YugabyteDB Documentation

DocumentDB is a MongoDB-compatible open-source document database built on PostgreSQL. It enables CRUD operations on BSON (Binary JSON) data types within a PostgreSQL framework and supports full-text search, geospatial queries, and vector search. The system consists of three main components:

• pg_documentdb_core: A PostgreSQL extension that introduces BSON datatype support and operations for native Postgres.
• pg_documentdb: The public API surface providing CRUD functionality on documents.
• pg_documentdb_gw: A gateway protocol translation layer that converts MongoDB APIs into PostgreSQL queries.

The hll module introduces a HyperLogLog data structure, a fixed-size, set-like structure used for estimating the count of distinct values with tunable precision. It uses a promotion hierarchy of algorithms to balance accuracy, memory, and performance:

• EMPTY: A constant representing an empty set.
• EXPLICIT: A sorted list of unique integers (exact representation) maintained up to a fixed cardinality.
• SPARSE: A map-based probabilistic implementation that stores only non-zero registers.
• FULL: A fully-materialized, list-based implementation storing every register.

Cardinality estimates for EMPTY and EXPLICIT are exact, while SPARSE and FULL provide probabilistic guarantees.

pgrx is a framework that allows developers to write PostgreSQL extensions using Rust instead of C. It provides safe abstractions to handle complex PostgreSQL requirements, such as:

• Binding Postgres headers and creating loadable dynamic libraries.
• Generating SQL CREATE FUNCTION and CREATE TYPE declarations.
• Managing C language invariants and SQL invariants.
• Handling Postgres-specific invariants like the function argument ABI.

By using Rust, pgrx leverages procedural macros and type-safe abstractions to encode these invariants, reducing the need for expert-level knowledge of Postgres internals, C, and FFI.

Kiwi is a C library designed for handling PostgreSQL protocol-level messages. It provides functionality for both Frontend (client-to-server) and Backend (server-to-client) communication, making it suitable for writing client or server simulation applications.

Key Constraints:

• It does not support networking. It is strictly limited to buffer management and packet validation.
• It is intended to be used in conjunction with the machinarium framework.

pg_hint_plan

pg_hint_plan is an extension that allows you to manually influence PostgreSQL execution plans by using "hints" embedded within SQL comments.

While PostgreSQL uses a cost-based optimizer that estimates the lowest-cost execution plan based on data statistics, the optimizer may not always account for certain data properties (such as correlation between columns). pg_hint_plan provides a mechanism to override these decisions using syntax like /*+ SeqScan(a) */ to ensure specific execution strategies are used.

The documentdb_rum module is a fork of the standard RUM access method designed to support document-centric scenarios.

Key characteristics:

• Compatibility: It maintains the same storage layout and disk content as the standard RUM index, making it fully backwards compatible.
• Implementation: It is based on the GIN access method code and is designed such that changes are restricted to the query path, volatile path, or ways that remain compatible with indexes built with the default RUM repository.