Bills of materials, logistics networks and supplier chains are graphs stored in relational tables. Walking them from the application is slow; walking them inside the engine is not. This release adds a DataBlade that does the walk where the data is.
The DataBlade
- Named graphs over edge tables A graph is registered over an ordinary edge table, which remains the source of truth; a compressed adjacency snapshot is built in named memory on demand and refreshed by a routine.
- Iterator routines Neighbours, reachable nodes by breadth-first search with a minimum depth, and shortest path by Dijkstra, consumed as table functions in a SELECT. Scalars for reachability, path cost and degree. Create, drop, refresh and statistics for management.
Proven on real shapes
- Tests Bill-of-materials explosion and where-used, a 660-route European logistics network, supplier-recall impact, and a quantity-weighted cost roll-up, with expected results kept beside the SQL.
- Engine facts recorded How Informix 14.10 passes 64-bit integers, parses text-mode results and spells recursive common table expressions, documented for the next DataBlade.
Vector search was the first extension the platform built inside Informix. Graph traversal is the second, and it answers a question an ERP asks every day.