Considerations when Deleting Lots of Data

Ed Elliott takes us through things to think about before deleting a few million rows from a table:

Fragmentation
Fragmentation occurs when we delete from pages, and there is still data surrounding our data. If we have 100 rows and delete every odd row, we would have perfect fragmentation in that we have doubled the size of the data that we need. If we delete rows 1-49, even though we remove the same number of rows we don’t have any fragmentation as the data is in a continuous block. Knowing how the data is stored on disk and how the data will be deleted, is it the first x records or every x record is vital so that we know whether, after the delete, we should also reorganise the indexes to remove the deleted records.

Ed has quality insights here, so check it out.

Puzzling Through Older Problems

Kenneth Fisher shares a couple of interview puzzles:

The year is 2004. You’re taking a tech test as an interview for a SQL development job. They have a page in their application that displays up to 20 rows of information. They need a piece of code that will return the rows from a given page. Oh, and it may not always be 20 rows per page. You need to write a piece of code where they can pass in a page number and page size and get back results. So for example, if the page size is 20 and the page is 3 then you need to return back rows 41 to 60.

The answers aren’t on the page, but then again, that’s the point of a puzzle.

Testing Maximum Rows in Table Value Constructors

Solomon Rutzky dives into how big a table value constructor can be in terms of rows:

On 2019-05-08, a helpful individual, Michael B, commented on my answer saying that the 1000-row limit only existed when using a TVC as the VALUES clause of an INSERT statement. And, that there was no limit when using a TVC as a derived table.

Could this be true?

Solomon finds out. Click through and so can you.

T-SQL Tips Regarding Subqueries

Itzik Ben-Gan provides quality information on working with subqueries in SQL Server:

In this plan you see a Nested Loops (Left Semi Join) operator, with a scan of the clustered index on Customers as the outer input and a seek in the index on the customerid column in the Orders as the inner input. You also see an outer reference (correlated parameter) based on the custid column in Customers, and the seek predicate Orders.customerid = Customers.custid.

So why are you getting the plan in Figure 1 and not the one in Figure 2? If you haven’t figured it out yet, look closely at the definitions of both tables—specifically the column names—and at the column names used in the query. You will notice that the Customers table holds customer IDs in a column called custid, and that the Orders table holds customer IDs in a column called customerid. However, the code uses custid in both the outer and inner queries. 

Itzik covers three specific scenarios, all of which can cause trouble to database developers who haven’t been burned yet. And sometimes even those who have.

Splitting Arrays with OPENJSON

Dave Mason continues a journey into parsing JSON with T-SQL:

Starting with SQL Server 2016, Microsoft provided a STRING_SPLIT function. It is a table-valued function that splits a string into rows of substrings, based on a specified separator character. It’s been a welcome addition that we waited a long time for. It has one shortcoming, though: the order of the output rows is not guaranteed to match the order of the substrings in the input string.

Microsoft also provided support for parsing JSON data starting with SQL Server 2016. I discovered the OPENJSON function can be used to split strings, and it can also return the ordinal position of each substring from the original input string.

There are some limitations which you’d expect, namely around requirements for valid JSON.

Parsing JSON with T-SQL

Dave Mason has a primer on JSON parsing using T-SQL:

Microsoft added support for JSON data beginning with SQL Server 2016. JSON is an open-standard file format consisting of attribute–value pairs and array data types. It is commonly used to transmit data objects for asynchronous browser–server communication. But it is also used for storing unstructured data in files or NoSQL databases such as Microsoft Azure Cosmos DB. For most of us, SQL Server’s support for JSON probably means two things: we can convert relational data to JSON and vice versa. In this post, I’ll focus on converting JSON to relational data and share what I’ve learned from a recent experience.

I’ve been pleasantly surprised with the way JSON support works in SQL Server. It’s supported every complicated scenario I’ve had to deal with so far, including nesting, deciding with or without arrays for the outer element, quotes or no quotes around numbers, etc.

Minimal Logging When Inserting into Heaps

Paul White gives us the lowdown on minimal logging when performing INSERT..SELECT operations into heap tables:

When inserting rows using INSERT...SELECT into a heap with no nonclustered indexes, the documentation universally states that such inserts will be minimally logged as long as a TABLOCK hint is present. This is reflected in the summary tables included in the Data Loading Performance Guide and the Tiger Team post. The summary rows for heap tables without indexes are the same in both documents (no changes for SQL Server 2016):

But it’s not quite that straightforward, as Paul shows. Read the whole thing.

Replacing Text Across SQL Agent Jobs

Max Vernon shares a script to perform a find-and-replace across SQL Agent jobs:

Once in a while you might need to make common changes to a lot of SQL Server Agent Jobs. For example, if you change the path where you store SQL Server backup files, you might need to update many jobs to point at \\SERVERB\Backups instead of \\SERVERA\Backups. The script below provides a simple instance-wide find-and-replace for SQL Server Agent job-step commands. It modifies the command text for all jobs that contain the matching @Find parameter, replacing it with the provided @Replace value. You can exclude jobs by adding them to the list of values in the #excludeJobs table.

Click through for the script.

Breaking Up Queries with UNION ALL

Bert Wagner takes us through a scenario where it can be faster to combine queries with UNION ALL rather than using IN:

Even though this query reads the whole clustered index to get the Benefactor rows, the total number of logical reads is still smaller than the seek/key lookup pattern seen in the combined query with IN(). This UNION ALL version gives SQL Server the ability to build a hybrid execution plan, combining two different techniques to generate a plan with fewer overall reads.

Click through for the example.

Finding Percentage of NULL Values on a Column

Denis Gobo shows how you can find the percentage of NULL (or non-NULL) values in a column:

This came up the other day, someone wanted to know the percentage of NULL values in a column

Then I said “I bet you I can run that query without using a NULL in the WHERE clause, as a matter of fact, I can run that query without a WHERE clause at all!!”

Bonus lesson: maybe don’t bet against Denis.

Categories

May 2019
MTWTFSS
« Apr  
 12345
6789101112
13141516171819
20212223242526
2728293031