Add LAGraph_RegularPathQuerry

AGENTS.md

@@ -14,21 +14,28 @@ pathrex/
 ├── Cargo.toml                  # Crate manifest (edition 2024)
 ├── build.rs                    # Links LAGraph + LAGraphX; optionally regenerates FFI bindings
 ├── src/
-│   ├── lib.rs                  # Public modules: graph, formats, lagraph_sys; utils is pub(crate)
+│   ├── lib.rs                  # Public modules: graph, formats, rpq, sparql, lagraph_sys, utils
 │   ├── main.rs                 # Binary entry point (placeholder)
 │   ├── lagraph_sys.rs          # FFI module — includes generated bindings
 │   ├── lagraph_sys_generated.rs# Bindgen output (checked in, regenerated in CI)
-│   ├── utils.rs                # Internal helpers: CountingBuilder, CountOutput, VecSource,
-│   │                           #   grb_ok! and la_ok! macros
+│   ├── utils.rs                # Public helpers: CountingBuilder, CountOutput, VecSource,
+│   │                           #   grb_ok! and la_ok! macros, build_graph
 │   ├── graph/
 │   │   ├── mod.rs              # Core traits (GraphBuilder, GraphDecomposition, GraphSource,
 │   │   │                       #   Backend, Graph<B>), error types, RAII wrappers, GrB init
 │   │   └── inmemory.rs         # InMemory marker, InMemoryBuilder, InMemoryGraph
+│   ├── rpq/
+│   │   ├── mod.rs              # RPQ evaluation trait (RpqEvaluator), RpqResult, RpqError
+│   │   ├── nfarpq.rs           # NFA-based RPQ evaluator using LAGraph_RegularPathQuery
+│   │   └── rpqmatrix.rs        # Plan-based RPQ evaluator using LAGraph_RPQMatrix
+│   ├── sparql/
+│   │   └── mod.rs              # SPARQL parsing (spargebra), PathTriple extraction, parse_rpq
 │   └── formats/
 │       ├── mod.rs              # FormatError enum, re-exports
 │       └── csv.rs              # Csv<R> — CSV → Edge iterator (CsvConfig, ColumnSpec)
 ├── tests/
-│   └── inmemory_tests.rs       # Integration tests for InMemoryBuilder / InMemoryGraph
+│   ├── inmemory_tests.rs       # Integration tests for InMemoryBuilder / InMemoryGraph
+│   └── nfarpq_tests.rs         # Integration tests for NfaRpqEvaluator
 ├── deps/
 │   └── LAGraph/                # Git submodule (SparseLinearAlgebra/LAGraph)
 └── .github/workflows/ci.yml   # CI: build GraphBLAS + LAGraph, cargo build & test
@@ -204,6 +211,79 @@ Configuration is via [`CsvConfig`](src/formats/csv.rs:17):
 [`ColumnSpec`](src/formats/csv.rs:11) is either `Index(usize)` or `Name(String)`.
 Name-based lookup requires `has_header: true`.
 
