Nov 23, 2006
/*Subject: How to build a procedure that returns differentnumbers of columns as a result based on a parameter.You can copy/paste this whole post in SQL Query Analyzeror Management Studio and run it once you've made surethere is no harmful code.Currently we have several stored procedures which finalresult is a select with several joins that returns manycolumns (150 in one case, maybe around 50 the average).We have analyzed our application and found out that mostof the time not all the columns are used. We haveidentified 3 different sets of columns needed indifferent parts of the application.Let's identify and name these sets as:1- simple set, return the employee list for example2- common set, return the employee information (whichinclude the simple set)3- extended set, return the employee information (whichinlude the common set which itself includes the simpleset) + additional information from other tables, maybeeven some SUM aggregates and so on (I don't know forexample, how much sales the employee did so far).So the bigger sets contain the smaller ones. Please keepreading all the way to the bottom to better understandtechnically what we are trying.Here is a code sample of how our current procedureswork. Please note that the passing parameter we can eitherpass a Unique Identifier (PK) to retrieve a single record,or if we pass for example -1 or NULL we retrieve all theemployee records.*/create table a ( apk int primary key, af1 int, af2 int, af3 int, af4int, af5 int, af6 int)create table b ( bpk int primary key, bf1 int, bf2 int, bf3 int, bf4int, bf5 int, bf6 int)create table c ( cpk int primary key, cf1 int, cf2 int, cf3 int, cf4int, cf5 int, cf6 int)create table d ( dpk int primary key, df1 int, df2 int, df3 int, df4int, df5 int, df6 int)insert a values (1,1111,1112,1113,1114,1115,1116)insert a values (2,1211,1212,1213,1214,1215,1216)insert a values (3,1311,1312,1313,1314,1315,1316)insert a values (4,1411,1412,1413,1431,1415,1416)insert a values (5,1511,1512,1513,1514,1515,1516)insert a values (6,1611,1612,1613,1614,1615,1616)insert b values (1,2111,2112,2113,2114,2115,2116)insert b values (2,2211,2212,2213,2214,2215,2216)insert b values (3,2311,2312,2313,2314,2315,2316)insert b values (4,2411,2412,2413,2431,2415,2416)insert b values (5,2511,2512,2513,2514,2515,2516)insert b values (6,2611,2612,2613,2614,2615,2616)insert c values (1,3111,3112,3113,3114,3115,3116)insert c values (2,3211,3212,3213,3214,3215,3216)insert c values (3,3311,3312,3313,3314,3315,3316)insert c values (4,3411,3412,3413,3431,3415,3416)insert c values (5,3511,3512,3513,3514,3515,3516)insert c values (6,3611,3612,3613,3614,3615,3616)insert d values (1,4111,4112,4113,4114,4115,4116)insert d values (2,4211,4212,4213,4214,4215,4216)insert d values (3,4311,4312,4313,4314,4315,4316)insert d values (4,4411,4412,4413,4431,4415,4416)insert d values (5,4511,4512,4513,4514,4515,4516)insert d values (6,4611,4612,4613,4614,4615,4616)gocreate procedure original_proc @pk int asif @pk = -1set @pk = nullselecta.af1, a.af2, a.af3, a.af4, b.bf1, b.bf2, b.bf3, b.bf4, c.cf1, c.cf2,c.cf3, c.cf4, d.df1, d.df2, d.df3, d.df4fromajoin b on a.apk = b.bpkjoin c on b.bpk = c.cpkjoin d on c.cpk = d.dpkwherea.apk = ISNULL(@pk, a.apk)goexec original_proc 1go/*Currently the above SP is a single SP that is basicallyreturning ALL possible needed data. However most of thetime we might need to call and retrieve a simple employeelist.So we thought about modifying the stored procedure byadding an extra parameter that will indicate which setof columns to return.For modifying the stored procedure in order to get avariable name of columns returned and avoidingrepeating code, we built 4 objects: the storedprocedure being called, one table function and 2 views.One table function so that we are able to pass a parameter.The views since they do not accept parameters they arealways joined at least with the inline table function.The stored procedure generates in its body a dynamicSQL statement, where it queries the table function andthe views, depending which set is required. Here is acode sample of our current design (you need to run theprevious code in order for this to work).*/create function _1_set(@pk int)returns tableas return(select a.apk, a.af1, a.af2, a.af3, a.af4, b.bf1, b.bf2from ajoin b on a.apk = b.bpkwhere a.apk = ISNULL(@pk, a.apk))gocreate view _2_set asselect b.bpk, b.bf3, b.bf4, c.cf1, c.cf2from bjoin c on b.bpk = c.cpkgocreate view _3_set asselect c.cpk, c.cf3, c.cf4, d.df1, d.df2, d.df3, d.df4from cjoin d on c.cpk = d.dpkgocreate procedure new_proc @pk int, @set int asdeclare @sql nvarchar(4000)if @pk = -1set @pk = nullset @sql = 'select * from _1_set(@pk) fs 'if @set 1set @sql = @sql + 'join _2_set ss on fs.apk = ss.bpk 'if @set 2set @sql = @sql + 'join _3_set ts on ss.bpk = ts.cpk 'exec sp_executesql @sql, N'@pk int', @pkgoexec new_proc 1, 3go/*For executing the new procedure, we pass parameter 1for the smaller set, 2 for the medium size set or 3for the complete set.For example when we want to retrieve the common setwe pass the Unique Identifier of the employee to theSP and then we pass the type of set we want to useas the second parameter (1 for simple set, 2 forcommon set and 3 for extended set).The SP has the IF and dynamic SQL to add more JOINs.We would like to know what you think of this approachand if you know a simpler way of doing it.For cleaning up the test objects run the following code.*/drop procedure original_procdrop procedure new_procdrop function _1_setdrop view _2_setdrop view _3_setdrop table adrop table bdrop table cdrop table dAs always I would appreciate any feedback, opinion,comments, ideas and suggestions.Thank you
View 9 Replies
View Related
