Recently, I've been working on some bits for the iceberg connector that will support sampled statistics for the optimizer (#20937, #21296, #20993, and more). However, for sample maintenance to occur, we'll also need to write out those samples to a table in the filesystem somewhere.
Currently, in our java runtime implementation, the tables are written out to separate Iceberg tables at a known subdirectory. It would be ideal if we could keep the implementation the same between Java and C++ runtimes. This would require implementing write support for Iceberg tables in C++/Prestissimo.
ZacBlanco
commented
1 year ago
Recently, I've been working on some bits for the iceberg connector that will support sampled statistics for the optimizer (#20937, #21296, #20993, and more). However, for sample maintenance to occur, we'll also need to write out those samples to a table in the filesystem somewhere.
Currently, in our java runtime implementation, the tables are written out to separate Iceberg tables at a known subdirectory. It would be ideal if we could keep the implementation the same between Java and C++ runtimes. This would require implementing write support for Iceberg tables in C++/Prestissimo.
Presto Component, Service, or Connector
Prestissimo/Iceberg