Skip to content

Changelog

All notable changes to additory. Format based on Keep a Changelog.


0.1.3a11 — 2026-04-29

Added

  • Scan Diff & Reconciliation — add.scan('@diff', ...) for DataFrame comparison
    • Summary mode: inline "old → new" change markers with _diff_status column
    • Detail mode: one row per changed cell with _key, _column, _old_value, _new_value
    • Auto key detection: finds unique single or two-column composite keys
    • Duplicate handling: collapses identical duplicates, flags non-identical
    • Reconciliation via .add files: aliases and groups
    • exclude and carry columns in strategy dict
  • Pipe-Friendly Dynamic API
    • add.to() pipe compatibility (df.pipe(add.to, ...))
    • add.<dynamic>() API — module-level __getattr__ resolution for expression-based column operations
    • Column auto-matching, explicit mapping with **kwargs
    • add.scan('@set', ...) — runtime expression loading
  • Dynamic Expressions — Unified TOML Format
    • Unified .add file format — each top-level TOML table is one expression
    • Added category field to expressions
    • Rust expression parser, formatter, scanner, and config system
  • Rust Migration (partial)
    • Rust core for diff types, strategy parsing, key detection, duplicate handling
    • Config types, parser, and three-tier resolver
    • CLI crate (additory-cli) with subcommands: to, transform, synthetic, diff, config, expressions

Changed

  • add.synthetic() signature refactor — DataFrame-first, pipe-friendly
    • First argument is now df_or_mode: Union[DataFrame, str]
    • Mode inferred from type: DataFrame → augment, '@new' → new
    • Old mode= kwarg raises TypeError with migration message
  • Expression data model — removed sha, format, requires fields
  • Inbuilt expression files trimmed — deleted advanced.add and medical_extended.add

Fixed

  • 45 Rust test compilation errors fixed
  • 12 compiler warnings eliminated
  • Invalid regex escape sequences in expression safe pattern
  • Synthetic test parameter structure updated

Performance

  • 404 Rust tests + 157 Python tests = 561 total tests passing
  • Zero compiler warnings on cargo build --lib

0.1.3 — 2026-03-09

Added

  • Lineage Tracking — lineage=True parameter on all core functions
    • Session-only lineage metadata stored in DataFrame native format
    • View lineage reports with add.scan('@lineage', df)
    • Track column sources, row mappings, and operation history
    • Mutual exclusion validation with as_type parameter
  • add.scan() Function — unified scanning interface
    • add.scan('@analyze', df) — statistical analysis and data quality
    • add.scan('@lineage', df) — view lineage tracking reports
    • Replaces standalone add.analyze() and add.analyse()

Changed

  • API Consolidation — add.analyze() → add.scan('@analyze')
  • Parameter Order Convention — standardized: logging, lineage, as_type
  • Code Distribution — ~95% Rust, ~5% Python

Fixed

  • add.to() parameter validation bug (wrong parameters sent to Rust)
  • String-to-list conversion for bring and against parameters
  • 11 compiler warnings fixed with cargo fix

Performance

  • Lineage overhead: <3ms per operation
  • All 341 tests passing

0.1.3a10 — 2026-02-10

Added

  • Pure Rust KNN Imputation — no scikit-learn dependency
    • Multiple distance metrics (Euclidean, Manhattan, Cosine)
    • Uniform and distance-weighted averaging
    • 2x+ performance improvement over Python
  • Pure Rust Label Deduction — TF-IDF based
    • Unigrams and bigrams, cosine similarity
    • 2x+ performance improvement over Python
  • Python Synthetic Wrapper — PyO3 bindings for all 10 strategies

Removed

  • scikit-learn dependency — 50+ MB package size reduction

Performance

Metric Result
KNN imputation (1000×10) <500ms
Label deduction (1000×5) <1000ms
Synthetic generation (10k rows) <200ms
Package size reduction 50+ MB

0.1.3a6 — 2026-02-13

Fixed

  • Added pyarrow as dependency for pandas users

0.1.3a5 — 2026-02-13

Added

  • Complete PyO3 bindings for Rust-Python integration
  • add.to() — lookup, multi-column fetch, aggregation, position control
  • add.transform() — all 12 modes: @calc, @filter, @sort, @aggregate, @transpose, @split, @extract, @onehot, @label, @harmonize, @deduce, @round
  • add.synthetic() — @new mode (7 distributions, patterns, sequences) and augment mode
  • Polars and Pandas support

Changed

  • Renamed by parameter to against in add.to()

Performance

  • 3-5x faster transformations vs pure Python
  • 5-10x faster data joining operations
  • 10-20x faster synthetic data generation

Version History

Version Date Highlights
0.1.3a11 2026-04-29 Diff, dynamic API, pipe compatibility, reconciliation
0.1.3 2026-03-09 Lineage tracking, add.scan(), 95% Rust
0.1.3a10 2026-02-10 Pure Rust KNN/TF-IDF, no scikit-learn
0.1.3a6 2026-02-13 pyarrow dependency fix
0.1.3a5 2026-02-13 First beta with complete PyO3 bindings
0.1.3a4 2026-02-12 Alpha with partial Rust
0.1.3a3 2026-02-11 Synthetic data features
0.1.3a2 2026-02-10 Transform modes
0.1.3a1 2026-02-09 Initial alpha release