Enum SourceSpec 

pub enum SourceSpec {
    ScanVertices {
        rows: Vec<Vec<Value>>,
        col_names: Vec<String>,
    },
    CteScan {
        cte_name: String,
        col_names: Vec<String>,
    },
    StandaloneValues {
        values: Vec<Vec<ContextualExpression>>,
        col_names: Vec<String>,
    },
    StorageScanVertices {
        space_name: String,
        limit: Option<usize>,
        col_names: Vec<String>,
        projected_properties: Vec<String>,
        identity_only: bool,
        skip_rows: bool,
        predicate: Vec<ScanPredicate>,
        tag: Option<String>,
        partition_range: Option<Range<i64>>,
        shard_slice: Option<ShardSlice>,
    },
    ScanEdges {
        rows: Vec<Vec<Value>>,
        col_names: Vec<String>,
    },
    StorageScanEdges {
        space_name: String,
        limit: Option<usize>,
        edge_type: Option<String>,
        col_names: Vec<String>,
        projected_properties: Vec<String>,
        identity_only: bool,
        predicate: Vec<ScanPredicate>,
        shard_slice: Option<ShardSlice>,
    },
    GetVertices {
        space_name: String,
        tag: String,
        vertex_ids: Option<Vec<Value>>,
        projected_properties: Vec<String>,
        col_names: Vec<String>,
    },
    GetEdges {
        space_name: String,
        edge_type: Option<String>,
        src: Option<String>,
        dst: Option<String>,
        rank: i64,
        projected_properties: Vec<String>,
    },
    GetNeighbors {
        space_name: String,
        tag: String,
        direction: String,
        projected_properties: Vec<String>,
    },
    IndexScan {
        space_name: String,
        index_name: String,
        index_id: u64,
        tag: String,
        predicate: Box<BoundIndexPredicate>,
        projection: IndexProjection,
        residual_filter: Option<Expression>,
        output_layout: Arc<SlotLayout>,
    },
    Argument {
        col_names: Vec<String>,
    },
    GetProp {
        space_name: String,
        entity_slot: usize,
        prop_names: Vec<String>,
        is_vertex: bool,
        output_layout: Arc<SlotLayout>,
    },
    Start,
}

Immutable config for source operators.

Mutable state (cursor, buffer, current_index, partition_id, partition_range) lives in SourceState.

Variants

ScanVertices

Fields

rows: Vec<Vec<Value>>

col_names: Vec<String>

CteScan

Recursive-CTE working-table scan.

Reads the rows published by the enclosing fixpoint operator under the mangled cte_name (see crate::cte). Produced when a step pattern labels a node with the CTE name; never touches storage.

Fields

cte_name: String

col_names: Vec<String>

StandaloneValues

Standalone DML values — evaluated once per execution in the source operator so that volatile expressions (e.g. now()) are resolved at execution time, not at plan build time.

Fields

values: Vec<Vec<ContextualExpression>>

col_names: Vec<String>

StorageScanVertices

Fields

space_name: String

limit: Option<usize>

col_names: Vec<String>

projected_properties: Vec<String>

identity_only: bool

Identity mode: the scan variable is never consumed as a whole entity downstream, so the entity column carries lightweight Value::VertexId instead of boxed vertices.

skip_rows: bool

Rowless mode (set with identity_only by the planner when the direct consumer is column-capable): skip the row view entirely and emit typed columns only. Valid only when the columnar path is on; the scan falls back to rows when it is off.

predicate: Vec<ScanPredicate>

Scan predicates pushed into the storage layer (pure pre-filter; the original filter still runs on top).

tag: Option<String>

Tag-restricted scan: only rows of this tag are scanned at the storage layer (the matching contains(labels(v), ...) residual conjunct may then be elided).

partition_range: Option<Range<i64>>

Optional vertex-id range restricting the scan to one partition. Set only on per-partition copies produced by the partitioned physical-plan builder; None means a full scan.

shard_slice: Option<ShardSlice>

Optional shard slice dividing a partitioned scan’s enumeration across partitions. Set together with a cleared partition_range on per-partition copies whose union must cover the table without repeating full-table walks; None enumerates the whole table.

ScanEdges

Fields

rows: Vec<Vec<Value>>

col_names: Vec<String>

StorageScanEdges

Fields

space_name: String

limit: Option<usize>

edge_type: Option<String>

col_names: Vec<String>

projected_properties: Vec<String>

identity_only: bool

Identity mode: the scan variable is never consumed as a whole entity downstream, so the entity column carries lightweight Value::EdgeHeader instead of boxed edges.

predicate: Vec<ScanPredicate>

Scan predicates pushed into the storage layer (pure pre-filter; the original filter still runs on top).

shard_slice: Option<ShardSlice>

Optional shard slice dividing a partitioned edge scan’s source enumeration across partitions. Set on per-partition copies whose union must cover every edge without repeating full-table walks; None walks the whole source space.

GetVertices

Fields

space_name: String

tag: String

vertex_ids: Option<Vec<Value>>

projected_properties: Vec<String>

col_names: Vec<String>

Entity variable name for the fetched vertex (e.g. the tag name).

GetEdges

Fields

space_name: String

edge_type: Option<String>

src: Option<String>

dst: Option<String>

rank: i64

projected_properties: Vec<String>

Property names to keep on the materialized edge; empty reads all.

GetNeighbors

Fields

space_name: String

tag: String

direction: String

projected_properties: Vec<String>

IndexScan

Index scan with typed predicate and projection.

The predicate and projection are validated at build time against the index schema. Stale row IDs are skipped, not treated as EOF.

Fields

space_name: String

index_name: String

index_id: u64

tag: String

Tag owning the indexed rows; covering vertex rows are rebuilt with it.

predicate: Box<BoundIndexPredicate>

projection: IndexProjection

residual_filter: Option<Expression>

output_layout: Arc<SlotLayout>

Argument

Produces one singleton row from which correlated apply can pull. col_names mirrors the outer (left) layout so the emitted chunk slots align with the correlation frame read at runtime.

Fields

col_names: Vec<String>

GetProp

Unary property retrieval: reads properties from the input entity IDs.

Unlike the source-variant GetProp (which is a zero-input stub), this unary variant takes an input child and produces one output row per input row with additional property columns.

Fields

space_name: String

entity_slot: usize

Slot number of the entity (vertex or edge) column in the input chunk.

prop_names: Vec<String>

Property names to read.

is_vertex: bool

Whether the entity is a vertex or edge.

output_layout: Arc<SlotLayout>

Output layout (input columns + new property columns).

Start

Produces one zero-column row on first next(), then None. Used as the seed for command-type operators (DDL, DML, etc.) that require a source but have no real input.

Trait Implementations

impl Clone for SourceSpec

fn clone(&self) -> Self

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for SourceSpec

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

Formats the value using the given formatter.