VECTOR
A fixed-length vector of FP16 (half-precision) floating-point values. Used for similarity search and ML embedding workloads. Declared as VECTOR(n) where n is the number of dimensions.
Casting
| From | Example | Notes |
|---|---|---|
| from ARRAY |
float_array_col::VECTOR(384) |
Quantizes each element to FP16 |
Comparisons
This type does not support direct comparisons with =, <, or >. Extract or cast values first.
Notes
Similarity search uses dedicated functions such as COSINE_DISTANCE(a, b) and COSINE_SIMILARITY(a, b). Standard comparison operators are not supported on VECTOR.
Limitations
- Vector columns cannot be used with standard comparison operators (=, <, >, etc.).
- The dimension count must match between vectors in any operation.