Block graphs in practice

Motivated by the rapidly increasing size of genomic databases, code repositories and versioned texts, several compression schemes have been proposed that work well on highly-repetitive strings and also support fast random access: e.g., LZ-End, RLZ, GDC, augmented SLPs, and block graphs. Block graphs have good worst-case bounds but it has been an open question whether they are practical. We describe an implementation of block graphs that, for several standard datasets, provides better compression and faster random access than competing schemes.