The devtool experts at JetBrains have added full-line code completion in the latest 2024.1 updates to popular JetBrains integrated development environments (IDEs) – including IntelliJ IDEA and other recently updated IDEs available from QBS.
JetBrains’ AI-powered full-line code completion complements the JetBrains AI Assistant in that it aims to speed up devs’ work even further. Together, they go beyond the standard code completion in JetBrains IDEs that’s been available for some time.
With the addition of full-line code completion, devs can:
- Complete lines of code with single-line suggestions based on current-file context.
- Benefit from specialised language models trained specifically for different languages and frameworks, without having to send code online.
- Code faster with greater accuracy, freeing up time and head space for innovation while supporting efficiency and productivity across production teams.
FULL-LINE CODE COMPLETION VS AI ASSISTANT
- Both offerings help teams benefit from AI functionality in JetBrains IDEs, but JetBrains AI Assistant offers additional features such as context-aware smart chat and the ability to generate tests or write documentation.
- Full-line code completion, on the other hand, runs locally without sending any data over the internet – broadening access to AI benefits in coding workflows, such as when connecting online may be risky for data protection and compliance.
- As more organisations build AI into daily workflows, few can afford to be left behind. JetBrains full-line code completion recognises and addresses the fact that AI can also introduce risk into certain workflows, giving an alternative way to secure AI benefits.
With JetBrains full-line code completion – available integrated with updates of Java, Kotlin, Python, JavaScript, TypeScript, CSS, PHP, Go, and Ruby IDEs, with functionality extended to C#, Rust, and C++ soon – the benefits include:
- Appropriate formatting of all suggestions.
- IDEs check for balanced brackets and quotes.
- Static analysis and JetBrains’ understanding of code deployed to filter out incorrect suggestions.
- Each supported language has its own set of suggested code correctness checks, helping ensuring IDEs only suggest existent variables and methods.
- Auto-import is supported.
- Smart filtering smooths workflows, eliminating suggestions likely to be cancelled explicitly or immediately deleted.
- Model inferences run in a separate process, and are optimised for target machine architecture – running on the CPU when it comes to x86-64, or on the GPU for ARM64 architecture, for example.
- Post-processing steps including syntax and semantics checks, smart filtering, formatting, parenthesis balancing, and various other manipulations to ensure good user experience.
- Control the potential for multiple review requirements by delivering only single full lines of code completion.
- Reduction of potential cloud-related expenses.
WORKING WITH JETBRAINS CODE COMPLETION
JetBrains full-line code completion can be tweaked relatively easily. It should be noted that this feature’s design made it possible to reduce the size of language models without quality taking a significant hit.
For instance, in the 2024.1 version of JetBrains IDEs, the company developed a language model 100 million parameter language model with a maximum context size of 1,536 tokens, or about 170 lines of code.
- To configure full line code completion: Simply go to Settings | Editor | General | Code Completion and check settings under the Machine Learning-Assisted Completion section.
- The feature itself can be disabled by unticking the Enable Full Line suggestions box, or by disabling the plugin via in Settings | Plugins | Installed.
Full line code completion is already available for IntelliJ IDEA Ultimate, PyCharm Professional, WebStorm, PhpStorm, GoLand, and RubyMine, and eventually also Rider, RustRover, and CLion. It is included with active JetBrains IDEs subscriptions on v2024.1 or later.
No JetBrains AI features are trained on customers’ code, but if teams seek to use AI to accelerate coding workflows while still complying with strict data privacy policies and regulations, full-code completion can fit the bill – a programming-language specific language model, which is trained in house using a dataset of open-source code with permissive licenses.
DO MORE WITH JETBRAINS TOOLS
Many great JetBrains tools for IDE projects are available through QBS, such as the Code With Me Enterprise collaborative coding and pair programming service. Code With Me means teams can easily collaborate and learn in real time on IDE projects.
Code With Me is also perfect for swarm programming to develop, debug, and fix code as well, including for IntelliJ IDEA, AppCode, CLion, GoLand, PhpStorm, PyCharm, RubyMine and WebStorm – all available via QBS.
Additional solutions (to name just three) include static code analysis tool Qodana for controlling code quality, JetBrains Academy to get teams up and running fast on builds, or the JetBrains All Products Pack. Check them all out today.
Email QBS at:sales@qbssoftware.com
Telephone QBS on: +44(0) 20 8733 7100