timescale/pg_textsearch

Features

• Full Text Search - Provides high-performance full-text search integration for PostgreSQL with BM25 relevance ranking.
• Segment Spilling - Uses a memtable system to buffer index updates and spill them to immutable disk segments to handle massive datasets.
• BM25 Full-Text Indices - Implements specialized indices using the BM25 ranking function for efficient keyword-based retrieval from text columns.
• Memtable-Based Segment Spilling - Builds scalable indices using a memtable architecture that spills to disk segments to handle massive datasets.
• Log-Structured Merge-Trees - Implements a log-structured merge architecture to combine index segments, reducing scan overhead and improving query performance.
• Large Scale Indexing - Builds and maintains search indexes for massive datasets using parallel workers and memory-to-disk spilling.
• Multilingual Search Engines - Provides built-in support for multilingual search through language-specific tokenization, stemming, and stop-word removal.
• BM25+ Scoring - Calculates BM25 scores using term frequency and inverse document frequency to return the most relevant documents.
• Document Chunking Strategies - Splits documents exceeding token limits into smaller pieces to ensure consistent tokenization and ranking.
• Memory-Efficient Chunking - Splits oversized text bodies into smaller pieces during tokenization to maintain memory efficiency and consistency.
• WAL-Based Standby Replication - Utilizes the PostgreSQL write-ahead log to synchronize search index data across primary and standby nodes.
• Partition-Local Statistics - Maintains term frequencies and document counts per partition to ensure accurate BM25 scoring across distributed data.
• Database Performance Tuning - Optimizes query speeds through segment consolidation, memory management, and skipping irrelevant data blocks.
• Expression Indexes - Supports indexing computed results from stable functions to enable search on JSONB fields and concatenated columns.
• Asynchronous Segment Merging - Merges multiple internal index segments into a single segment to reduce scan overhead and increase query speed.
• Index Memory Management - Controls shared memory usage for in-memory caches and manages thresholds for data spilling to disk.
• Language-Specific Partial Indexing - Creates smaller partial indexes for subsets of data to apply unique language configurations for multilingual tables.
• Parallel Construction - Accelerates index creation for large tables using parallel worker processes based on available system memory.
• Top-k Retrieval Optimization - Improves retrieval speed by skipping irrelevant data blocks when queries include order-by clauses and limits.
• Cascading Replication - Synchronizes search data across nodes to support cascading replication and point-in-time recovery.
• Search Index Recovery - Automatically restores search index integrity and availability after crashes or unexpected shutdowns.
• Index Memory Management - Manages search indexes with configurable memory budgets and per-index limits to optimize system performance.
• Search Result Filtering - Combines search scoring with pre-filtering via separate indexes or post-filtering for non-indexed columns.
• Local Indexes - Maintains partition-local statistics for document counts and term frequencies to ensure accurate scoring.
• Build Acceleration - Accelerates the creation of indexes for large tables and partitions using parallel worker processes.
• Bitset-Based Deletion Tracking - Uses per-segment bitsets to prune deleted documents during vacuuming without requiring expensive full index rebuilds.
• Database High Availability - Ensures data availability across primary and standby nodes through index replication and automatic recovery.

Istoric stele