEMA from INFORMATION_SCHEMA.TABLES
OPEN CUR_SCHEMA
FETCH NEXT FROM CUR_SCHEMA
INTO #CURRENT_SCHEMA
WHILE ##FETCH_STATUS = 0
BEGIN
DECLARE CUR_TAB CURSOR FOR
SELECT ist.TABLE_NAME from INFORMATION_SCHEMA.TABLES ist
WHERE ist.TABLE_SCHEMA = #CURRENT_SCHEMA
AND EXISTS (
SELECT TOP 1 name
FROM sys.tables
where name = ist.TABLE_NAME)
ORDER BY ist.TABLE_NAME
OPEN CUR_TAB
FETCH NEXT FROM CUR_TAB
INTO #CURRENT_TABLE
WHILE ##FETCH_STATUS = 0
BEGIN
PRINT '
IF OBJECT_ID('''+#CURRENT_TABLE+''') IS NULL
BEGIN
SET #SQL = ''
CREATE TABLE [' + #CURRENT_TABLE +'] (placeholder bit)''
EXEC sp_executesql #SQL
END
'
DECLARE CUR CURSOR FOR
SELECT COLUMN_NAME, DATA_TYPE
FROM INFORMATION_SCHEMA.COLUMNS
WHERE TABLE_NAME = #CURRENT_TABLE
AND TABLE_SCHEMA = #CURRENT_SCHEMA
ORDER BY ORDINAL_POSITION asc
OPEN CUR
FETCH NEXT FROM CUR
INTO #CURRENT_COLUMN, #CURRENT_DATATYPE
SET #COLUMN_LITERAL = '[' + #CURRENT_COLUMN + ']'
WHILE ##FETCH_STATUS = 0
BEGIN
SET #SQLB = ''
SET #COMPUTED = 0
/* Check if column is computed */
IF (SELECT is_computed FROM sys.columns
WHERE object_id = OBJECT_ID(#CURRENT_TABLE)
AND name = #CURRENT_COLUMN) = 1
BEGIN
SET #SQLB = #SQLB + 'IF NOT EXISTS(SELECT TOP 1 sc.name FROM sys.columns sc
INNER JOIN sys.tables st ON st.object_id = sc.object_id
INNER JOIN sys.schemas ss ON ss.schema_id = st.schema_id
WHERE sc.Name = ''' + #CURRENT_COLUMN + '''
AND st.Object_ID = OBJECT_ID('''+ #CURRENT_TABLE+ ''')
AND ss.name = ''' + #CURRENT_SCHEMA + ''')
BEGIN
ALTER TABLE ' + #CURRENT_SCHEMA + '.[' + #CURRENT_TABLE + ']
ADD ' + #CURRENT_COLUMN + ' AS ' +
(SELECT definition FROM sys.computed_columns
WHERE object_id = OBJECT_ID(#CURRENT_TABLE)
AND name = #CURRENT_COLUMN)
SET #COMPUTED = 1
END
/* Check for identity */
IF (SELECT is_identity FROM sys.columns WHERE object_id = OBJECT_ID(#CURRENT_TABLE)
AND name = #CURRENT_COLUMN) = 1
BEGIN
SET #SQLB = #SQLB + ' IDENTITY (' +
CAST((SELECT IDENT_SEED(#CURRENT_SCHEMA + '.[' + #CURRENT_TABLE + ']')) AS VARCHAR(4)) + ',' +
CAST((SELECT IDENT_INCR(#CURRENT_SCHEMA + '.[' + #CURRENT_TABLE + ']')) AS VARCHAR(4)) + ')'
END
/* Check if NULL is allowed */
IF (SELECT sc.is_nullable from sys.columns sc
INNER JOIN sys.tables st ON st.object_id = sc.object_id
INNER JOIN sys.schemas ss ON ss.schema_id = st.schema_id
INNER JOIN sys.types sp ON sp.system_type_id = sc.system_type_id
WHERE st.name = #CURRENT_TABLE
AND sc.name = #CURRENT_COLUMN
AND ss.name = #CURRENT_SCHEMA
AND sp.name = #CURRENT_DATATYPE
) = 0
BEGIN
SET #SQLB = #SQLB + ' NOT NULL'
END
ELSE SET #SQLB = #SQLB + ' NULL'
/* Check for defaults */
IF (SELECT COLUMN_DEFAULT FROM INFORMATION_SCHEMA.COLUMNS
WHERE COLUMN_NAME = #CURRENT_COLUMN
