First release of the 6.38 series.
🗒️ Release notes
💾 Install instructions
We are proud to announce the Development release ROOT 6.38.00 🏆
The focus for this release was on expanding usability from Python and interoperability with its ecosystem, data analysis and modernisation. However, there is much more than that: ease of installation, runtime performance, stability (>250 items in our trackers have been addressed). See the full release notes and highlights below!
📦 Distribution
- In alpha mode, you can install 6.38 releases with pip on Linux: pip install root 🚀
📈 RDataFrame.
- Ergonomics: You don’t need to write template arguments for Snapshot any more, i.e.
myDf.Snapshot<int, float>(... - You can snapshot systematic variations to TTree in experimental mode. See systematic variations for the details!
- Distributed RDataFrame: RNTuples are treated internally as TTrees. Without API changes, your processing of RNTuples will be even faster than before!
🐍 Python Interface
- In RDataFrame, you can pass to
DefineandFiltera set of Python functions wider than before. - An enhanced Unified Histogram Interface (UHI) implementation is now even more stable, thanks to a reinforced testing suite extending the UHI testing helper.
7️⃣ ROOT7
- New highly performant histograms are available for testing: see these examples!
- ROOT::Experimental::RFile, a minimal and modern interface to ROOT files, is there!
💾 RNTuple
- Clarity: you can support RNTuple with the classic browser, profiting from several visual improvements, including a new treemap visualization of field sizes.
- Parallelism: The parallel writer became part of the stable API. See it in action in this example.
- Performance: (vectors of) fixed-size arrays are now treated in bulk
- Usability: automatic schema evolution is now akin to the classic I/O behavior. You can read expert documentation in tree/ntuple/doc/SchemaEvolution.md.
🧑🏽💻 C++ Interpreter
- The LLVM powering Cling has been upgraded to version 20: more runtime performance and features thanks to the new compiler!
- You can use SYCL interactively if you build ROOT with -Dexperimental_adaptivecpp=ON
📊 RooFit
- With the RooMultiPdf, you can minimize floating categorical variables, e.g. for models with variables that toggle between different background shapes.
Please note that 6.38 is a short term support release. As such, it is not meant to be used for data taking. It will be superseded by the 6.40 release, which is foreseen in May 2026. Patch releases of 6.38 will be provided until June 2026.