Tag: create temp table sql

One Metric Week

This went on a bit longer than I thought it would, but to someone who tries to blog five days a week, that’s a good thing.

While this isn’t the most glamorous subject in the world, it seems it’s still a necessary one to write about. I don’t quite know how to feel about that, as I’ve read and written about it many times over the past 10 or so years.

• SARGability Week: What Is It, Anyway?
• SARGability Week: Using Computed Columns To Fix Non-SARGable Queries
• SARGability Week: Using Temp Tables To Fix Non-SARGable Queries
• SARGability Week: Using Index Column Order To Fix Non-SARGable Queries
• SARGability Week: Using Dynamic SQL To Fix Non-SARGable Queries
• SARGability Week: Max Data Type Columns And Predicates Aren’t SARGable
• SARGability Week: Rewriting Scalar User Defined Functions To Make Them SARGable
• SARGability Week: Using Indexed Views To Make Predicates SARGable
• SARGability Week: Why Implicit Conversions Aren’t SARGable

Thanks for reading!

Going Further

If this is the kind of SQL Server stuff you love learning about, you’ll love my training. I’m offering a 75% discount to my blog readers if you click from here. I’m also available for consulting if you just don’t have time for that and need to solve performance problems quickly.

President Precedence

Data types are one of those core things you need to get right. Whether it’s matching types between join columns, or between parameters and predicates, not doing so can really squash application performance in quite similar ways to writing non-SARGable predicates.

That’s because — wait for it — a lot of the stuff we’ve talked about over the last week that can happen with poorly written predicates can happen with poorly matched data types, too.

The main thing to remember, is that aside from max datatypes, what matters most in a situation with implicit conversion is that it doesn’t take place on a column. If it happens on a parameter or variable it’s far less of an issue, but it can still cause oddities in query plans, and with cardinality estimation.

Speak And Spell

Most of the problems I see with implicit conversion is with other datatypes being compared to nchar/nvarchar types, but not always.

For example, this query results in an index seek despite a variable declared as nvarchar(11).

DECLARE 
    @i nvarchar(11) = N'22656';

SELECT
   u.*
FROM dbo.Users AS u
WHERE u.AccountId = @i;
GO 

Note that the convert_implicit function is applied to the variable, and not the AccountId column.

Cahn

To show an example of when implicit conversions act the same way was non-SARGable predicates, let’s make a crappy copy of some columns from the Users table.

DROP TABLE IF EXISTS
    dbo.UsersBad;

SELECT u.Id,
       ISNULL
       (
           CONVERT
           (
               varchar(40), 
               u.DisplayName
           ), 
           ''
       ) AS DisplayName,
       ISNULL
       (
           CONVERT
           (
               nvarchar(11), 
               u.Reputation
           ), 
           ''
       ) AS Reputation
INTO dbo.UsersBad
FROM dbo.Users AS u;

ALTER TABLE dbo.UsersBad 
    ADD CONSTRAINT PK_UsersBad_Id 
        PRIMARY KEY CLUSTERED (Id);

CREATE INDEX ur ON dbo.UsersBad(Reputation);
CREATE INDEX ud ON dbo.UsersBad(DisplayName);

Here, we’re converting DisplayName from nvarchar, and Reputation from an int. We’re also creating some indexes that will be rendered nearly useless by implicit conversions.

CREATE OR ALTER PROCEDURE 
    dbo.BadUsersQuery
(
    @DisplayName nvarchar(40),
    @Reputation  int
)
AS
BEGIN

    SELECT 
        u.DisplayName
    FROM dbo.UsersBad AS u
    WHERE u.DisplayName = @DisplayName;

    SELECT
        Reputation = 
            MAX(u.Reputation)
    FROM dbo.UsersBad AS u
    WHERE u.Reputation = @Reputation;

END;
GO 

EXEC dbo.BadUsersQuery 
    @DisplayName = N'Eggs McLaren',
    @Reputation = 1787;

Imaging

For both of these, the convert_implicit winds up on the column rather than on the parameter.

And that’s what makes for the type of implicit conversion that causes most performance problems I see.

This is a relatively small table, so the hit isn’t too bad, but comparatively it’s much worse, like in all the other examples of SARGability we’ve seen lately.

Aware

I think most people who do performance tuning regularly are quite aware of this problem. There’s even a query plan warning about it, for those who don’t.

