Skip to content

CHANGELOG.md

Latest commit

 

History

History
221 lines (172 loc) · 10.4 KB

CHANGELOG.md

File metadata and controls

221 lines (172 loc) · 10.4 KB
Raw

Changelog

All notable changes to this project will be documented in this file. The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

Unreleased

0.9.0 - 2026-06-18

Added

  • Added support for passing DB functions into search queries.

0.8.0 - 2026-05-21

Changed

  • BREAKING: Match has been split into two classes: MatchAny and MatchAll.
  • BREAKING: Search modifiers are now composable functions applied to search values or query expressions: Boost(...), Const(...), Fuzzy(...), Slop(...), and Tokenized(...). Query expressions no longer accept modifier kwargs like boost=, const=, distance=, slop=, or tokenizer=.

0.7.0 - 2026-05-05

  • Add equality comparison for Tokenizer types so that repeated makemigrations calls do not repeatedly create the same migration.

0.6.0 - 2026-03-23

Added

  • Added support for icu and edge_ngram tokenizers.

Changed

  • BREAKING: Removed the Empty, ParseWithField, and Range query expressions from the public search API.
  • BREAKING: Updated tokenizers to be configured with a function-based API instead of using strings. Before: "simple('lowercase=false') After: Tokenizer.simple(options={"lowercase": False})

0.5.0 - 2026-03-23

Added

  • IndexExpression class for indexing computed expressions in BM25Index. Supports Django expressions (e.g., Lower('title'), F('rating') + 1) with optional tokenizer configuration. For non-text expressions, uses pdb.alias.

Changed

  • BREAKING: More like this queries now go through the main ParadeDB path like all other queries
  • BREAKING: Proximity queries have a new API that exposes the full SQL api: Proximity("running").within(1, ProxRegex("sho.*"), ordered=True).within(2, "stores").boost(1.2)

0.4.0 - 2026-02-28

Added

  • Match query expression requiring explicit operator='AND' or operator='OR' for text-match queries, replacing implicit bare-string matching. (#37)
  • 6 new query expression types: Empty (match-nothing), Exists (field existence), FuzzyTerm (dedicated fuzzy term via pdb.fuzzy_term()), ParseWithField (field-scoped query parsing via pdb.parse_with_field()), Range (range queries via pdb.range() with typed PostgreSQL ranges), and TermSet (set membership via pdb.term_set() with typed arrays). (#44)
  • Diagnostics helpers: paradedb_indexes(), paradedb_index_segments(), paradedb_verify_index(), paradedb_verify_all_indexes() wrapping ParadeDB pdb.* diagnostic table functions. (#39)
  • Django management commands: paradedb_indexes, paradedb_index_segments, paradedb_verify_index, paradedb_verify_all_indexes (requires "paradedb" in INSTALLED_APPS). (#39)
  • Non-exact facets: exact parameter on Agg and facets() — pass exact=False for approximate facet aggregations. (#39)
  • Partial index support: BM25Index now accepts a condition parameter (Django Q object) to create partial BM25 indexes with a WHERE clause. (#41)
  • Concurrent index creation: BM25Index.create_sql() now correctly forwards the concurrently kwarg so AddIndexConcurrently works. (#41)
  • Top-level package re-exports: all public API symbols are now importable directly from paradedb (e.g., from paradedb import ParadeDB, Match, Score, BM25Index). (#43)
  • Centralised API constants: api.json, api.py, and api.pyi providing a single source of truth for all ParadeDB SQL operators, functions, and types. (#44)
  • Schema compatibility checking: scripts/check_schema_compat.py and apiignore.json for bidirectional validation against ParadeDB's SQL schema. (#44)
  • Schema compatibility CI workflow triggered on each ParadeDB release with failure notifications via GitHub Issues and Slack. (#46)
  • Inline fuzzy parameters (distance, prefix, transposition_cost_one) on Match and Term query expressions. (#40)
  • ProxRegex items (pdb.prox_regex(...)) can now be mixed with plain strings inside ProximityArray. (#48)

Changed

  • BREAKING: ParadeDB('shoes') (bare strings) no longer works for text matching — use ParadeDB(Match('shoes', operator='AND')) instead. (#37)
  • BREAKING: Fuzzy search is no longer a standalone Fuzzy(...) expression — use Match('shoez', operator='OR', distance=1) or Term('shoez', distance=1). (#37, #40)
  • BREAKING: operator kwarg removed from ParadeDB(...) — pass it via Match(...) instead. (#51)
  • JSON field aggregations now use native json_fields index configuration with dot notation (metadata.color) instead of column aliases. (#42)
  • Tokenizer quoting improved to support invocation syntax like whitespace('lowercase=false') without over-quoting. (#40)
  • Tokenizer resolution now uses a validated lookup backed by api.json constants. (#44)
  • All hardcoded ParadeDB SQL strings replaced with constants from paradedb.api. (#44)
  • Match now validates invalid combinations (tokenizer + fuzzy, multi-term fuzzy + boost/const) eagerly at construction time. (#57)
  • TermSet enforces homogeneous element types. (#50)
  • MoreLikeThis validation tightened for key field, fields, stopwords, and numeric options. (#50, #58)
  • README quickstart rewritten to use MockItemDjango model with mock_items dataset. (#36)

