We have a SQL query that pulls a large number of fields from many tables/views from a database. We need to put a spec together for integration with a 3rd party, what is the quickest way to compile the data types of the result set?
Clarifications:
- There are 25+ tables/views involved, so functions at the table level will still be cumbersome.
- All work is currently being done in Microsoft SQL Server Management Studio.
You can run the query with SET FMTONLY ON, but that might not help you to easily determine the data types returned, since you're working in management studio only. If it was me, I think I'd create a view temporarily with the same body as the stored procedure (you may have to declare variables for any parameters). You can then look at the columns returned by the view with the INFORMATION_SCHEMA queries already discussed.
You might use few quick SQL statements to view result column type, by using temp table.
Temp tables is a little better then a view, as they are connection-local scope and will be cleared once disconnect.
All you need is inject few keyword as follow
SELECT
TOP 0 -- to speed up without access data
your,original,columns
INTO #T -- temp table magic
FROM originalTablesJoins
Order by anything
exec tempdb.sys.sp_columns #T
drop table #T
or;
SELECT TOP 0 * FROM (
select your,original,columns from originalTablesJoins -- remove order by if any
) x
exec tempdb.sys.sp_columns #T
drop table #T
Note: inspired by View schema of resultset in SQL Server Management Studio
If you're using SQL Server, metadata from various tables is available in the information_schema table. For instance, to get column metadata for table Foo, issue this query:
SELECT * FROM information_schema.columns WHERE table_name = 'Foo'
If you were using C# you could access it right from the field in the DataRow object:
Type columnNameType = row["columnName"].GetType();
And for an additional alternative you can use
sp_help 'Table_Name'
EDIT: Also, sp_help can be used for any object ( i.e. it would indicate the return type of stored procedure input and output variables)
Here's one, if you can just SELECT ... INTO #Temp (remember #Temp is scoped to the session at worst) with some stolen from https://stackoverflow.com/a/14328779/162273:
SELECT
c.name AS UsefulRawName,
',' + c.name + ' ' + UPPER(t.name) +
CASE
WHEN t.name IN ('char', 'nchar', 'varchar', 'nvarchar') THEN '(' + CAST(c.max_length AS VARCHAR(3) ) + ')'
WHEN t.name IN ('decimal', 'numeric') THEN '(' + CAST(c.[precision] AS VARCHAR(3) ) + ', ' + CAST(c.[scale] AS VARCHAR(3) ) + ')'
ELSE '' END + CASE WHEN c.Is_Nullable = 1 THEN ' NULL' ELSE ' NOT NULL' END AS SQLColumnType,
'public ' +
CASE
WHEN t.name IN ('varchar', 'nvarchar', 'text', 'ntext', 'char', 'nchar', 'xml', 'sysname') THEN 'string'
WHEN t.name IN ('binary', 'varbinary', 'image') THEN 'byte[]' + CASE WHEN c.Is_Nullable = 1 THEN '?' ELSE '' END
WHEN t.name IN ('uniqueidentifier') THEN 'Guid' + CASE WHEN c.Is_Nullable = 1 THEN '?' ELSE '' END
WHEN t.name IN ('datetimeoffset') THEN 'DateTimeOffset' + CASE WHEN c.Is_Nullable = 1 THEN '?' ELSE '' END
WHEN t.name IN ('date', 'time', 'datetime2', 'smalldatetime', 'datetime') THEN 'DateTime' + CASE WHEN c.Is_Nullable = 1 THEN '?' ELSE '' END
WHEN t.name IN ('numeric', 'decimal', 'real', 'money', 'smallmoney') THEN 'decimal' + CASE WHEN c.Is_Nullable = 1 THEN '?' ELSE '' END
WHEN t.name IN ('float') THEN 'float' + CASE WHEN c.Is_Nullable = 1 THEN '?' ELSE '' END
