Apache Arrow began as a language-independent specification for in-memory columnar data and has since grown into a broad, multi-language software ecosystem for analytics and data interchange. That breadth is powerful, but it can also be heavy for projects that only need a small subset of functionality: read Arrow data, write it, or exchange it efficiently with another library.

The Arrow C Data Interface was created to solve part of that problem by defining a very small, stable C interface for sharing Arrow-compatible data structures without a build-time dependency on the full Arrow library stack. But the interface is intentionally low-level, which makes direct use awkward for everyday application code.

This talk presents Sparrow as the missing middle layer. Sparrow provides C++20 libraries that map these low-level Arrow interfaces to expressive, C++ idiomatic APIs, making it practical to build and exchange Arrow data in projects that rely on standard modern C++ patterns. We will show how Sparrow integrates with standard algorithms, how it supports interoperability through Arrow’s interface specifications, and where the ecosystem is heading next, including cross-language bindings and future evolution.