+### SPARQL parsing (`src/sparql/mod.rs`)
+
+The [`sparql`](src/sparql/mod.rs) module uses the [`spargebra`](https://crates.io/crates/spargebra)
+crate to parse SPARQL 1.1 query strings and extract the single property-path
+triple pattern that pathrex's RPQ evaluators operate on.
+
+**Supported query form:** `SELECT` queries with exactly one triple or property
+path pattern in the `WHERE` clause, e.g.:
+
+```sparql
+SELECT ?x ?y WHERE { ?x <knows>/<likes>* ?y . }
+```
+
+Key public items:
+
+- [`parse_query(sparql)`](src/sparql/mod.rs:51) — parses a SPARQL string into a
+  [`spargebra::Query`].
+- [`extract_path(query)`](src/sparql/mod.rs:73) — validates a parsed `Query` is a
+  `SELECT` with a single path pattern and returns a [`PathTriple`](src/sparql/mod.rs:62).
+- [`parse_rpq(sparql)`](src/sparql/mod.rs:196) — convenience function combining
+  `parse_query` + `extract_path` in one call.
+- [`PathTriple`](src/sparql/mod.rs:62) — holds the extracted `subject`
+  ([`TermPattern`]), `path` ([`PropertyPathExpression`]), and `object`
+  ([`TermPattern`]).
+- [`ExtractError`](src/sparql/mod.rs:31) — error enum for extraction failures
+  (`NotSelect`, `NotSinglePath`, `UnsupportedSubject`, `UnsupportedObject`,
+  `VariablePredicate`).
+- [`RpqParseError`](src/sparql/mod.rs:204) — combined error for [`parse_rpq`]
+  wrapping both [`SparqlSyntaxError`] and [`ExtractError`].
+- [`DEFAULT_BASE_IRI`](src/sparql/mod.rs:44) — `"http://example.org/"`, the
+  default base IRI constant.
+
+The module also handles spargebra's desugaring of sequence paths
+(`?x <a>/<b>/<c> ?y`) from a chain of BGP triples back into a single
+[`PropertyPathExpression::Sequence`].
+
+### RPQ evaluation (`src/rpq/`)
+
+The [`rpq`](src/rpq/mod.rs) module provides an abstraction for evaluating
+Regular Path Queries (RPQs) over edge-labeled graphs using GraphBLAS/LAGraph.
+
+Key public items:
+
+- [`RpqEvaluator`](src/rpq/mod.rs:47) — trait with a single method
+  [`evaluate(subject, path, object, graph)`](src/rpq/mod.rs:48) that takes
+  SPARQL [`TermPattern`] endpoints, a [`PropertyPathExpression`] path, and a
+  [`GraphDecomposition`], returning an [`RpqResult`](src/rpq/mod.rs:42).
+- [`RpqResult`](src/rpq/mod.rs:42) — wraps a [`GraphblasVector`] of reachable
+  vertices.
+- [`RpqError`](src/rpq/mod.rs:21) — error enum covering parse errors, extraction
+  errors, unsupported paths, missing labels/vertices, and GraphBLAS failures.
+
+#### `NfaRpqEvaluator` (`src/rpq/nfarpq.rs`)
+
+[`NfaRpqEvaluator`](src/rpq/nfarpq.rs:265) implements [`RpqEvaluator`] by:
+
+1. Converting a [`PropertyPathExpression`] into an [`Nfa`](src/rpq/nfarpq.rs:27)
+   via Thompson's construction ([`Nfa::from_property_path()`](src/rpq/nfarpq.rs:35)).
+2. Eliminating ε-transitions via epsilon closure
+   ([`NfaBuilder::epsilon_closure()`](src/rpq/nfarpq.rs:198)).
+3. Building one `LAGraph_Graph` per NFA label transition
+   ([`Nfa::build_lagraph_matrices()`](src/rpq/nfarpq.rs:43)).
+4. Calling [`LAGraph_RegularPathQuery`] with the NFA matrices, data-graph
+   matrices, start/final states, and source vertices.
+
+Supported path operators: `NamedNode`, `Sequence`, `Alternative`,
+`ZeroOrMore`, `OneOrMore`, `ZeroOrOne`. `Reverse` and `NegatedPropertySet`
+return [`RpqError::UnsupportedPath`].
+
+Subject/object resolution: a [`TermPattern::Variable`] means "all vertices";
+a [`TermPattern::NamedNode`] resolves to a single vertex via
+[`GraphDecomposition::get_node_id()`](src/graph/mod.rs:195).
+
 ### FFI layer
 
 [`lagraph_sys`](src/lagraph_sys.rs) exposes raw C bindings for GraphBLAS and
@@ -212,10 +292,11 @@ LAGraph. Safe Rust wrappers live in [`graph::mod`](src/graph/mod.rs):
 - [`LagraphGraph`](src/graph/mod.rs:48) — RAII wrapper around `LAGraph_Graph` (calls
   `LAGraph_Delete` on drop). Also provides
   [`LagraphGraph::from_coo()`](src/graph/mod.rs:85) to build directly from COO arrays.
-- [`GraphblasVector`](src/graph/mod.rs:124) — RAII wrapper around `GrB_Vector`.
+- [`GraphblasVector`](src/graph/mod.rs:128) — RAII wrapper around `GrB_Vector`
+  (derives `Debug`).
 - [`ensure_grb_init()`](src/graph/mod.rs:39) — one-time `LAGraph_Init` via `std::sync::Once`.
 
