Struct Reader

Source

pub struct Reader { /* private fields */ }

Expand description

Read transaction over a pinned snapshot.

A read transaction owns its own Arc<Snapshot> and never borrows the Db, so it stays valid across a later begin_write/checkpoint on the same handle (it cloned the snapshot before the write borrowed &mut). It is Send + Sync (asserted below).

§Performance

Creating and cloning a read transaction is O(1): it shares the pinned snapshot through an Arc, not by copying.

Implementations§

Source §

impl Reader

Source

pub fn pin(&self) -> ReadPin

Returns this transaction’s reader pin.

§Performance

This method is O(1).

Source

pub fn catalog(&self) -> &Catalog

Returns catalog metadata.

§Performance

This method is O(1).

Source

pub fn element_count(&self) -> usize

Returns visible element count.

§Performance

This method is O(base + overlay change).

Source

pub fn relation_count(&self) -> usize

Returns visible relation count.

§Performance

This method is O(base + overlay change).

Source

pub fn incidence_count(&self) -> usize

Returns visible incidence count.

§Performance

This method is O(base + overlay change).

Source

pub fn element_ids(&self) -> Vec<ElementId>

Returns every visible element id in id order.

§Performance

This method is O(element count).

Source

pub fn relation_ids(&self) -> Vec<RelationId>

Returns every visible relation id in id order.

§Performance

This method is O(relation count).

Source

pub fn contains_element(&self, id: ElementId) -> bool

Returns whether an element exists.

§Performance

This method is O(log change + log n).

Source

pub fn contains_relation(&self, id: RelationId) -> bool

Returns whether a relation exists.

§Performance

This method is O(log change + log n).

Source

pub fn contains_incidence(&self, id: IncidenceId) -> bool

Returns whether an incidence exists.

§Performance

This method is O(log change + log n).

Source

pub fn element(&self, id: ElementId) -> Option<Element>

Returns an owned element view — id, labels, and all properties read in one call.

§Performance

This method is O(log n + label count + property count).

Source

pub fn relation(&self, id: RelationId) -> Option<Relation>

Returns an owned relation view — id, type, labels, and all properties read in one call.

§Performance

This method is O(log n + label count + property count).

Source

pub fn incidence(&self, id: IncidenceId) -> Option<IncidenceRecord>

Returns an owned incidence record.

§Performance

This method is O(log change + log n).

Source

pub fn element_incidences(&self, id: ElementId) -> Vec<IncidenceRecord>

Returns every visible incidence attached to an element, in ascending incidence-id order.

The merged set mixes overlay-owned and base-borrowed records, so this returns an owned Vec (IncidenceRecord is Copy, so the copy is cheap).

§Performance

This method is O(base incidences + overlay incidence change).

Source

pub fn endpoints(&self, relation: RelationId) -> Option<(ElementId, ElementId)>

Returns a binary relation’s two endpoint elements, ordered by ascending incidence id.

Reads the relation’s incidences from the reverse-adjacency index and returns the elements carried by its first two incidences in id order. A relation with fewer than two visible incidences returns None. This reports endpoints structurally, without consulting any projection’s source/target roles — use Self::neighbors when role direction matters.

§Performance

This method is O(degree) over the relation’s incidences.

Source

pub fn neighbors( &self, element: ElementId, relation_type: RelationTypeId, direction: Direction, ) -> Vec<ElementId>

Returns the elements reachable from element along relations of relation_type, in ascending element-id order.

Direction selects the role element must play on each relation. Endpoint roles are encoded by incidence-creation order: a binary relation’s source is its lower incidence id and its target the higher (see Self::endpoints). Outgoing requires element to be the source (and yields the target), Incoming requires it to be the target (and yields the source), and Both yields the opposite endpoint either way. Resolved over the reverse-adjacency index — each incidence of element whose relation has the requested type contributes that relation’s other endpoint — so this works for any binary relation without a materialized projection.

§Performance

This method is O(degree of element + sum of touched relation degrees).

Source

pub fn property( &self, subject: PropertySubject, key: PropertyKeyId, ) -> Option<PropertyValue>

Returns one owned property value.

§Performance

This method is O(log subjects + log keys).

Source

pub fn element_by_key<T: ValueType>( &self, index: EqualityIndex<T>, value: impl Assignable<T>, ) -> Result<Option<Element>, DbError>

Returns the owned element whose value in index equals value, or None when no element matches.

