Struct CsrMutableEdges 

pub struct CsrMutableEdges { /* private fields */ }

Mutable edge storage combining CSR base with delta slab

Provides efficient edge mutations with automatic rebuild when delta grows too large.

Implementations

impl CsrMutableEdges

pub fn new() -> Self

Create new mutable edges with empty base

pub fn with_coords(coords: Vec<AbsCoord>) -> Self

Create with initial vertex coordinates

pub fn add_edge(&mut self, from: VertexId, to: VertexId)

Add an edge, rebuilding if threshold reached

pub fn remove_edge(&mut self, from: VertexId, to: VertexId)

Remove an edge, rebuilding if threshold reached

pub fn out_edges(&self, v: VertexId) -> Vec<VertexId>

Get outgoing edges for a vertex (merged view)

pub fn out_edges_ref(&self, v: VertexId) -> Option<&[VertexId]>

Borrow outgoing edges when no delta mutations are pending.

This is a zero-allocation hot path for read-heavy evaluation/scheduling phases.

pub fn in_edges(&self, v: VertexId) -> &[VertexId]

Get incoming edges from base CSR (delta not applied for performance) After rebuild, this will include all changes

pub fn in_edges_ref(&self, v: VertexId) -> Option<&[VertexId]>

Borrow incoming edges when no delta mutations are pending.

This is a zero-allocation hot path for read-heavy evaluation/scheduling phases.

pub fn delta_size(&self) -> usize

Get the current delta size

pub fn num_edges_exact(&self) -> usize

Return the exact number of logical outgoing dependency edges, including pending delta mutations.

This is intended for read-only observability. When the delta slab is non-empty, the implementation walks the known vertex ids and merges each outgoing edge list, so callers should avoid putting it on hot evaluation paths.

pub fn rebuild(&mut self)

Force a rebuild of the CSR structure

pub fn begin_batch(&mut self)

Enter batch mode - defer rebuilds until the outer end_batch() call.

pub fn end_batch(&mut self)

Exit batch mode and rebuild if needed.

pub fn add_vertex(&mut self, coord: AbsCoord, vertex_id: u32) -> usize

Add a new vertex with its coordinate and ID

pub fn add_vertices_batch(&mut self, items: &[(AbsCoord, u32)])

Add many vertices at once; single rebuild at end.

pub fn update_coord(&mut self, vertex_id: VertexId, new_coord: AbsCoord)

Update coordinate for a vertex in the cache Marks for rebuild to maintain sort order

pub fn adjacency_with_carried_forward_edges( &self, adjacency: Vec<(u32, Vec<u32>)>, ) -> Vec<(u32, Vec<u32>)>

Return a copy of adjacency extended with the current base+delta out-edges of every existing vertex that the input does not cover.

Named-range pass-through vertices (NamedScalar/NamedArray) emit edges to their underlying cells via add_edge during load; those edges live in base/delta but are not part of the formula-target adjacency that bulk-ingest’s finalize hands to [build_from_adjacency]. Feeding the raw adjacency straight to that (pure) builder would therefore silently drop the pass-through vertices’ out-edges, and build_demand_subgraph could never reach the underlying cells. Callers run this first to merge those edges back in, then pass the result to build_from_adjacency.

Must be called BEFORE build_from_adjacency, which overwrites base/delta/vertex_ids.

pub fn build_from_adjacency( &mut self, adjacency: Vec<(u32, Vec<u32>)>, coords: Vec<AbsCoord>, vertex_ids: Vec<u32>, )

Build underlying CSR directly from adjacency and provided coords/ids. This replaces the current base and clears the delta slab.

Pure builder: it uses exactly the edges in adjacency and does not consult the existing base/delta. To preserve edges for vertices absent from adjacency (e.g. named-range pass-through vertices), run [adjacency_with_carried_forward_edges] first and pass its result in.

Trait Implementations

impl Debug for CsrMutableEdges

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

Formats the value using the given formatter.

impl Default for CsrMutableEdges

fn default() -> Self

Returns the “default value” for a type.