SQL Server stored procedure optimal query method? - sql-server

Question

I have 5 meta tables that have the same format but depend on 5 other tables. Each Meta table looks like this:
CREATE TABLE [dbo].[SiteMetas]
(
[SiteMetaId] [bigint] IDENTITY(1,1) NOT NULL,
[SiteId] [bigint] NOT NULL,
FOREIGN KEY([SiteId]) REFERENCES [dbo].[Sites] ([SiteId]),
[MetaGroup] [nvarchar] (64) NOT NULL,
[MetaName] [nvarchar] (128) NOT NULL,
[MetaType] [char] NOT NULL DEFAULT 0, -- t, i, r, d, s, b
[MetaBool] [bit] DEFAULT NULL, -- t
[MetaInteger] [bigint] DEFAULT NULL, -- i
[MetaReal] [real] DEFAULT NULL, -- r
[MetaDateTime] [datetime] DEFAULT NULL, -- d
[MetaString] [nvarchar] (MAX) DEFAULT NULL, -- s
[MetaBinary] [varbinary] (MAX) DEFAULT NULL, -- b
[MetaCreated] [datetime] NOT NULL DEFAULT (GETUTCDATE()),
[MetaExpires] [datetime] DEFAULT NULL,
[MetaUpdated] [datetime] DEFAULT NULL,
PRIMARY KEY CLUSTERED ([SiteMetaId] ASC) WITH (IGNORE_DUP_KEY = ON),
UNIQUE NONCLUSTERED ([SiteId] ASC, [MetaGroup] ASC, [MetaName] ASC) WITH (IGNORE_DUP_KEY = ON)
);
This is for Site but there's 4 more. Like Users, ...
And I want to read the Binary meta value from Site. So wrote this stored procedure:
CREATE PROCEDURE [dbo].[GetSiteMetaBinary]
#SiteId AS bigint,
#Group AS nvarchar(64),
#Name AS nvarchar(128)
AS
BEGIN
SELECT TOP 1 [MetaBinary]
FROM [dbo].[SiteMetas]
WHERE [SiteId] = #SiteId
AND [MetaGroup] = #Group
AND [MetaName] = #Name
AND [MetaType] = 'b';
END;
This stored procedure has duplicates for User too... and the rest of the tables. That just replaces Site with User in its body.
But thinking that I have too many of these I wrote this one:
CREATE PROCEDURE [dbo].[GetMeta]
#Set AS nvarchar(64),
#Id AS bigint,
#Group AS nvarchar(64),
#Name AS nvarchar(128),
#Type AS nvarchar(16)
AS
BEGIN
DECLARE #Flag nchar(1);
DECLARE #Sql nvarchar(MAX);
SET #Flag = CASE #Type
WHEN 'Bool' THEN 't'
WHEN 'Integer' THEN 'i'
WHEN 'Real' THEN 'r'
WHEN 'DateTime' THEN 'd'
WHEN 'String' THEN 's'
WHEN 'Binary' THEN 'b'
ELSE NULL
END;
SET #Sql = N'SELECT TOP 1 [Meta' + #Type + N'] FROM [dbo].[' + #Set + N'Metas]' +
N'WHERE [' + #Set + N'Id] = #Id AND [MetaGroup] = #Group AND [MetaName] = #Name AND [MetaType] = #Flag;';
-- SELECT #Sql; -- DEBUG
EXEC sp_executesql #Sql,
N' #Id AS bigint, #Group AS nvarchar(64), #Name AS nvarchar(128), #Flag AS nchar(1)',
#Id, #Group, #Name, #Flag
;
END;
which is a general use stored procedure to read any data typed stored in a column based on input arguments. I use it like this [dbo].[GetMeta] 'Site', 1, 'group', 'name', 'Binary' the difference being that the actual query is dynamically generated so it's not known before hand by SQL Server like the first specialized variant.
Which of the two choices is better from a performance point of view and friendlier to SQL Server's internals? A dedicated one for each table and column data type of a general one that internally builds a query based on fed arguments.
I can use either. I like the last as it does not pollute my stored procedure space. :) The first one is more clear and SQL Server might be able to optimize it better. Not sure...
PS: I'm quite new to SQL Server

Answer 1

Static procedures are usually faster because the SQL engine can cache the compiled SP's execution plan.
However, unless this SP will be called a lot or is time-critical, it probably isn't worth worrying about it because the time savings of only having to maintain one SP will make up for the very small amount of time difference spent waiting for the SP to finish.