§Errors

Returns DbError when the index is unknown or is not an equality index.

§Performance

This method is O(log n + label count + property count).

Source

pub fn count(&self, index: IndexId) -> Result<usize, DbError>

Returns the number of subjects carried by a membership index (a label or relation-type index).

§Errors

Returns DbError when the index is unknown or does not support membership enumeration.

§Performance

This method is O(indexed family size).

Source

pub fn lookup_property_equal( &self, key: PropertyKeyId, value: &PropertyValue, ) -> Result<Vec<PropertySubject>, DbError>

Looks up subjects with a property value.

§Errors

Returns DbError when the property key is unknown or value does not match the key schema.

§Performance

This method is O(property subject count).

Source

pub fn lookup_property_range( &self, key: PropertyKeyId, min: &PropertyValue, max: &PropertyValue, ) -> Result<Vec<PropertySubject>, DbError>

Looks up subjects with a property inside an inclusive range.

§Errors

Returns DbError when the property key is unknown or either bound does not match the key schema.

§Performance

This method is O(property subject count).

Source

pub fn lookup( &self, index: IndexId, lookup: Match<'_>, ) -> Result<Vec<PropertySubject>, DbError>

Executes an index lookup.

§Errors

Returns DbError when the index is unknown, the lookup shape does not match the index kind, or supplied property values do not match catalog schemas.

§Performance

This method is O(indexed family size).

Source

pub fn graph_projection( &self, id: ProjectionId, ) -> Result<GraphProjection, DbError>

Materializes a graph projection.

§Errors

Returns DbError when the projection is unknown, is not a graph, or fails validation against current topology.

§Performance

This method is O(relation count * incidence count).

Source

pub fn graph_projection_by_name( &self, name: &str, ) -> Result<GraphProjection, DbError>

Materializes a graph projection by catalog name.

§Errors

Returns DbError when the projection is unknown, is not a graph, or fails validation against current topology.

§Performance

This method is O(log projection count + relation count * incidence count).

Source

pub fn walk( &self, projection: ProjectionId, seeds: &[ElementId], walk: Walk, ) -> Result<Subgraph, DbError>

Walks a cataloged graph projection from canonical seed elements, returning the discovered nodes AND the projection edges among them.

Nodes are unique canonical elements in BFS first-discovery order; depth is the shortest discovered hop count from any seed. Edges connect two discovered nodes, ordered deterministically and unique by relation, so the Subgraph never references a node it omitted.

§Errors

Returns DbError when the projection is unknown, is not a graph, cannot be materialized, or a seed element is not part of the projection.

§Performance

This method is O(relation count * incidence count + visited edges).

Source

pub fn personalized_pagerank( &self, projection: ProjectionId, seeds: &[ElementId], config: PageRankConfig<f64>, ) -> Result<Vec<(ElementId, f64)>, DbError>

Ranks a cataloged graph projection by personalized PageRank, returning every projection element paired with its rank, ordered highest first.

seeds are the restart (teleport) set: rank mass returns to them on each damping step, biasing the stationary distribution toward elements reachable from the seeds (random walk with restart). The seed weights are normalized internally, so passing the seed elements is sufficient. With no seeds this is the uniform-teleport PageRank over the projection.

§Errors

Returns DbError when the projection is unknown, is not a graph, cannot be materialized, or PageRank rejects the configuration (the PageRankConfig was invalid or the power iteration did not converge). Seeds absent from the projection are ignored rather than erroring; with no resolvable seed this is the uniform-teleport rank.

§Performance

This method is O(relation count * incidence count + iterations * (visible elements + visible edges) + visible elements * log(visible elements)) — the trailing term is the final rank sort.

Source

pub fn longest_path( &self, projection: ProjectionId, elements: &[ElementId], ) -> Result<Vec<ElementId>, DbError>

Returns the longest chain of canonical elements along the projection’s outgoing edges within the subgraph induced by elements.

Only edges whose endpoints are both in elements participate. The path lists each element once from start to end; its length in edges is path.len() - 1. An empty elements slice yields an empty path.

§Errors

Returns DbError when the projection is unknown, is not a graph, cannot be materialized, or the induced subgraph contains a cycle. Elements absent from the projection are ignored, so the chain is computed over the present subset.