Struct mz_sql::plan::Scope

pub struct Scope {
    pub items: Vec<ScopeItem>,
    pub ungrouped_columns: Vec<ScopeUngroupedColumn>,
    pub lateral_barrier: bool,
}

Fields

items: Vec<ScopeItem>

The items in this scope.

ungrouped_columns: Vec<ScopeUngroupedColumn>

The ungrouped columns in the scope.

lateral_barrier: bool

Implementations

impl Scope

pub fn empty() -> Self

pub fn from_source<I, N>( table_name: Option<PartialItemName>, column_names: I, ) -> Self

where I: IntoIterator<Item = N>, N: Into<ColumnName>,

pub fn column_names(&self) -> impl Iterator<Item = &ColumnName>

Constructs an iterator over the canonical name for each column.

pub fn len(&self) -> usize

fn all_items<'a>( &'a self, outer_scopes: &'a [Scope], ) -> impl Iterator<Item = ScopeCursor<'a>> + 'a

Iterates over all items in the scope.

Items are returned in order of preference, where the innermost scope has the highest preference. For example, given scopes A(B(C)), items are presented in the order C, B, A.

Items are returned alongside the column reference that targets that item and the item’s “lateral level”. The latter bears explaining. The lateral level is the number of lateral barriers between this scope and the item. See Scope::lateral_barrier for a diagram. Roughly speaking, items from different levels but the same lateral level are items from different joins in the same subquery, while items in different lateral levels are items from different queries entirely. Rules about ambiguity apply within an entire lateral level, not just within a single scope level.

NOTE(benesch): Scopereally shows its weaknesses here. Ideally we'd have separate types likeQueryScopeandJoinScope` that more naturally encode the concept of a “lateral level”, or at least something along those lines.

pub fn items_from_table<'a>( &'a self, outer_scopes: &'a [Scope], table: &PartialItemName, ) -> Result<Vec<(ColumnRef, &'a ScopeItem)>, PlanError>

Returns all items from the given table name in the closest scope.

If no tables with the given name are in scope, returns an empty iterator.

NOTE(benesch): This is wrong for zero-arity relations, because we can’t distinguish between “no such table” and a table that exists but has no columns. The current design of scope makes this difficult to fix, unfortunately.

fn resolve_internal<'a, M>( &'a self, outer_scopes: &[Scope], matches: M, table_name: Option<&PartialItemName>, column_name: &ColumnName, ) -> Result<ColumnRef, PlanError>

where M: FnMut(&ScopeCursor<'_>) -> bool,

Returns a matching ColumnRef, if one exists.

Filters all visible items against the provided matches closure, and then matches this filtered set against the provided column_name.

pub fn resolve_column<'a>( &'a self, outer_scopes: &[Scope], column_name: &ColumnName, ) -> Result<ColumnRef, PlanError>

Resolves references to a column name to a single column, or errors if multiple columns are equally valid references.

pub fn resolve_using_column( &self, column_name: &ColumnName, join_side: JoinSide, ) -> Result<ColumnRef, PlanError>

Resolves a column name in a USING clause.

pub fn resolve_table_column<'a>( &'a self, outer_scopes: &[Scope], table_name: &PartialItemName, column_name: &ColumnName, ) -> Result<ColumnRef, PlanError>

pub fn resolve<'a>( &'a self, outer_scopes: &[Scope], table_name: Option<&PartialItemName>, column_name: &ColumnName, ) -> Result<ColumnRef, PlanError>

pub fn resolve_expr<'a>(&'a self, expr: &Expr<Aug>) -> Option<ColumnRef>

Look to see if there is an already-calculated instance of this expr. Failing to find one is not an error, so this just returns Option

pub fn product(self, right: Self) -> Result<Self, PlanError>

pub fn project(&self, columns: &[usize]) -> Self

pub fn table_names(&self) -> BTreeSet<&PartialItemName>

Trait Implementations

impl Clone for Scope

fn clone(&self) -> Scope

Returns a copy of the value.

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

Performs copy-assignment from source.

impl Debug for Scope

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

Formats the value using the given formatter.