NCA pgBadger count study. All SQL/logs are fictional; no database is started or queried. Requires Python 3 and Perl (executed Perl v5.34.1). From extracted root run python3 scripts/run_parser_study.py. The harness invokes included unmodified official pinned pgBadger 13.2/13.3 sources, builds two synthetic stderr representations and writes four text reports. It makes no network calls. Prefix: %m [%p] %u@%d . No dbuser filtering; one parser job. Expected known ledger: 20 execution IDs, four query shapes with five executions each; single-record input 20 logical entries/30 physical lines; paired input 28 entries/58 physical lines. Actual original totals: 13.2 single20 paired28, 13.3 single20 paired20. Do not claim all environments overcount by40%, use SQL text as execution identity or interpret corrected counts as faster queries. Original code MIT, synthetic data CC0; upstream parser sources retain PostgreSQL license, with both LICENSE files included.