It’s definitely a good problem to solve, but it often leads to finding other problems. For example, we get the seek predicate warning regardless of if there’s an index we can seek to. If I drop all the indexes off of the UsersBad table and run a query like this, we still get a warning:

DECLARE
    @DisplayName nvarchar(40);

SELECT 
    u.DisplayName
FROM dbo.UsersBad AS u
WHERE u.DisplayName = @DisplayName;

It’s also worth noting that getting rid of the implicit conversion — much like fixing other non-SARGable predicates — may reveal missing index requests that weren’t there before.

DECLARE
    @DisplayName nvarchar(40) = N'Eggs McLaren';

SELECT 
    u.*
FROM dbo.UsersBad AS u
WHERE u.DisplayName = @DisplayName;
GO 

DECLARE
    @DisplayName varchar(40) = 'Eggs McLaren';

SELECT 
    u.*
FROM dbo.UsersBad AS u
WHERE u.DisplayName = @DisplayName;
GO

Haunch

Solving implicit conversion issues is just as important (and often easier) than solving other issues with SARGable predicates, and just as important.

Even though it’s one of the first performance problems people learn about, I still see it out there enough to write about it. I think a lot of the reason that it still crops up is because ORMs leave developers detached from the queries, and they don’t see how parameter types end up getting inferred when not strongly typed.

Anyway, that’s it for this series. Tomorrow’s post will be a wrap up with links.

Thanks for reading!

Going Further

If this is the kind of SQL Server stuff you love learning about, you’ll love my training. I’m offering a 75% discount to my blog readers if you click from here. I’m also available for consulting if you just don’t have time for that and need to solve performance problems quickly.

Boniface

There are some things that, in the course of normal query writing, just can’t be SARGablized. For example, generating and filtering on a windowing function, a having clause, or any other runtime expression listed here.

There are some interesting ways to use indexed views to our advantage for some of those things. While windowing functions and having clauses can’t be directly in an indexed view, we can give an indexed view a good definition to support them.

I Don’t Care For It

Starting with these indexes to help things along, they don’t really do as much as we’d hope.

CREATE INDEX c
ON dbo.Comments (PostId);

CREATE INDEX v
ON dbo.Votes (PostId);

CREATE INDEX p
ON dbo.Posts (Id, OwnerUserId, Score);

This query has to process a ton of rows, and no matter what we set the having expression to, the entire result set has to be generated before it can be applied. We could set it to > 0 or > infinity and it would take the same amount of time to have a working set to apply it to.

SELECT 
    p.OwnerUserId,
    TotalScore = 
        SUM(ISNULL(p.Score * 1., 0.)),
    records = 
        COUNT_BIG(*)
FROM dbo.Posts AS p
JOIN  dbo.Comments AS c
    ON c.PostId = p.Id
JOIN dbo.Votes AS v
    ON v.PostId = p.Id
GROUP BY 
    p.OwnerUserId
HAVING 
    SUM(ISNULL(p.Score * 1., 0.)) > 5000000.
ORDER BY
    TotalScore DESC;

Limited Liability

I know that having clause looks funny there, but it’s not my fault. The sum of Score ends up being a really big integer, and overflows the regular sized integers unless you explicitly convert it to a bigint or implicitly convert it to something floaty. The isnull is there because the column is NULLable, which is unacceptable to an indexed view.

So, here we are, forced to write something weird to conform.

Sizzling. Sparkling.

Ignoring the woefully misaligned and misleading operator times, we can see in the query plan that again(!) a late Filter operator is applied that aligns with the predicate in our having clause.

Sarging Ahead

Let’s work some magic, here.

CREATE VIEW 
    dbo.BunchaCrap
WITH 
    SCHEMABINDING
AS
SELECT 
    p.OwnerUserId,
    TotalScore = 
        SUM(ISNULL(p.Score * 1., 0.)),
    records = 
        COUNT_BIG(*)
FROM dbo.Posts AS p
JOIN  dbo.Comments AS c
    ON c.PostId = p.Id
JOIN dbo.Votes AS v
    ON v.PostId = p.Id
GROUP BY 
    p.OwnerUserId;
GO 

CREATE UNIQUE CLUSTERED INDEX bc ON dbo.BunchaCrap(OwnerUserId);

This gives us an indexed view with the TotalScore expression materialized, which means we can search on it directly now without all the 50 some-odd seconds of nonsense leading up to it.

At The Mall

Our options now are either to query the indexed view directly with a noexpand hint, or to run the query as initially designed and rely on expression matching to pick up on things.

SELECT 
    bc.*
FROM dbo.BunchaCrap AS bc WITH (NOEXPAND)
WHERE bc.TotalScore > 5000000.
ORDER BY bc.TotalScore DESC;

In either case, we’ll get this query plan now:

Which looks a whole heck of a lot nicer.

