Along with a Visual Studio Code update to version 1.77, Microsoft has previewed fresh features including a Data Wrangler for data scientists working in Python that will automatically generate code to transform data, and new Copilot integration that supports AI chat.

VS Code updates on a monthly cycle, but this is a quiet month for new features that are generally available – though it is notable that this is the first version to ship with TypeScript 5, released in the middle of last month. TypeScript 5.0 introduces Decorator Expressions, a way of customizing classes in a reusable manner, as we examined here.

TypeScript 5.0 is not a native feature of VS Code though. The most significant activity this month lies in preview releases which are part of an updated VS Code Insiders build, which includes features likely to come to the production release soon.  

One that will grab attention is Copilot Chat, a chat-based UI for Microsoft’s AI-driven programming assistant. Microsoft took care to announce this first for Visual Studio, its commercial IDE, even though developers say that Copilot works better in the free VS Code. Copilot for VS Code has 4.7 million installs and four stars in the Visual Studio Code marketplace, whereas Copilot for Visual Studio has 237,000 installs and two and half stars.

According to Product Manager Chris Dias, the idea for Chat came when “the VS Code team had an internal hackathon to explore how it could integrate AI into VS Code more deeply,” and “kept circling around the Chat experience.” The idea was to provide ChatGPT but within the context of a specific code base, to provide features like explaining code. “Know that Copilot isn’t perfect and will give wrong answers,” warns Dias, but nevertheless he says that building, debugging, deploying and maintaining applications with AI assistance will become the norm. Developers can now try Copilot Chat view in VS Code, via the Copilot Nightly extension.

The other major feature in preview is called Data Wrangler. As described by Microsoft Data Science product manager Jeffrey Mew, Data Wrangler is for visualizing and transforming data, including the ability to “generate code using open-source Python libraries for the data transformations you perform.” Data Wrangler integrates with Pandas (Python Data Analysis Library). Mew said that Data Wrangler draws on the same technology from Microsoft Research that drives Excel’s Flash Fill feature, which apparently is called PROSE (Program Synthesis Using Examples).  

There are two ways to launch Data Wrangler. One is from a Jupyter Notebook that uses Panda data frames, which now shows a Launch Data Wrangler button, and the second is by selecting a CSV data file, right-clicking, and choosing Open in Data Wrangler.

Initial reaction is positive. “The code generation has been saving me a ton of trips to StackOverflow”, said one early adopter.