Struct sql::plan::scope::ScopeItem

pub struct ScopeItem {
    pub table_name: Option<PartialName>,
    pub column_name: ColumnName,
    pub exprs: HashSet<Expr<Aug>>,
    pub from_single_column_function: bool,
    pub allow_unqualified_references: bool,
    pub lateral_error_if_referenced: bool,
    pub is_exists_column_for_a_table_function_that_was_in_the_target_list: bool,
    _private: (),
}

Fields

table_name: Option<PartialName>

The name of the table that produced this scope item, if any.

column_name: ColumnName

The name of the column.

exprs: HashSet<Expr<Aug>>

The expressions from which this scope item is derived. Used by GROUP BY.

from_single_column_function: bool

Whether the column is the return value of a function that produces only a single column. This accounts for a strange PostgreSQL special case around whole-row expansion.

allow_unqualified_references: bool

Controls whether the column is only accessible via a table-qualified reference. When false, the scope item is also excluded from SELECT *.

This should be true for almost all scope items. It is set to false for join columns in USING constraints. For exmaple, in t1 FULL JOIN t2 USING a, t1.a and t2.a are still available by fully-qualified reference, but a bare a refers to a new column whose value is coalesce(t1.a, t2.a). This is a big special case because normally having three columns in scope named a would result in “ambiguous column reference” errors.

lateral_error_if_referenced: bool

Whether reference the item should produce an error about the item being on the wrong side of a lateral join.

Per PostgreSQL (and apparently SQL:2008), we can’t simply make these items unnameable. These items need to exist because they might shadow variables in outer scopes that would otherwise be valid to reference, but accessing them needs to produce an error.

is_exists_column_for_a_table_function_that_was_in_the_target_list: bool

For table functions in scalar positions, this flag is true for the ordinality column. If true, then this column represents an “exists” flag for the entire row of the table function. In that case, this column must be excluded from * expansion. If the corresponding datum is NULL, then * expansion should yield a single NULL instead of a record with various datums.

_private: ()

Implementations

impl ScopeItem

pub fn empty() -> ScopeItem

pub fn from_column_name<N>(column_name: N) -> ScopeItem where
    N: Into<ColumnName>, 

Constructs a new scope item from an unqualified column name.

pub fn from_name<N>(
    table_name: Option<PartialName>,
    column_name: N
) -> ScopeItem where
    N: Into<ColumnName>, 

Constructs a new scope item from a name.

pub fn from_expr(expr: impl Into<Option<Expr<Aug>>>) -> ScopeItem

Constructs a new scope item with no name from an expression.

pub fn is_from_table(&self, table_name: &PartialName) -> bool

Trait Implementations

impl Clone for ScopeItem

fn clone(&self) -> ScopeItem

Returns a copy of the value.

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

Performs copy-assignment from source.

impl Debug for ScopeItem

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

Formats the value using the given formatter.