Why would call to scalar function inside a Table Value Function be slower than outside the TVF?

Question

I am writing a Table Value Function, calling the function takes 10x as long as directly running the code. I traced this to a call to a multi-line scalar function inside the TVF. The call to the scalar function is excessively slow WHEN called from within the TVF. The scalar function takes 3 int parameters and returns a single int result.

What would cause it to be slower from within a TVF?

First, the TVF is basically a sort of pivot table, returning just one row, with 13 columns.

The scalar function is a multiline scalar, it looks for a matched set of keys (ie input is columnA, output is columnB, where input = columnA and active) in one of two tables. Searching table 1, then table 2 and finally table 1 again with a slightly changed where clause.

Usage was originally a cte query with a join clause:

  With cte as ( select count(*) over (partition by c.c1) recs,
                       c.setId, c.c1, c.c2, n.name
                From tbl c
                     Inner join tbl n on n.id = schma.scal(c.c1, c.c2, c.c3)
                Where c.setId=@setId and c.endDt is null
   )

This cte was originally called 11 times, returning from (about half the time) zero to maybe 3 rows (with a where clause like c1=42). So, I thought, hey, it's getting called a lot, since there are really only 4-15 rows in total, and the base tables are ~1,000,000, I'll cut that down to just 15 calls by putting the values into a table variable, and then doing an update. This wasn't really any faster but it did let me prove that it was the call to the scalar function that was slowing things down.

That looked like:

   Insert into @cte(recs, setId, c1, c2)
    select count(*) over (partition by c.c1) recs, c.setId, c.c1, c.c2
    From tbl c
    Where c.setId=@setId and c.endDt is null;

    Update c
     Set nId =schma.scal(c.c1, c.c2, c.c3)
    From @cte c

The @cte had, as I said, from 4 to 15 rows (11 in my most used setId). Putting a return first after, and then before this update is how I concluded that it was the scalar function causing the problem. Basically it was tens of milliseconds, more than a second, another few milliseconds and done.

The query plan from SSMS wasn't useful, and I was thinking about seeing if I could get more details using the plans in the DMVs.

Knowing that TVFs can be used to replace scalars, and that this can improve performance, I gave that a try, and it worked (times in ~100 ms range for the whole query when called inside the original TVF).

I'm still scratching my head as to why it would take longer to call from within the TVF than from outside.

From outside the TVF, I get comparable times using the scalar as I do using the new TFV inside the original TFV.

Answer 1

Scalar functions are called once-per-row, when called as part of a query.

Consider the following example.

1. Create a new, blank database for our tests:

   USE master;
   IF EXISTS (SELECT 1 FROM sys.databases d WHERE d.name = 'mv')
   BEGIN
       ALTER DATABASE mv SET SINGLE_USER WITH ROLLBACK IMMEDIATE;
       DROP DATABASE mv;
   END
   GO
   CREATE DATABASE mv;
   GO
   

2. Create a table, a multi-statement function, and a table-valued-function:

   USE mv;
   GO
   CREATE TABLE dbo.t
   (
       t_id int NOT NULL
           CONSTRAINT PK_t
           PRIMARY KEY CLUSTERED
   );
   GO
   
   CREATE FUNCTION dbo.t_func
   (
       @t_id int
   )
   RETURNS bit
   WITH SCHEMABINDING
   AS
   BEGIN
       DECLARE @r bit;
       IF EXISTS (SELECT 1 FROM dbo.t WHERE t.t_id = @t_id)
           SET @r = 1
       ELSE
           SET @r = 0;
       RETURN @r;
   END
   GO
   
   CREATE FUNCTION dbo.t_tvf
   (
       @min_t_id int
       , @max_t_id int
   )
   RETURNS TABLE 
   WITH SCHEMABINDING
   AS
   RETURN (
       SELECT t_id = t.t_id
           , e = dbo.t_func(dbo.t.t_id)
       FROM dbo.t
       WHERE t.t_id >= @min_t_id
           AND t.t_id <= @max_t_id
   );
   GO
   

3. Insert some sample data into the table:

   INSERT INTO dbo.t (t_id)
   SELECT ROW_NUMBER() OVER (ORDER BY c.id, c.colid)
   FROM sys.syscolumns c;
   GO
   

4. Create a table to store function execution stats, and populate it with a start-row showing execution counts for the multi-statement-function, t_func:

   CREATE TABLE dbo.function_stats
   (
       run_num int NOT NULL
       , object_name sysname NOT NULL
       , execution_count int NULL 
       , CONSTRAINT PK_function_stats
           PRIMARY KEY CLUSTERED (run_num, object_name)
   );
   GO
   INSERT INTO dbo.function_stats (run_num, object_name, execution_count)
   SELECT 1
       , o.name
       , COALESCE(fs.execution_count, 0)
   FROM sys.objects o 
       LEFT JOIN sys.dm_exec_function_stats fs ON fs.object_id = o.object_id
   WHERE o.name = 't_func';
   GO
   

5. Run a query against the TVF:

   SELECT t.*
   FROM dbo.t_tvf(1, 2) t;
   GO
   

6. Capture the execution stats now:

   INSERT INTO dbo.function_stats (run_num, object_name, execution_count)
   SELECT 2
       , o.name
       , COALESCE(fs.execution_count, 0)
   FROM sys.objects o 
       LEFT JOIN sys.dm_exec_function_stats fs ON fs.object_id = o.object_id
   WHERE o.name = 't_func';
   

7. The function stats results:

   SELECT *
   FROM dbo.function_stats fs
   ORDER BY fs.run_num
       , fs.object_name;
   

╔═════════╦═════════════╦═════════════════╗
║ run_num ║ object_name ║ execution_count ║
╠═════════╬═════════════╬═════════════════╣
║       1 ║ t_func      ║               0 ║
║       2 ║ t_func      ║               2 ║
╚═════════╩═════════════╩═════════════════╝

As you can see, the multi-statement-function has execute twice, once per row for the source table accessed by the TVF.

I expect the mutli-statement-function is being called many, many times by the TVF, giving the impression that it is running slowly, whereas in fact it is simply being called many times.