Tomorrow, we’ll look at how implicit conversion can look a lot like non-SARGable predicates.

Thanks for reading!

Going Further

If this is the kind of SQL Server stuff you love learning about, you’ll love my training. I’m offering a 75% discount to my blog readers if you click from here. I’m also available for consulting if you just don’t have time for that and need to solve performance problems quickly.

Cheap Replica

The bottom line on scalar UDFs is that they’re poison pills for performance.

They’re bad enough in the select list, but they get even worse if they appear in join or where clause portions of the query.

The example query we’re going to use doesn’t even go out and touch other tables, which can certainly make things worse. It does all its processing “in memory”.

CREATE FUNCTION dbo.nonsargable(@d datetime)
RETURNS bit
WITH SCHEMABINDING,
     RETURNS NULL ON NULL INPUT
AS 
BEGIN

RETURN
(
    SELECT 
        CASE 
            WHEN DATEDIFF
                 (
                     DAY, 
                     @d, 
                     GETDATE()
                 ) > 30
            THEN 1
            ELSE 0
        END
);

END;
GO

Much Merch

When we run this query, the plan is a messmare.

SELECT
    c = COUNT_BIG(*)
FROM dbo.Users AS u
WHERE u.Reputation > 100
AND   dbo.nonsargable(u.LastAccessDate) = 1;

The Filter operator is a familiar face at this point.

The function has to run once per row that leaves the Users table, and the predicate has to be calculated and evaluated later at the filter operation.

Twisty

If we look in dm_exec_function_stats, we can see just how many times, and how much CPU the function used when we ran just this one query:

If you pay extra special attention, you’ll noticed that the execution_count here exactly matches the number of rows that pass through the Filter operator above.

The way to fix this is to  rewrite the function as an inline table valued function.

CREATE FUNCTION dbo.nonsargable_inline(@d datetime)
RETURNS TABLE
WITH SCHEMABINDING
AS 
RETURN

    SELECT 
        b = 
        CASE 
            WHEN DATEDIFF
                 (
                     DAY, 
                     @d, 
                     GETDATE()
                 ) > 30
            THEN 1
            ELSE 0
        END;

GO

Now we don’t have all those scalar problems.

Save The Wheels

We have to call our function a little bit differently, but that’s far less of a big deal.

SELECT
    c = COUNT_BIG(*)
FROM dbo.Users AS u
WHERE u.Reputation > 100
AND   (
          SELECT
              * 
          FROM dbo.nonsargable_inline(u.LastAccessDate)
       ) = 1;

And our query can go parallel, and take way less than 10 seconds.

While the predicate isn’t exactly ideal, in this case I’ll deal with it, and maybe even add some better indexes if I need to. At 183 milliseconds, I don’t need to do that right now.

Save The Feels

Scalar UDFs are still the downfall of many queries, and because this one has getdate in it, it can’t be inlined automatically in SQL Server 2019.

With simple enough functions, rewrites to inline table valued functions are easy enough. The more nonsense you put into the body of a function, the harder quick fixes get. I’ve spent days rewriting some that ran into the thousands of lines, and it’s not a pretty process.

Tomorrow we’re going to look at how indexed views can help you solve SARGability issues.

Thanks for reading!

Going Further

If this is the kind of SQL Server stuff you love learning about, you’ll love my training. I’m offering a 75% discount to my blog readers if you click from here. I’m also available for consulting if you just don’t have time for that and need to solve performance problems quickly.

Stucco

When you sit down to choose data types, max really should be last on your list. There are very few legitimate uses for them.

Start by asking yourself if someone may ever throw in a string over 4000 or 8000 characters, you’re storing XML or JSON, or some other foolish endeavor.

If the answer is “no”, or “just to be safe“, stop right there. Put that max datatype down. You’re really gonna hate when they end up doing to your memory grants.

But you’ll also hate what they do to queries that try to filter on them.

Columns

No matter how much you scream, holler, and curse, when you try to filter data in a column with a max type, that predicate can’t be pushed to when you touch the index.

Leaving aside that max data types can’t be in the key of an index, because that would be insane, even if you stick them in the includes you’re looking at a lot of potential bulk, and not the good kind that makes you regular.

Let’s look at these two queries. The first one hits the Body column, and the second one hits the Title column.

SELECT
    c = COUNT_BIG(*)
FROM dbo.Posts AS p
WHERE p.Body = N'A';

SELECT
    c = COUNT_BIG(*)
FROM dbo.Posts AS p
WHERE p.Title = N'A';

Both of these queries suck because we have to scan the entire Posts table, but at least the second one would be indexable if we cared enough to add one.

