Clarification on Delta Table schema type naming (“bool” vs “boolean”)

[shanisolomon]

Hi,

I wanted to confirm whether a Delta table whose metadata action defines a column with the data type "bool" is considered valid. Based on the Delta protocol and Parquet specifications, my understanding is that "boolean" is the officially supported type name.

More generally, are the data type names listed in the protocol documentation
the only supported ones, or should implementations accept type name variants (for example, "integer" vs "int", "boolean" vs "bool", etc.)?

Could you please clarify whether "bool" and similar variants are valid, and whether Delta readers are expected to recognize or handle them?

Thanks,
Shani

[shanisolomon]

+@larsk-db who might know the answer to this one :)

[larsk-db]

Hmm, looking at that table, I don't think those are even the official names. They are just indicative of the Delta type, not the literal spellings. For example, the table says int but the actual value Delta produces in schemas is integer.

We should probably update that table.
But in the meantime, I think readers should support the schema as it is in the golden files.

[larsk-db]

It looks like these are the actual type names: https://github.com/delta-io/delta/blob/master/PROTOCOL.md#primitive-types

[larsk-db]

Except the actual name for "timestamp without timezone" is timestamp_ntz apparently.