AND TABLE_SCHEMA = #CURRENT_SCHEMA
AND TABLE_NAME = #CURRENT_TABLE) IS NOT NULL
BEGIN
SET #CURRENT_DEFAULT = ' DEFAULT ' + (SELECT COLUMN_DEFAULT FROM INFORMATION_SCHEMA.COLUMNS
WHERE COLUMN_NAME = #CURRENT_COLUMN
AND TABLE_SCHEMA = #CURRENT_SCHEMA
AND TABLE_NAME = #CURRENT_TABLE)
END
ELSE SET #CURRENT_DEFAULT = ''
IF #CURRENT_DATATYPE in ('date','datetime2','datetime','time',
'smalldatetime','datetimeoffset','text','ntext',
'varchar','char','nchar','nvarchar')
BEGIN
/* Check for date related data types */
IF #CURRENT_DATATYPE in ('date','datetime2','datetime','time',
'smalldatetime','datetimeoffset')
BEGIN
SET #SQLA = '
IF NOT EXISTS(SELECT TOP 1 sc.name FROM sys.columns sc
INNER JOIN sys.tables st ON st.object_id = sc.object_id
INNER JOIN sys.schemas ss ON ss.schema_id = st.schema_id
WHERE sc.Name = ''' + #CURRENT_COLUMN + '''
AND st.Object_ID = OBJECT_ID('''+ #CURRENT_TABLE+ ''')
AND ss.name = ''' + #CURRENT_SCHEMA + ''')
BEGIN
ALTER TABLE ' + #CURRENT_SCHEMA + '.['+ #CURRENT_TABLE + ']
ADD '+#COLUMN_LITERAL+'' + ' ' + ''+#CURRENT_DATATYPE+' '+#CURRENT_DEFAULT
END
/* Check for MAX column length */
IF (SELECT sc.max_length FROM sys.columns sc
INNER JOIN sys.tables st ON st.object_id = sc.object_id
INNER JOIN sys.schemas ss ON ss.schema_id = st.schema_id
INNER JOIN sys.types sp ON sp.system_type_id = sc.system_type_id
WHERE st.name = #CURRENT_TABLE
AND sc.name = #CURRENT_COLUMN
AND ss.name = #CURRENT_SCHEMA
AND sp.name = #CURRENT_DATATYPE) = -1
BEGIN
SET #SQLA = '
IF NOT EXISTS(SELECT TOP 1 sc.name FROM sys.columns sc
INNER JOIN sys.tables st ON st.object_id = sc.object_id
INNER JOIN sys.schemas ss ON ss.schema_id = st.schema_id
WHERE sc.Name = ''' + #CURRENT_COLUMN + '''
AND st.Object_ID = OBJECT_ID('''+ #CURRENT_TABLE+ ''')
AND ss.name = ''' + #CURRENT_SCHEMA + ''')
BEGIN
ALTER TABLE ' + #CURRENT_SCHEMA + '.['+ #CURRENT_TABLE + ']
ADD '+#COLUMN_LITERAL+'' + ' ' + ''+#CURRENT_DATATYPE+'(MAX)'+' ' + #CURRENT_DEFAULT
END
/* Check for string data types */
ELSE IF #CURRENT_DATATYPE in ('varchar','char','nchar','nvarchar')
BEGIN
SET #SQLA = '
IF NOT EXISTS(SELECT TOP 1 sc.name FROM sys.columns sc
INNER JOIN sys.tables st ON st.object_id = sc.object_id
INNER JOIN sys.schemas ss ON ss.schema_id = st.schema_id
WHERE sc.Name = ''' + #CURRENT_COLUMN + '''
AND st.Object_ID = OBJECT_ID('''+ #CURRENT_TABLE+ ''')
AND ss.name = ''' + #CURRENT_SCHEMA + ''')
BEGIN
ALTER TABLE ' + #CURRENT_SCHEMA + '.[' + #CURRENT_TABLE + ']
ADD '+#COLUMN_LITERAL+'' + ' ' + ''+#CURRENT_DATATYPE+''
+ '(' +
CAST(
( SELECT
CASE WHEN #CURRENT_DATATYPE IN ('nchar', 'nvarchar') THEN MAX(sc.max_length)/2
