Hey guys!

Next Monday, 26/05, at 9pm (Brasília), I will be live on Canal .NET, together with the legend Renato Groffe, for a free online event with a series of practical tips on databases and modern development.

💡 Let's talk about:
– What’s new in SQL Server 2025
– How to use OpenTelemetry with .NET
– Tools like Gitleaks for code security
– Good practices and useful frameworks for those who develop with data

📺 The broadcast will be on YouTube!
🔗 Sign up here: https://www.meetup.com/pt-BR/dotnet-sao-paulo/events/307501819/

 

Stream link:

The new SQL Server Management Studio (SSMS) and Copilot

The first big change we noticed is not in the “engine”, but in the working tool. SSMS is now built on top of the modern Visual Studio shell. Before it was an independent product and somewhat visually outdated. Now, in addition to decent native support for Dark Mode, it uses the same Visual Studio installer.

But the star here is the Copilot in SSMS. It is already in preview and allows you to ask questions in natural language about the health of the database, backups and even generate complex queries.

Emphasis: To set up Copilot, you'll need an Azure Open AI endpoint, deployment name, and API key. It is not “magic” and free; it consumes resources from your Azure.

Fabric Mirroring: Transparent Replication

Microsoft is betting all its chips on Fabric. Mirroring allows you to replicate your on-premises SQL Server 2025 database to the Fabric Warehouse almost instantly and without impact on your transactional environment. It uses the log and the internal CDC to send the data. Is this the end of complex ETLs just to bring data to the analytics environment? Perhaps.

Native Regular Expressions (Regex): The 20-Year Wait is Over!

Finally, after two decades depending on CLR or workarounds with LIKE and PATINDEX, SQL Server 2025 brings native support for regular expressions. We have functions like REGEXP_LIKE, REGEXP_REPLACE, REGEXP_INSTR and others.

PERFORMANCE ALERT: The new Regex functions only work if the database is in compatibility mode 170 (SQL Server 2025).
Additionally, Regex consumes much more CPU than simple string operations. If using on gigantic tables, consider indexing persisted calculated columns with the result of the expression.

Native REST calls: sp_invoke_external_rest_endpoint

This functionality already existed in Azure SQL Database and now comes to the On-Premises environment. You can make HTTP requests (GET, POST, etc.) directly from a Stored Procedure.

Safety Tip: Don't leave API keys exposed in your T-SQL code. Use the feature Database Scoped Credentials to encapsulate the URL and Secret, ensuring the developer just calls the endpoint without seeing the key.

The New JSON Data Type

Until SQL Server 2022, JSON was stored as NVARCHAR(MAX). In SQL Server 2025, we have the data type JSON native. It automatically validates the syntax in INSERT or UPDATE. If you try to write a poorly formatted JSON, the bank will spit out an error right away.

SQL Server 2025: Standard Developer Edition

This is one of Microsoft's smartest changes for us developers. Previously, the version Developer was always a copy of the Enterprise. The problem? You develop something using expensive resources (such as advanced compression or specific partitions) and, when it comes to production, the client uses Standard, the code broke.

Now, when installing SQL Server 2025, you can choose the Standard Developer. It is free for development, but limits functionality to what the Standard license allows. This ensures that if it worked on your machine, it will work on the production server without any unpleasant surprises.

Generative AI and Vectors: The Future of T-SQL

SQL Server 2025 now natively handles Vector Data. This is the basis for creating semantic search applications (RAG) directly at the bank. Functions were introduced to generate “embeddings” and calculate the vector distance between texts.

Why does this matter? Instead of searching for exact words (LIKE ‘%carro%’), you search for context. If the user searches for “four-wheeled vehicle”, the vector search can identify the relationship with “car” in a mathematical way.