03 / adoption boundary
Limitations
What reachq does not promise yet, and the questions to answer before using it as production infrastructure.
Edit this pageUpdated with the source
What is supported
- Pure-Python directed-graph primitives and deterministic seeded constructions.
- Exact reachability-preservation checks for shortcut-set workflows.
- Weighted hopset construction with an explicit epsilon parameter and baseline comparison.
What is not promised
- No universal practical speedup follows from beta or hopbound alone.
- The hopset path is sequential in the current Python implementation.
- Research algorithms and optional acceleration backends can change without a stable API promise.
- There is no published wheel or hosted graph service yet.
- Exact transitive closure is output-quadratic and guarded by a pair budget.
Use NetworkX or igraph for general graph manipulation. Use reachq when its shortcut-set or hopset constructions are the specific thing you need.
Benchmark honestly
Record the commit, Python version, operating system, CPU, graph source, vertex and edge counts, seed, algorithm parameters, construction time, query time, memory, output size, and approximation error. A theoretical quantity is not a measured product speedup.