ELSE MAX(sc.max_length) END AS 'max_length' FROM sys.columns sc
INNER JOIN sys.tables st ON st.object_id = sc.object_id
INNER JOIN sys.schemas ss ON ss.schema_id = st.schema_id
INNER JOIN sys.types sp ON sp.system_type_id = sc.system_type_id
WHERE st.name = #CURRENT_TABLE
AND sc.name = #CURRENT_COLUMN
AND ss.name = #CURRENT_SCHEMA
AND sp.name = #CURRENT_DATATYPE
)
AS VARCHAR(10)) +')'+#CURRENT_DEFAULT
END
/* Check for text and ntext types (no column width) */
ELSE IF #CURRENT_DATATYPE in ('text','ntext')
BEGIN
SET #SQLA = '
IF NOT EXISTS(SELECT TOP 1 sc.name FROM sys.columns sc
INNER JOIN sys.tables st ON st.object_id = sc.object_id
INNER JOIN sys.schemas ss ON ss.schema_id = st.schema_id
WHERE sc.Name = ''' + #CURRENT_COLUMN + '''
AND st.Object_ID = OBJECT_ID('''+ #CURRENT_TABLE+ ''')
AND ss.name = ''' + #CURRENT_SCHEMA + ''')
BEGIN
ALTER TABLE ' + #CURRENT_SCHEMA + '.[' + #CURRENT_TABLE + ']
ADD '+#COLUMN_LITERAL+'' + ' ' + ''+#CURRENT_DATATYPE+' '+#CURRENT_DEFAULT
END
END
ELSE
/* Check for decimal and numeric types */
IF #CURRENT_DATATYPE in ('decimal','numeric')
BEGIN
SET #SQLA = '
IF NOT EXISTS(SELECT TOP 1 sc.name FROM sys.columns sc
INNER JOIN sys.tables st ON st.object_id = sc.object_id
INNER JOIN sys.schemas ss ON ss.schema_id = st.schema_id
WHERE sc.Name = ''' + #CURRENT_COLUMN + '''
AND st.Object_ID = OBJECT_ID('''+ #CURRENT_TABLE+ ''')
AND ss.name = ''' + #CURRENT_SCHEMA + ''')
BEGIN
ALTER TABLE ' + #CURRENT_SCHEMA + '.[' + #CURRENT_TABLE + ']
ADD '+#COLUMN_LITERAL+'' + ' ' + ''+#CURRENT_DATATYPE+''+'(' + CAST( (SELECT MIN(NUMERIC_PRECISION) FROM INFORMATION_SCHEMA.COLUMNS
WHERE TABLE_NAME = #CURRENT_TABLE
AND COLUMN_NAME = #CURRENT_COLUMN
AND TABLE_SCHEMA = #CURRENT_SCHEMA
AND DATA_TYPE = #CURRENT_DATATYPE
) AS VARCHAR(10)) + ',' +
CAST( (SELECT MIN(NUMERIC_SCALE) FROM INFORMATION_SCHEMA.COLUMNS
WHERE TABLE_NAME = #CURRENT_TABLE
AND COLUMN_NAME = #CURRENT_COLUMN
AND DATA_TYPE = #CURRENT_DATATYPE
) AS VARCHAR(10)) + ')'+ #CURRENT_DEFAULT
END
ELSE
BEGIN
SET #SQLA = '
IF NOT EXISTS(SELECT TOP 1 sc.name FROM sys.columns sc
INNER JOIN sys.tables st ON st.object_id = sc.object_id
INNER JOIN sys.schemas ss ON ss.schema_id = st.schema_id
WHERE sc.Name = ''' + #CURRENT_COLUMN + '''
AND st.Object_ID = OBJECT_ID('''+ #CURRENT_TABLE+ ''')
AND ss.name = ''' + #CURRENT_SCHEMA + ''')
BEGIN
ALTER TABLE ' + #CURRENT_SCHEMA + '.[' + #CURRENT_TABLE + ']
ADD '+#COLUMN_LITERAL+'' + ' ' + ''+#CURRENT_DATATYPE+''+#CURRENT_DEFAULT
END
IF #COMPUTED = 0
BEGIN
PRINT #SQLA + #SQLB + '
END
'
END
FETCH NEXT FROM CUR
INTO #CURRENT_COLUMN, #CURRENT_DATATYPE
SET #COLUMN_LITERAL = '[' + #CURRENT_COLUMN + ']'
END
CLOSE CUR;
