Struct NullableIndexExprResult 

pub struct NullableIndexExprResult {
    pub lower: NullableRowAddrMask,
    pub upper: NullableRowAddrMask,
    /* private fields */
}

Result of an index search before NULL rows are dropped. Each endpoint is a NullableRowAddrMask carrying SQL three-valued logic info.

Fields

lower: NullableRowAddrMask

Rows the index guarantees are TRUE.

upper: NullableRowAddrMask

Rows that may be TRUE. Rows outside upper are guaranteed to be FALSE / NULL (and so not in a WHERE answer set).

Implementations

impl NullableIndexExprResult

pub fn exact(mask: NullableRowAddrMask) -> Self

Precise result — every row in mask is in the answer and every row outside is not. Equivalent to the old Exact variant.

pub fn at_most(mask: NullableRowAddrMask) -> Self

Upper-bound-only result — rows outside mask are guaranteed not to match; rows inside may match and require a recheck. Equivalent to the old AtMost variant.

pub fn at_least(mask: NullableRowAddrMask) -> Self

Lower-bound-only result — rows in mask are guaranteed to match; rows outside may match too and require a recheck. Equivalent to the old AtLeast variant.

pub fn new(lower: NullableRowAddrMask, upper: NullableRowAddrMask) -> Self

Construct an interval result from lower/upper bounds.

lower rows are guaranteed TRUE and upper rows may be TRUE, with NULL state preserved at both endpoints. Equal endpoints are canonicalized to Self::exact so is_exact() remains structural.

pub fn is_exact(&self) -> bool

True if the result is exact — the answer is precisely the lower (== upper) mask.

This is a structural check on the canonical form produced by the constructors / algebra: an Exact(m) built with Self::exact holds equal masks, and elementwise & / | / ! preserve that. It is not a semantic emptiness test — a hand-constructed IndexExprResult whose endpoints are representationally distinct but semantically equal (e.g. AllowList(universe) vs BlockList(empty)) will report is_exact() == false. All in-tree code paths construct results through the canonical builders, so this is sound in practice.

The three shape predicates are not mutually exclusive — see the note on Self::is_at_least for the precedence convention.

pub fn is_at_most(&self) -> bool

True if lower matches no rows (canonical AllowList(∅)) — the index gives only an upper bound on the answer.

Like Self::is_exact, this is a structural check on the canonical form. See that doc for the caveat.

pub fn is_at_least(&self) -> bool

True if upper covers every row (canonical BlockList(∅)) — the index gives only a lower bound on the answer.

Precedence convention for consumers branching on shape: check Self::is_exact first (Exact-of-empty satisfies both is_exact and is_at_most; Exact-of-universe satisfies both is_exact and is_at_least); then is_at_least; finally treat the residual as is_at_most or Refined. The branches in filtered_read::apply_index_to_fragment follow this order.

pub fn drop_nulls(self) -> IndexExprResult

Project NULL rows out of the result.

Under a WHERE clause NULL is treated as FALSE, so drop_nulls folds NULL rows out of the answer at each endpoint.

Trait Implementations

impl BitAnd for NullableIndexExprResult

type Output = NullableIndexExprResult

The resulting type after applying the & operator.

fn bitand(self, rhs: Self) -> Self

Performs the & operation.

impl BitOr for NullableIndexExprResult

type Output = NullableIndexExprResult

The resulting type after applying the | operator.

fn bitor(self, rhs: Self) -> Self

Performs the | operation.

impl Clone for NullableIndexExprResult

fn clone(&self) -> NullableIndexExprResult

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for NullableIndexExprResult

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

Formats the value using the given formatter.

impl Not for NullableIndexExprResult

type Output = NullableIndexExprResult

The resulting type after applying the ! operator.

fn not(self) -> Self

Performs the unary ! operation.