Arguments

Starting off clear: This will happen regardless of if your search argument is a variable or a parameter, regardless of recompile hints.

DECLARE @A nvarchar(MAX) = N'A';
SELECT
    c = COUNT_BIG(*)
FROM dbo.Posts AS p
WHERE p.Title = @A;
GO 

DECLARE @A nvarchar(MAX) = N'A';
SELECT
    c = COUNT_BIG(*)
FROM dbo.Posts AS p
WHERE p.Title = @A
OPTION(RECOMPILE);
GO

I see this quite often in ORMs where people don’t explicitly define datatypes, and stored procedures where people are being daft.

Surgeons

This is the sort of stuff you have to deal with when you use max data types.

They really are a pain, and the larger your tables are, the harder it can be to make changes later. Add in any sort of data synchronization and it all gets much worse.

In tomorrow’s post, we’ll look at how user defined functions can make things horrible.

As usual.

Thanks for reading!

Going Further

If this is the kind of SQL Server stuff you love learning about, you’ll love my training. I’m offering a 75% discount to my blog readers if you click from here. I’m also available for consulting if you just don’t have time for that and need to solve performance problems quickly.

Specific

The non-SARGABLE pattern that dynamic SQL helps you deal with is the the catch all query pattern, which can look like:

• col = @parameter or @parameter is null
• col = isnull(@parameter, col)
• col = coalesce(@parameter, col)

Or any similar variation of null/not null checking of a parameter (or variable) in the where/join clause at runtime.

Dynamic SQL allows you to build up the specific where clause that you need for the non-NULL set of search filters.

But First

Let’s look at one of my favorite demos, because it very simply shows the goofy kind of things that can go wrong when you don’t practice basic query hygiene.

I’m going to create these two indexes:

CREATE INDEX osc ON dbo.Posts
    (OwnerUserId, Score, CreationDate);

CREATE INDEX po ON dbo.Posts
    (ParentId, OwnerUserId);

They are fundamentally and obviously different indexes.

The query has a where clause on OwnerUserId and CreationDate, and an order by on Score.

The select list is, of course, everything.

DECLARE 
    @OwnerUserId int = 22656,
    @CreationDate datetime = '20190101',
    @SQLString nvarchar(MAX) = N'
SELECT 
    p.*
FROM dbo.Posts AS p
WHERE (p.OwnerUserId   = @OwnerUserId  OR @OwnerUserId IS NULL)
AND   (p.CreationDate >= @CreationDate OR @CreationDate IS NULL)
ORDER BY p.Score DESC;
';

EXEC sys.sp_executesql 
    @SQLString,
  N'@OwnerUserId INT,
    @CreationDate DATETIME',
    @OwnerUserId,
    @CreationDate;

But when we execute it, it uses the index on ParentId and OwnerUserId.

This is completely bizarre given the requirements of the query.

Now Second

Yes yes, I know, Captain Recompile. A hint will fix this problem. But then you might have another problem. Or a whole bunch of other problems.

Here’s an example of nice, safe dynamic SQL that gets the correct index used and a much more efficient query overall.

DECLARE 
    @OwnerUserId int = 22656,
    @CreationDate datetime = '20190101',
    @SQLString nvarchar(MAX) = N'
SELECT 
    p.*
FROM dbo.Posts AS p
WHERE 1 = 1' + NCHAR(10)

IF @OwnerUserId IS NOT NULL
BEGIN
    SET @SQLString += N'AND p.OwnerUserId   = @OwnerUserId' + NCHAR(10)
END

IF @CreationDate IS NOT NULL
BEGIN
    SET @SQLString += N'AND p.CreationDate >= @CreationDate' + NCHAR(10)
END

SET @SQLString += N'ORDER BY p.Score DESC;'

PRINT @SQLString;

EXEC sys.sp_executesql 
    @SQLString,
  N'@OwnerUserId INT,
    @CreationDate DATETIME',
    @OwnerUserId,
    @CreationDate;

And On The Seventh Day

The title of this series is SARGability week, and at first I had five posts set to go on this. As I was writing, I realized there were a few other things that I wanted to cover.

Next week I’m going to talk about max data types, user defined functions, and implicit conversions, then wrap things up.

Unless I think of something else. After all, I’m writing this on the 9th. Time travel, baby!

Thanks for reading

Going Further

If this is the kind of SQL Server stuff you love learning about, you’ll love my training. I’m offering a 75% discount to my blog readers if you click from here. I’m also available for consulting if you just don’t have time for that and need to solve performance problems quickly.