DEALLOCATE CUR;
PRINT '
IF EXISTS
(SELECT TOP 1 sc.name FROM sys.columns sc
INNER JOIN sys.tables st ON st.object_id = sc.object_id
INNER JOIN sys.schemas ss ON ss.schema_id = st.schema_id
WHERE sc.Name = ''placeholder''
AND st.Object_ID = OBJECT_ID('''+ #CURRENT_TABLE+ ''')
AND ss.name = ''' + #CURRENT_SCHEMA + ''')
BEGIN
ALTER TABLE '+#CURRENT_SCHEMA+'.['+#CURRENT_TABLE+'] DROP COLUMN [placeholder]
END
'
FETCH NEXT FROM CUR_TAB
INTO #CURRENT_TABLE
END
CLOSE CUR_TAB
DEALLOCATE CUR_TAB
END
CLOSE CUR_SCHEMA
DEALLOCATE CUR_SCHEMA
Questions:
Could my use of a "placeholder" column be avoided? (I added it because I couldn't create empty columns in case they didn't exist).
Is the use of three cursors accepted? I believe this could be simplified, perhaps with temp tables, or table variables.
Is my approach to catch exceptional formatting cases (like the numeric data type definition or the maximum column character length) cohesive?
Is the script fully correct? I intensively tested it against a copy of a real database and also tested the script it generated against a blank database, and it seems to have produced the expected results.
Is the number of variables I used excessive? Is any of my variables irrelevant?
Is the use of both INFORMATION_SCHEMA and system tables acceptable? (I used INFORMATION_SCHEMA some times to avoid excess of table joining).
Am I using cursors correctly?
Would you suggest a different approach to some section of my script?
Thanks, and sorry for asking so many questions. Answer just one or some if you don't wish to answer all of them!
** NOTES **
I wrote this for an SQL Server 2008 database, but you could point out alternatives for newer versions to help improve my knowledge
I know this script doesn't replicate stored procedures, triggers, and other things, but that could be scripted automatically with SSMS afterwards so I included just column properties on the script.

Answer 1

So I would propose the usage of the Sql Server Data Tools and database project. It allows you to import a schema of existing database or create a new database project. You can link it to the TFS or Git, your preference for source control. In this case you do not have to have a separate SQL script to generate schema. If you decide to move to the schema development from SSDT then the codebase will always have a latest known schema. You can then generate scripts for deployment of the new codebase and SSDT or Visual Studio will figure the SQL out, I would strongly recommend checking those scripts before deployment to prod.
It is also possible to publish changes to target databases from this tool.
SSDT