Subgraph Pattern Matching via Bitmap Label Encoding

The patent applies preliminary action by pre-ordering query vertices based on their label selectivity and pre-encoding labels as bitmaps before traversal begins. This preparation work is done in advance to optimize the subsequent traversal process, reducing cache misses and improving throughput during actual pattern matching operations.

The patent changes parameters by representing vertex and edge labels as bitmaps instead of traditional string or integer representations. This parameter transformation enables efficient bitwise operations for label matching, reduces memory footprint, and improves cache utilization during graph traversal, directly addressing the cache thrashing problem.

The patent transforms label data from traditional representations to bitmap representations, where each label is represented as a bit position in a compact integer. This parameter change maintains full matching accuracy through bitwise operations while dramatically reducing memory usage, as bitmaps require far less space than traditional label storage methods.

The patent creates compact bitmap copies of label information that can be efficiently processed during traversal. Instead of storing and processing full label strings or complex data structures, the system uses lightweight bitmap copies that preserve matching semantics while occupying minimal memory, thus maintaining accuracy with reduced memory footprint.