Struct InsertSeed 

pub struct InsertSeed<D, T> { /* private fields */ }

Implementations

impl<D, T: Table> InsertSeed<D, T>

pub fn values<R: InsertRow<Table = T>>(self, row: R) -> Insert<D, R>

pub fn select<Scope, Sel, SelIdx, TgtIdx>( self, query: &Select<D, Scope, Sel>, ) -> InsertSelect<D, T>

where D: Dialect, T: WrittenColumns, T::Columns: ColumnList<WrittenTable<T>, TgtIdx>, <<T as WrittenColumns>::Columns as ColumnList<WrittenTable<T>, TgtIdx>>::Fields<RowNil>: ColumnNames, Sel: Selection<Scope, SelIdx>, Sel::Output: SameShape<Row<<<T as WrittenColumns>::Columns as ColumnList<WrittenTable<T>, TgtIdx>>::Fields<RowNil>>>,

INSERT INTO t (..) SELECT .. — the rows a query produces, checked against the target’s own row by row::SameShape, the same one comparison a UNION branch and a CTE body go through.

The query fills every column the target lets a statement write: all of them but the generated ones, which the database writes itself and refuses a value for. SameShape compares name and type cell by cell, so the source’s columns must be spelled and typed as the target’s — SQL would widen an INTEGER into a BIGINT and take a NOT NULL value for a nullable column, and neither is accepted here. A source column under another name takes a label!{} one.

Known limitations: the source is a Select, so a SetOp (UNION) or a DynSelect cannot be one; a one-column target still needs a one-tuple (select((t::only,))), since a bare selection is a value rather than a row.

pub fn values_all<R: InsertRow<Table = T> + Insertable>( self, rows: impl IntoIterator<Item = R>, ) -> Result<Insert<D, R>, NothingToInsert>

Every row of a collection at once — the shape a bulk import has, where the rows are already in a Vec and the first one isn’t special. INSERT with no rows has no SQL form, so an empty collection is refused here rather than rendered.