WHEN t.name IN ('tinyint', 'smallint') THEN 'short' + CASE WHEN c.Is_Nullable = 1 THEN '?' ELSE '' END
WHEN t.name IN ('int') THEN 'int' + CASE WHEN c.Is_Nullable = 1 THEN '?' ELSE '' END
WHEN t.name IN ('bit') THEN 'bool' + CASE WHEN c.Is_Nullable = 1 THEN '?' ELSE '' END
WHEN t.name IN ('bigint') THEN 'long' + CASE WHEN c.Is_Nullable = 1 THEN '?' ELSE '' END
WHEN t.name IN ('timestamp') THEN 'ulong'
ELSE 'object' END + ' ' + c.name + ' { get; set; }' AS CSColumnType,
c.name + ' = ' +
CASE
WHEN t.name IN ('varchar', 'nvarchar', 'text', 'ntext', 'char', 'nchar', 'xml', 'sysname') THEN 'reader["' + c.name + '"] as string,'
WHEN t.name IN ('binary', 'varbinary', 'image') THEN CASE
WHEN c.Is_Nullable = 1 THEN 'reader["' + c.name + '"] as byte[]?,'
ELSE '(byte[])reader["' + c.name + '"],' END
WHEN t.name IN ('uniqueidentifier') THEN CASE
WHEN c.Is_Nullable = 1 THEN 'reader["' + c.name + '"] as Guid?,'
ELSE '(Guid)reader["' + c.name + '"],' END
WHEN t.name IN ('datetimeoffset') THEN CASE
WHEN c.Is_Nullable = 1 THEN 'reader["' + c.name + '"] as DateTimeOffset?,'
ELSE '(DateTimeOffset)reader["' + c.name + '"],' END
WHEN t.name IN ('date', 'time', 'datetime2', 'smalldatetime', 'datetime') THEN CASE
WHEN c.Is_Nullable = 1 THEN 'reader["' + c.name + '"] as DateTime?,'
ELSE '(DateTime)reader["' + c.name + '"],' END
WHEN t.name IN ('numeric', 'decimal', 'real', 'money', 'smallmoney') THEN CASE
WHEN c.Is_Nullable = 1 THEN 'reader["' + c.name + '"] as decimal?,'
ELSE '(decimal)reader["' + c.name + '"],' END
WHEN t.name IN ('float') THEN CASE
WHEN c.Is_Nullable = 1 THEN 'reader["' + c.name + '"] as float?,'
ELSE '(float)reader["' + c.name + '"],' END
WHEN t.name IN ('tinyint', 'smallint') THEN CASE
WHEN c.Is_Nullable = 1 THEN 'reader["' + c.name + '"] as short?,'
ELSE '(short)reader["' + c.name + '"],' END
WHEN t.name IN ('int') THEN CASE
WHEN c.Is_Nullable = 1 THEN 'reader["' + c.name + '"] as int?,'
ELSE '(int)reader["' + c.name + '"],' END
WHEN t.name IN ('bit') THEN CASE
WHEN c.Is_Nullable = 1 THEN 'reader["' + c.name + '"] as bool?,'
ELSE '(bool)reader["' + c.name + '"],' END
WHEN t.name IN ('bigint') THEN CASE
WHEN c.Is_Nullable = 1 THEN 'reader["' + c.name + '"] as long?,'
ELSE '(long)reader["' + c.name + '"],' END
WHEN t.name IN ('timestamp') THEN '(ulong)reader["' + c.name + '"],'
ELSE 'reader["' + c.name + '"] == DBNull.Value ? null : reader["' + c.name + '"],' END AS ReaderStatements
FROM tempDb.sys.columns c
INNER JOIN sys.types AS t ON c.system_type_id = t.system_type_id AND t.system_type_id = t.user_type_id
WHERE [object_id] = OBJECT_ID('tempdb..#Temp')
ORDER BY column_id
A more motivated person could do some smarter mapping of the types to .Net equivalents automatically but this isn't so bad and keeps the snippet short.
Converting these types manually was enough of a pain in the butt that extending the script was worthwhile. Not sure my conversions are all correct.
Note that result set metadata is different from table metadata, because SQL queries can include expressions, whose data types may not be the same as those in the tables they query.
Many SQL query interfaces give you some function to retrieve information about the result set metadata (datatypes, etc.).
The specific API functions you need to use depend on what language and query interface you're using. You haven't stated this.
For instance, if you're using ODBC, the SQLDescribeCol() function can give you information about the result set metadata.
来源：https://stackoverflow.com/questions/440123/how-to-determine-the-datatypes-of-the-results-of-a-sql