-### Macros (`src/utils.rs`)
+### Macros & helpers (`src/utils.rs`)
 
 Two `#[macro_export]` macros handle FFI error mapping:
 
@@ -225,20 +306,28 @@ Two `#[macro_export]` macros handle FFI error mapping:
   appending the required `*mut i8` message buffer, and maps failure to
   `GraphError::LAGraph(info, msg)`.
 
+A convenience function is also provided:
+
+- [`build_graph(edges)`](src/utils.rs:184) — builds an `InMemoryGraph` from a
+  slice of `(&str, &str, &str)` triples (source, target, label). Used by
+  integration tests.
+
 ## Coding Conventions
 
 - **Rust edition 2024**.
-- Error handling via `thiserror` derive macros; two main error enums:
-  [`GraphError`](src/graph/mod.rs:15) and [`FormatError`](src/formats/mod.rs:24).
+- Error handling via `thiserror` derive macros; three main error enums:
+  [`GraphError`](src/graph/mod.rs:15), [`FormatError`](src/formats/mod.rs:24),
+  and [`RpqError`](src/rpq/mod.rs:21).
 - `FormatError` converts into `GraphError` via `#[from] FormatError` on the
   `GraphError::Format` variant.
-- Unsafe FFI calls are confined to `lagraph_sys`, `graph/mod.rs`, and
-  `graph/inmemory.rs`. All raw pointers are wrapped in RAII types that free
-  resources on drop.
+- Unsafe FFI calls are confined to `lagraph_sys`, `graph/mod.rs`,
+  `graph/inmemory.rs`, and `rpq/nfarpq.rs`. All raw pointers are wrapped in
+  RAII types that free resources on drop.
 - `unsafe impl Send + Sync` is provided for `LagraphGraph` and
   `GraphblasVector` because GraphBLAS handles are thread-safe after init.
 - Unit tests live in `#[cfg(test)] mod tests` blocks inside each module.
-  Integration tests that need GraphBLAS live in [`tests/inmemory_tests.rs`](tests/inmemory_tests.rs).
+  Integration tests that need GraphBLAS live in [`tests/inmemory_tests.rs`](tests/inmemory_tests.rs)
+  and [`tests/nfarpq_tests.rs`](tests/nfarpq_tests.rs).
 
 ## Testing
 
@@ -256,7 +345,13 @@ native libraries.
 
 Tests in `src/formats/csv.rs` are pure Rust and need no native dependencies.
 
-Tests in `src/graph/inmemory.rs` and [`tests/inmemory_tests.rs`](tests/inmemory_tests.rs)
+Tests in `src/sparql/mod.rs` are pure Rust and need no native dependencies.
+
+Tests in `src/rpq/nfarpq.rs` (NFA construction unit tests) are pure Rust and need no
+native dependencies.
+
+Tests in `src/graph/inmemory.rs`, [`tests/inmemory_tests.rs`](tests/inmemory_tests.rs),
+and [`tests/nfarpq_tests.rs`](tests/nfarpq_tests.rs)
 call real GraphBLAS/LAGraph and require the native libraries to be present.
 
 ## CI

Cargo.toml

@@ -8,6 +8,7 @@ csv = "1.4.0"
 libc = "0.2"
 oxrdf = "0.3.3"
 oxttl = "0.2.3"
+spargebra = "0.4.6"
 thiserror = "1.0"
 
 [features]

build.rs

@@ -83,6 +83,7 @@ fn regenerate_bindings() {
         .allowlist_function("LAGraph_Delete")
         .allowlist_function("LAGraph_Cached_AT")
         .allowlist_function("LAGraph_MMRead")
+        .allowlist_function("LAGraph_RegularPathQuery")
         .default_enum_style(bindgen::EnumVariation::Rust {
             non_exhaustive: false,
         })

src/graph/mod.rs

@@ -125,6 +125,7 @@ impl Drop for LagraphGraph {
 unsafe impl Send for LagraphGraph {}
 unsafe impl Sync for LagraphGraph {}
 
