update_treestats#

barecat.Index.update_treestats()[source]#

Recompute size_tree and num_files_tree for all directories.

Uses a bottom-up recursive CTE that expands each file to all its ancestor directories, then aggregates. This is O(files * avg_depth) instead of the naive GLOB/LIKE join which is O(dirs * files).

Performance comparison (10k files, 769 dirs, avg depth ~3):

Method | Time | vs CTE |

|--------|——–|-----------| | GLOB | 2.53s | 73x slower| | LIKE | 0.67s | 19x slower| | CTE | 0.04s | baseline |

On a real dataset (21M files, 341k dirs):

  • CTE took 170s

  • GLOB/LIKE would take hours (O(dirs * files) = 7 trillion comparisons)