Hi Andrew, Thrift messages are defined in the Thrift Interface Definition Language[1] and the binary encoding used in Parquet is the "Thrift Compact Protocol" [2].
There is some additional detail in the "Background: Apache Thrift" section of the blog[3] for anyone curious for more details Andrew [1]: https://thrift.apache.org/docs/idl [2]: https://github.com/apache/thrift/blob/master/doc/specs/thrift-compact-protocol.md [3]: https://arrow.apache.org/blog/2025/10/23/rust-parquet-metadata/ On Thu, Oct 23, 2025 at 6:25 PM Andrew Bell <[email protected]> wrote: > Hi Andrew, > > I'm having trouble locating documentation of the binary encoding of Thrift. > Can you point me to it? > > Thanks, > > On Thu, Oct 23, 2025 at 1:54 PM Andrew Lamb <[email protected]> > wrote: > > > Breaking this off into its own thread. > > > > In case anyone is interested, I just published a blog[1] post about the > new > > metadata decoder we released for the Rust implementation of Parquet that > > explains background, results we achieved, and how it works. > > > > Andrew > > > > [1]: https://arrow.apache.org/blog/2025/10/23/rust-parquet-metadata/ > > > > -- > Andrew Bell > [email protected] >