+#[derive(Debug)]
 pub struct GraphblasVector {
     pub inner: GrB_Vector,
 }

src/lagraph_sys_generated.rs

@@ -261,3 +261,37 @@ unsafe extern "C" {
         msg: *mut ::std::os::raw::c_char,
     ) -> ::std::os::raw::c_int;
 }
+unsafe extern "C" {
+    pub fn LAGraph_RegularPathQuery(
+        reachable: *mut GrB_Vector,
+        R: *mut LAGraph_Graph,
+        nl: usize,
+        QS: *const GrB_Index,
+        nqs: usize,
+        QF: *const GrB_Index,
+        nqf: usize,
+        G: *mut LAGraph_Graph,
+        S: *const GrB_Index,
+        ns: usize,
+        msg: *mut ::std::os::raw::c_char,
+    ) -> ::std::os::raw::c_int;
+}
+#[repr(u32)]
+#[derive(Debug, Copy, Clone, Hash, PartialEq, Eq)]
+pub enum RPQMatrixOp {
+    RPQ_MATRIX_OP_LABEL = 0,
+    RPQ_MATRIX_OP_LOR = 1,
+    RPQ_MATRIX_OP_CONCAT = 2,
+    RPQ_MATRIX_OP_KLEENE = 3,
+    RPQ_MATRIX_OP_KLEENE_L = 4,
+    RPQ_MATRIX_OP_KLEENE_R = 5,
+}
+#[repr(C)]
+#[derive(Debug, Copy, Clone)]
+pub struct RPQMatrixPlan {
+    pub op: RPQMatrixOp,
+    pub lhs: *mut RPQMatrixPlan,
+    pub rhs: *mut RPQMatrixPlan,
+    pub mat: GrB_Matrix,
+    pub res_mat: GrB_Matrix,
+}

src/lib.rs

@@ -1,6 +1,8 @@
 pub mod formats;
 pub mod graph;
+pub mod rpq;
+pub mod sparql;
 #[allow(unused_unsafe, dead_code)]
-pub(crate) mod utils;
+pub mod utils;
 
 pub mod lagraph_sys;

src/rpq/mod.rs

@@ -0,0 +1,54 @@
+//! Regular Path Query (RPQ) evaluation over edge-labeled graphs.
+//! ```rust,ignore
+//! use pathrex::sparql::parse_rpq;
+//! use pathrex::rpq::{RpqEvaluator, nfarpq::NfaRpqEvaluator};
+//!
+//! let triple = parse_rpq("SELECT ?x ?y WHERE { ?x <knows>/<likes>* ?y . }")?;
+//! let result = NfaRpqEvaluator.evaluate(&triple.subject, &triple.path, &triple.object, &graph)?;
+//! ```
+
+pub mod nfarpq;
+
+use crate::graph::GraphDecomposition;
+use crate::graph::GraphblasVector;
+use crate::sparql::ExtractError;
+use spargebra::SparqlSyntaxError;
+use spargebra::algebra::PropertyPathExpression;
+use spargebra::term::TermPattern;
+use thiserror::Error;
+
+#[derive(Debug, Error)]
+pub enum RpqError {
+    #[error("SPARQL syntax error: {0}")]
+    Parse(#[from] SparqlSyntaxError),
+
+    #[error("query extraction error: {0}")]
+    Extract(#[from] ExtractError),
+
+    #[error("unsupported path expression: {0}")]
+    UnsupportedPath(String),
+
+    #[error("label not found in graph: '{0}'")]
+    LabelNotFound(String),
+
+    #[error("vertex not found in graph: '{0}'")]
+    VertexNotFound(String),
+
+    #[error("GraphBLAS/LAGraph error: {0}")]
+    GraphBlas(String),
+}
+
+#[derive(Debug)]
+pub struct RpqResult {
+    pub reachable: GraphblasVector,
+}
+
+pub trait RpqEvaluator {
+    fn evaluate<G: GraphDecomposition>(
+        &self,
+        subject: &TermPattern,
+        path: &PropertyPathExpression,
+        object: &TermPattern,
+        graph: &G,
+    ) -> Result<RpqResult, RpqError>;
+}