I don’t really think about Jupyter Notebooks often, I’ve seen some nice demos of use cases for us SQL Server people (See this session by Rob Sewell for a few examples) but never given them much thought beyond that.
Which possibly shows how unglamorous they can be given that I use one virtually every day at work.
Just the one though, it’s a PowerShell notebook and as PowerShell can be used to run git commands I use it to hold the scripts that either take me too long to type out (like refreshing all the branches for our shared UAT and live environments in one go) or that I forget the syntax for because I don’t use it often enough and there’s no friendly button in Visual Studio to save me.
I thought I’d focus on something a little more interesting for this blog instead and so went back to one of the use cases that interested me when I first encountered SQL Notebooks. Using them to combine both text and code into a nice neat demo of a T-SQL feature. When I first investigated this most of my ideas fell flat because each code block was considered a different session and so temporary tables couldn’t be carried between them, this prevented you building up a decent example data set and then using it in subsequent steps. The good news is that that issue has now been fixed and so I thought I’d try converting my series of blog posts on querying XML in SQL Server into a set of notebooks.
The process was pleasantly simple, it’s just a matter of copying and pasting over and most of the formatting for headings, bullet points and links carried over without issue. A few times I had to delve into the markdown code to strip out a rogue change of font but each post took me 5-10 minutes to adapt.
You can access the results here they’re even viewable online at GitHub although the formatting of results tables isn’t quite as neat as in Azure Data Studio and there’s no way to run the sample code.
There’s a lot to be said for the approach, the long bits of code that set up the sample data can be hidden and results can be optionally saved with the notebook. As long as you have access to SQL Server you can click to run the sample code or start to experiment by changing it.
There are a few down sides. The most obvious is that while temp tables work between blocks of code, intellisense isn’t smart enough to know this and so throws up a lot of red wiggly lines. I’d also love to be able to collapse results sets without wiping them completely in the same way that you can do for code. Sometimes I wanted the results to be there but they were too long winded to want them to be filling up the page by default.
Limits come from Azure Data Studio as well, it’s not an issue for these posts but the lack of query plans will likely send me running back to good old SSMS at some point in the future.
Overall I like the result. I have something that I can easily point people to that allows them to run my example code but it also allows for more control over the surrounding text than is possible in a classic SQL script. Given that it’s relatively easy to convert a blog post into a notebook I’m going to keep at it for now and where appropriate attach a notebook version to each blog post.
Querying XML In SQL Server Series
- Part 1 – Introduction
- Part 2 – Simple Paths and Attributes With the Value Method
- Part 3 – Handling Repeating Regions With the Nodes Method
- Part 4 – Filtering With The Exist Method
- Part 5 – Introduction to the Query Method and XQuery
- Part 6 – Sequencing Elements and Finding Positions
- Part 7 – Creating XML