SubgraphMatching | Memory Subgraph Matching : An In-depth Study

The filtering methods that generate candidate vertex sets. Note that (1) every filtering method enables LDF by default; and (2) We optimize the filtering method of TurboIso with the technique of CFL. The ordering methods that generate matching order. The enumeration methods that find all results. We can execute a query by specifying the filtering method, the ordering method, and the enumeration method. For example, we evaluate the query with the GraphQL algorithm in the following command. Apart from evaluating queries with a specific algorithm, we can also execute queries by integrating techniques from different algorithms. For example, we generate the candidates with the label and degree filter, obtain the matching order with the ordering method of CFL, and finally enumerate the results with the set-intersection based local candidates computation method. Note that the source code cannot support the filtering/ordering/enumeration orders of CECI to integrate with other algorithms. In addition to setting the filtering, ordering and enumeration methods, you can also configure the set intersection algorithms and optimization techniques by defining macros in 'configuration/config.h'. Please see the comments in 'configuration/config.h' for the detail. We briefly introduce these methods in the following. Furthermore, our source code can support the spectrum analysis on a query. Specifically, we evaluate a query by generate a number of matching orders by performing the permutation of query vertices. For each order, we set a time limit for evaluating the query. In the following command line, we generate 1000 matching orders and set the time limit for each order as 60 seconds. Note that you need to first define SPECTRUM in 'configuration/config.h' file before executing the command.