Struct ColumnInfo 

pub struct ColumnInfo {
    pub name: String,
    pub data_type: SqlDataType,
    pub nullable: bool,
    pub is_primary_key: bool,
    pub default: Option<String>,
    pub raw_type: Option<String>,
    pub int_width: Option<IntWidth>,
    pub float_width: Option<FloatWidth>,
}

Metadata about a single column.

Fields

name: String

data_type: SqlDataType

nullable: bool

is_primary_key: bool

default: Option<String>

Default value expression (e.g. “UUID_V7”, “ULID”, “NANOID(10)”, “0”, “‘active’”).

raw_type: Option<String>

Raw type string as stored in the catalog (e.g. "JSON", "TEXT", "FLOAT64"). None for columns synthesized by the planner (e.g. auto-injected id). Columnar INSERT converters use this to reconstruct the exact ColumnType so JSON / Geometry / UUID columns are not incorrectly inferred as String. Write coercion reads a VECTOR(dim) declaration from it on a column whose data_type advertises text.

int_width: Option<IntWidth>

Declared width of an integer column, resolved once from the catalog at adapter-construction time. None for non-integer columns and for integer columns whose declared type carried no width.

This is deliberately a resolved [IntWidth] rather than another raw string: it is read on both the write path (range validation) and the read path (RowDescription OID and binary payload width), and those two must agree exactly. Resolving once at the catalog boundary is what makes disagreement unrepresentable.

float_width: Option<FloatWidth>

Declared width of a floating-point column, resolved once from the catalog at adapter-construction time. None for non-float columns and for float columns whose declared type the catalog never recorded.

The float analogue of ColumnInfo::int_width, and resolved from the same source. Read on the read path to select the RowDescription OID (700 vs 701) and the binary payload width (4 vs 8 bytes), and read on the write path by check_declared_float_ranges — but only for the narrower failure mode a float has: narrowing an f64 to f32 rounds rather than wraps (PostgreSQL accepts-and-rounds a double literal into a real column, and so does nodedb), so only a finite value that would overflow to infinity is rejected, not the full out-of-range check int_width gets.

Trait Implementations

impl Clone for ColumnInfo

fn clone(&self) -> Self

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for ColumnInfo

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.