Fixed

  • BM25Index.create_sql() now properly passes concurrently kwarg, fixing silent failures with Django's AddIndexConcurrently. (#41)
  • Multi-term fuzzy queries generate correct SQL (ARRAY[...]::pdb.fuzzy(N) instead of invalid per-element casts). (#42, #57)
  • Tokenizer + fuzzy ordering for single-term queries corrected. (#42)
  • api.json loaded from package data when installed (source-tree fallback for dev). (#49)
  • boost/const validated as finite numbers; snippet numeric options validated as non-negative integers. (#49)

Removed

  • BREAKING: PQ class removed — use Django Q objects with Match clauses. (#37, #40)
  • BREAKING: Fuzzy standalone expression removed — use inline fuzzy params on Match and Term. (#37, #40)
  • TERM removed from ParadeOperator literal type (only AND and OR remain). (#37)

0.3.0 - 2026-02-19

Added

  • Proximity query expression with unordered (##) and ordered (##>) forms
  • Match(...) query expression for explicit AND/OR literal matching
  • tokenizer= override support on Match(...) queries
  • tokenizer support on Phrase(...)
  • Scoring modifiers across query expressions via boost and const
  • Extended fuzzy options on Match(...)/Term(...): distance, prefix, transposition_cost_one
  • Parse(..., conjunction_mode=...) passthrough support
  • PhrasePrefix(...) support via pdb.phrase_prefix(...)
  • RegexPhrase(...) support via pdb.regex_phrase(...)
  • ProximityRegex(...) support via pdb.prox_regex(...) within proximity queries
  • ProximityArray(...) support via pdb.prox_array(...) within proximity queries
  • RangeTerm(...) support via pdb.range_term(...)
  • ParadeDB operators over ad-hoc SQL expressions

Changed

  • Added validation for incompatible argument combinations:
    • operator on non-string ParadeDB(...) terms is rejected
    • mixed fuzzy operators in one query are rejected
    • multiple Proximity(...) terms in one query are rejected (use ProximityArray(...))
  • Removed wrapper-side scoring validation for boost/const; scoring options are forwarded to ParadeDB/PostgreSQL.
  • Compatibility and packaging metadata were updated for broader support: Django 4.2+, Python 3.10+, and refreshed project metadata/dependencies.

0.2.0 - 2026-02-13

Added

  • Snippets annotation wrapper for pdb.snippets(...), including support for custom tags, max_num_chars, limit/offset, and sort_by (score or position).
  • SnippetPositions annotation wrapper for pdb.snippet_positions(...) to return byte-offset ranges for matched terms.
  • Expanded BM25Index tokenizer DSL with: tokenizers (multiple tokenizers per field), positional args, named_args, and alias-aware configuration.

Changed

  • BM25Index now enforces explicit tokenizer declarations when tokenizer config keys are provided (filters, stemmer, args, named_args, alias), and requires explicit tokenizers for json_keys.
  • Deprecated tokenizer options key was removed in favor of named_args.
  • Hybrid RRF example was refactored to a single ORM query approach (no Python post-join).
  • Project quality tooling moved from pre-commit to prek, with CI updates for linting/test matrix and Codecov reporting.

Fixed

  • Validation and error reporting for invalid tokenizer config shapes (mixed single/multi-tokenizer configuration and malformed tokenizers lists).
  • SQL generation coverage for tokenizer configuration and snippet functions.

0.1.1 - 2026-02-04

Fixed

  • Release automation was corrected to publish to PyPI from the release workflow path instead of relying on tag-push behavior.

0.1.0 - 2025-01-30

Added

  • BM25Index for declarative BM25 index creation in Django models
  • ParadeDB lookup wrapper with &&&, |||, ### operators
  • Boolean composition support via Django Q objects
  • Search expressions: Phrase, FuzzyTerm, Parse, Term, Regex, All
  • MoreLikeThis query filter for similarity search
  • Score annotation for BM25 relevance scores
  • Snippet annotation for highlighted text excerpts
  • ParadeDBQuerySet with .facets() method for aggregations
  • ParadeDBManager for easy model integration
  • JSON field key indexing support
  • Full Django ORM integration with Q objects and standard filters