Comparison
Hive-style partitioningvsOpen table format
Hive-style partitioning
the partition value is encoded in the folder name as dt=2026-01-14, and the column does not exist inside the files at all.
Encoding partition values in the object key path, so a reader infers them from the directory structure. It is simple, portable and universally understood, and it is why so much of the lake looks like a folder tree. Its weaknesses are what the table formats fix: renaming a partition scheme means rewriting every path, listing is expensive at scale, and the partition value is a string that must be parsed rather than a typed column.
Full entry →Open table format
there is a metadata layer over the files that says exactly which files are in the table right now, so two writers no longer overwrite each other.
A specification that turns a directory of files into a table with atomic commits, snapshot isolation and schema evolution, by keeping an explicit manifest of the files each version contains. It is what removes the lake's oldest failure — readers seeing a half-written table — and it makes deletes and updates expressible without rewriting everything. Because the metadata is the table, its maintenance becomes something you have to run.
Full entry →