pub trait GraphReader: Send + Sync {
Show 16 methods
// Required methods
fn get_node(&self, id: NodeId) -> Result<Option<Node>>;
fn get_edge(&self, id: EdgeId) -> Result<Option<Edge>>;
fn all_node_ids(&self) -> Result<Vec<NodeId>>;
fn nodes_by_label(&self, label: &str) -> Result<Vec<NodeId>>;
fn nodes_by_property(
&self,
label: &str,
property: &str,
value: &Property,
) -> Result<Vec<NodeId>>;
fn outgoing(&self, id: NodeId) -> Result<Vec<(EdgeId, NodeId)>>;
fn incoming(&self, id: NodeId) -> Result<Vec<(EdgeId, NodeId)>>;
// Provided methods
fn nodes_by_properties(
&self,
label: &str,
properties: &[String],
values: &[Property],
) -> Result<Vec<NodeId>> { ... }
fn edges_by_property(
&self,
_edge_type: &str,
_property: &str,
_value: &Property,
) -> Result<Vec<EdgeId>> { ... }
fn list_property_indexes(&self) -> Result<Vec<(String, Vec<String>)>> { ... }
fn list_edge_property_indexes(&self) -> Result<Vec<(String, Vec<String>)>> { ... }
fn list_point_indexes(&self) -> Result<Vec<(String, String)>> { ... }
fn list_edge_point_indexes(&self) -> Result<Vec<(String, String)>> { ... }
fn nodes_in_bbox(
&self,
label: &str,
property: &str,
srid: i32,
xlo: f64,
ylo: f64,
xhi: f64,
yhi: f64,
) -> Result<Vec<NodeId>> { ... }
fn edges_in_bbox(
&self,
_edge_type: &str,
_property: &str,
_srid: i32,
_xlo: f64,
_ylo: f64,
_xhi: f64,
_yhi: f64,
) -> Result<Vec<EdgeId>> { ... }
fn list_property_constraints(&self) -> Result<Vec<PropertyConstraintSpec>> { ... }
}Expand description
Read-side counterpart to crate::GraphWriter. Gives the executor a
uniform view of the graph regardless of whether the data behind it lives
entirely in a local storage engine (single-node or full-replica Raft
mode) or is sharded across cluster peers (routing mode, where a
partitioned reader fans out point reads to owners and scatter-gathers
bulk scans).
Methods are sync because the executor’s iterator model is sync. Async-
backed implementations (e.g. a remote reader that talks gRPC) bridge via
Handle::block_on; callers must run the executor inside spawn_blocking
so they don’t stall the tokio runtime.
Required Methods§
fn get_node(&self, id: NodeId) -> Result<Option<Node>>
fn get_edge(&self, id: EdgeId) -> Result<Option<Edge>>
fn all_node_ids(&self) -> Result<Vec<NodeId>>
fn nodes_by_label(&self, label: &str) -> Result<Vec<NodeId>>
Sourcefn nodes_by_property(
&self,
label: &str,
property: &str,
value: &Property,
) -> Result<Vec<NodeId>>
fn nodes_by_property( &self, label: &str, property: &str, value: &Property, ) -> Result<Vec<NodeId>>
Equality lookup via a property index. Callers (planner) must
have verified the (label, property) index exists before
emitting this call — fallback implementations are free to do a
label-scan-and-filter for impls that don’t maintain their own
property index, but the storage-backed reader treats a call on
a non-existent index as an empty result since no entries are
maintained.
fn outgoing(&self, id: NodeId) -> Result<Vec<(EdgeId, NodeId)>>
fn incoming(&self, id: NodeId) -> Result<Vec<(EdgeId, NodeId)>>
Provided Methods§
Sourcefn nodes_by_properties(
&self,
label: &str,
properties: &[String],
values: &[Property],
) -> Result<Vec<NodeId>>
fn nodes_by_properties( &self, label: &str, properties: &[String], values: &[Property], ) -> Result<Vec<NodeId>>
Composite form of Self::nodes_by_property. properties
and values are parallel slices of equal length — every
slot must be present for the call to land a match. The
default impl delegates to nodes_by_property for length-1
slices and returns empty otherwise, so readers that haven’t
wired a native composite seek degrade to no-results rather
than mis-answering. The storage-backed blanket overrides
with a real composite lookup.
Sourcefn edges_by_property(
&self,
_edge_type: &str,
_property: &str,
_value: &Property,
) -> Result<Vec<EdgeId>>
fn edges_by_property( &self, _edge_type: &str, _property: &str, _value: &Property, ) -> Result<Vec<EdgeId>>
Relationship-scope analogue of Self::nodes_by_property.
