precizer

precizer is a high-performance CLI integrity checker built to prove that large file trees are identical, especially after synchronization workflows that rely on metadata rather than full content verification.

At its core, precizer works in two phases: snapshot and compare. In snapshot mode, it recursively traverses a directory tree and builds an SQLite database that records each file under the scanned root (stored as a relative path) along with its computed checksum and related metadata. Storing relative paths means databases captured from different mount points (for example, /mnt1/... versus /mnt2/...) can still be compared consistently.

In compare mode, precizer compares two previously generated databases and produces a deterministic report: files that exist in A but not B (and vice versa), and files that exist in both snapshots but have different checksums.

precizer is designed for long-running verification jobs and large volumes of data. The database acts as persisted state, so runs are fault-tolerant and resumable: the process can be interrupted safely, and subsequent runs can continue using what’s already recorded rather than restarting from scratch. Checksums are computed using SHA-512 to reliably detect even single-byte differences.

For ongoing operation (daily or weekly scans), precizer supports refreshing an existing snapshot database via the --update option. This mode explicitly requires an opt-in flag to prevent accidental destructive refreshes; once enabled, the database is synchronized to the current on-disk state (additions, changes, deletions) and cleaned up as needed.

precizer also supports workflow-oriented controls that matter in real storage environments. It supports selective scope via PCRE2 regex filters: ignore paths with --ignore, override exclusions with --include, and optionally remove ignored entries from the database using --db-drop-ignored. It supports immutable archive protection with --lock-checksum: locked paths are written once and then treated as should-never-change; any later drift is reported and can be used for automation via exit codes. For deeper audits of locked data, --rehash-locked forces a full reread and SHA-512 recomputation for locked entries to perform an explicit integrity sweep