Skip to content

v0.27.0

Latest
Latest

Choose a tag to compare

View all tags
@rich-iannone rich-iannone released this 13 Aug 19:09
· 1 commit to main since this release
v0.27.0
2767f69
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Pointblank v0.27.0 brings referential integrity validation, statistical drift detection, and expanded data profiling capabilities. The new col_vals_in_table() validation method enables cross-table foreign key checks, while DataScan.compare() and the DataScanDiff class make it possible to detect schema and statistical drift between data profiles over time. Several skills are now packaged with the library for immediate use, contract adapters received important additions and fixes, and the MCP integration has been cleaned up with a leaner dependency footprint.

New Features

  • Referential integrity validation with col_vals_in_table() — New validation method that checks whether each value (or composite key) in a column exists in a reference table column. Supports single and multi-column keys, na_pass, segmentation, and all standard thresholds and actions, making foreign key–style consistency checks a first-class validation step. (#419)

  • Statistical drift measures via DataScanDiff — DataScan.compare(baseline) now returns a DataScanDiff object that provides programmatic access to schema changes (columns added, removed, or type-changed) and per-column statistical drift metrics — Population Stability Index (PSI) for numeric and categorical columns, and Kolmogorov-Smirnov statistic for numerics (with a pure-Python fallback when SciPy is unavailable). Results are surfaced via get_tabular_report(), which renders a styled Great Tables report with schema and drift columns. (#421)

  • Packaged skills — A set of skills is now distributed with the package (e.g., defining contracts, drafting validations, scanning and profiling data, writing validations, and more). (#417)

Enhancements

  • Extended DataScan class — DataScan can now be serialized and restored without access to the original data. New methods include to_dict(), to_json(), save_to_json(), from_dict(), from_json(), and load_from_json() for round-trip persistence of data profiles. Stored scans can then be compared against new scans using compare() to track drift over time. (#420)

  • Contract adapter additions and fixes — The contract adapter system received expanded functionality and correctness fixes across the supported formats (dbt, Frictionless, JSON Schema, ODCS). Adapter import and export coverage was improved, and ContractImport.to_contract() and ContractImport.to_python() output is now more reliable. (#418)

Bug Fixes

  • Removed the direct mcp[cli] dependency and replaced it with an updated fastmcp>=2.14.2 pin, resolving installation conflicts in environments that use the MCP optional extra. (#416)

New Contributors

  • @FBruzzezi made their first contribution in #416.
Assets 2
Loading