The planner only emits an EdgeSeek after confirming a
(edge_type, property) index is registered via
Self::list_edge_property_indexes. Default impl returns
empty so readers that haven’t wired a native seek path
degrade to no-results rather than mis-answering; the
storage-backed blanket overrides with a real lookup.
Sourcefn list_property_indexes(&self) -> Result<Vec<(String, Vec<String>)>>
fn list_property_indexes(&self) -> Result<Vec<(String, Vec<String>)>>
Snapshot the (label, property) pairs of every property
index visible through this reader. Used by SHOW INDEXES.
Default impl returns empty — the storage-backed reader
overrides via the blanket impl, and partitioned/overlay
readers delegate to their bases.
Sourcefn list_edge_property_indexes(&self) -> Result<Vec<(String, Vec<String>)>>
fn list_edge_property_indexes(&self) -> Result<Vec<(String, Vec<String>)>>
Relationship-scope analogue of Self::list_property_indexes.
Returns (edge_type, property) pairs for every registered
edge property index. Default impl returns empty; overlay
and partitioned readers delegate to their bases.
Sourcefn list_point_indexes(&self) -> Result<Vec<(String, String)>>
fn list_point_indexes(&self) -> Result<Vec<(String, String)>>
Snapshot the (label, property) pairs of every point /
spatial index visible through this reader. Used by SHOW POINT INDEXES. Default impl returns empty; storage-backed
readers override via the blanket impl.
Sourcefn list_edge_point_indexes(&self) -> Result<Vec<(String, String)>>
fn list_edge_point_indexes(&self) -> Result<Vec<(String, String)>>
Relationship-scope analogue of Self::list_point_indexes.
(edge_type, property) pairs. Default impl returns empty.
Sourcefn nodes_in_bbox(
&self,
label: &str,
property: &str,
srid: i32,
xlo: f64,
ylo: f64,
xhi: f64,
yhi: f64,
) -> Result<Vec<NodeId>>
fn nodes_in_bbox( &self, label: &str, property: &str, srid: i32, xlo: f64, ylo: f64, xhi: f64, yhi: f64, ) -> Result<Vec<NodeId>>
Axis-aligned bounding-box range query over the point index
(label, property). Returns every node carrying label
whose property is a Property::Point under srid that
falls inside [xlo..xhi] × [ylo..yhi].
Default impl does the naive scan (iterate nodes_by_label,
filter in memory) so that readers that don’t maintain a
spatial index — or that haven’t wired a native range query —
stay correct, just slow. Partitioned readers inherit this
default and get cluster-wide correctness through the
scatter-gather nodes_by_label underneath. The
storage-backed blanket overrides with the Z-order seek.
Sourcefn edges_in_bbox(
&self,
_edge_type: &str,
_property: &str,
_srid: i32,
_xlo: f64,
_ylo: f64,
_xhi: f64,
_yhi: f64,
) -> Result<Vec<EdgeId>>
fn edges_in_bbox( &self, _edge_type: &str, _property: &str, _srid: i32, _xlo: f64, _ylo: f64, _xhi: f64, _yhi: f64, ) -> Result<Vec<EdgeId>>
Relationship-scope analogue of Self::nodes_in_bbox.
Default impl returns empty — edge-scoped bbox queries aren’t
yet part of the planner’s rewrite surface, so no read path
exercises this on anything but the storage-backed blanket.
When the edge point-seek lowering lands this default should
grow a naive scan via edges_by_type (needs to be added to
the trait first).
Sourcefn list_property_constraints(&self) -> Result<Vec<PropertyConstraintSpec>>
fn list_property_constraints(&self) -> Result<Vec<PropertyConstraintSpec>>
Snapshot every registered constraint visible through this
reader, for SHOW CONSTRAINTS and db.constraints(). Default
impl returns empty; storage-backed readers override.
Implementors§
impl GraphReader for StorageReaderAdapter<'_>
impl<T: StorageEngine> GraphReader for T
Blanket impl: any sized type that implements StorageEngine
is automatically a GraphReader. Covers the concrete
meshdb_storage::RocksDbStorageEngine — a &RocksDbStorageEngine
coerces to &dyn GraphReader via this blanket.
Not covered: dyn StorageEngine itself. Rust does not transitively
coerce &dyn StorageEngine to &dyn GraphReader because trait
objects of unrelated traits carry different vtables and there’s no
supertrait relationship connecting them. Call sites that hold a
&dyn StorageEngine should use StorageReaderAdapter to wrap it
as a GraphReader.