Result Type Support
HarborSQL encodes Databricks SQL connector result pages directly from Arrow arrays. The current Thrift result type matrix is explicit:
| Arrow/DataFusion type | Databricks type metadata | Thrift value representation |
|---|---|---|
Boolean | boolean | boolean |
Int8 | tinyint | int |
Int16 | smallint | int |
Int32 | int | int |
Int64, UInt8, UInt16, UInt32 | bigint | bigint |
UInt64 | bigint | bigint only when the value fits in signed i64 |
Float32 | float | double |
Float64 | double | double |
Utf8, LargeUtf8, Utf8View | string | string |
Date32, Date64 | date | string |
Timestamp | timestamp | string |
Binary, LargeBinary, FixedSizeBinary | binary | binary |
Decimal128, Decimal256 | decimal with precision/scale qualifiers | string |
List, LargeList, FixedSizeList | array | string |
Map | map | string |
Struct | struct | string |
Decimal and nested values are rendered in the compact Databricks-style display form expected by connector compatibility tests. Arrays render as bracketed values, maps and structs render as JSON-like objects, string values are quoted, and nested date/timestamp values use Databricks-style textual dates and timestamps.
Unsupported Arrow types, including interval, dictionary, duration, list views,
and time-only values, return UNSUPPORTED_RESULT_TYPE instead of being coerced.
Query Surface
The default runtime is intentionally read-only. It targets SELECT workloads
against Unity Catalog Delta tables and does not currently provide Cloud Fetch,
durable result storage, streaming result storage, or broad Databricks SQL
protocol coverage.