GitHub user paleolimbot added a comment to the discussion: Proposal: ADBC file drivers with pluggable storage
I'll append this to day that ADBC isn't the right venue for hashing out the details of specifying a filesystem or filter unambiguously across a boundary: data sources are data sources and files are tables. Data source APIs (like GDAL's) typically lump databases and files together, and some file types *are* databases (e.g., geopackage is SQLite). Catalogs are increasingly compositions of all of these, and pushing a scan into a database is also the same as pushing a scan into a file (the question of which filters can or should be pushed down, and how to ensure they remain unambiguous across that boundary). GitHub link: https://github.com/apache/arrow-adbc/discussions/4877#discussioncomment-18987916 ---- This is an automatically sent email for [email protected]. To unsubscribe, please send an email to: [email protected]
