Struct ExprLowerer 

pub struct ExprLowerer<'a> { /* private fields */ }

Lowers IR expressions from an ExprArena into DataFusion [Expr] values.

Construct once per plan lowering pass and call lower for each ExprId you need to convert.

Implementations

impl<'a> ExprLowerer<'a>

pub fn new( arena: &'a ExprArena, ontology: Option<&'a OntologyHandle>, var_map: &'a VarMap, ) -> Self

Creates a new lowerer, building the PropId → name map from the ontology only (empty in exploratory mode).

• arena: the expression arena from the [GraphPlan] being lowered.
• ontology: the ontology handle, if one is loaded (may be None in exploratory mode).
• var_map: maps variable IDs to DataFusion column name strings.

Prefer with_prop_names when the caller has a map that also covers runtime-catalog (exploratory) property names.

pub fn with_prop_names( arena: &'a ExprArena, var_map: &'a VarMap, prop_names: HashMap<u32, String>, ) -> Self

Creates a lowerer with a prebuilt PropId → name map.

The GraphPlanLowerer builds the map once (merging ontology + runtime catalog) and clones it into each per-operator ExprLowerer (the maps are small).

pub fn with_prop_names_and_nodes( arena: &'a ExprArena, var_map: &'a VarMap, prop_names: HashMap<u32, String>, node_shapes: HashMap<u32, NodeShape>, type_id_to_entity_name: HashMap<u32, String>, props_authoritative: bool, ) -> Self

Like with_prop_names but also seeded with the VarId.0 → NodeShape map for bare-node-value materialization (#785).

props_authoritative is true when node_shapes’ property lists come from a real backing dataset (so an absent property is genuinely absent → null, #598); pass false for schema-only / explain lowering. See the field docs.

pub fn with_input_schema(self, schema: DFSchemaRef) -> Self

Attach the input plan’s schema so a relational op’s PropertyAccess can resolve a temporal-component accessor (d.year) by the base column’s Arrow type (ADR 0009 / #920).

pub fn with_elem_struct_col(self, col: String) -> Self

Push col onto the chain of synthetic per-element columns in scope, so a PropertyAccess on it lowers via struct-aware get_field rather than a dotted property-column name (#1004). Called once per enclosing loop, outermost first, so nested loops keep distinct bindings (#1021).

pub fn with_read_target(self, dir: PathBuf) -> Self

Attach the project directory for read-side lowering, so nodes(p) bakes a hydrating cypher_path_nodes (labels + property union, #1024).

pub fn lower(&self, id: ExprId) -> Result<DfExpr, LoweringError>

Lower the expression identified by id to a DataFusion [Expr].

Errors

Returns LoweringError if a variable is unbound, a function is unknown, or an expression variant is not yet supported.