Contents

• Module: Hash
  • Files
  • Responsibilities
    • 1. Hash Algorithm
    • 2. Per-Register Hash Functions
    • 3. Type Serialization
    • 4. NULL Handling
  • SQL Sources
  • Related Tests

Module: Hash

Purpose: Computes dim_hash — a 64-bit hash of the dimension value combination, used as the primary lookup key in totals and balance_cache tables.

Files

• hash.c — Hash computation, per-register hash function generation

Responsibilities

1. Hash Algorithm

• Default: xxhash64 (fast, low collision rate)
• Alternative: murmur3 (selectable via pg_accumulator.hash_function GUC)
• 64-bit output — collision probability below 1 in 10¹⁸ for realistic dimension counts

2. Per-Register Hash Functions

Auto-generates _hash_<register>(dim1, dim2, ...) functions for each register. These functions are declared as IMMUTABLE STRICT and implemented in C for maximum performance.

3. Type Serialization

Each dimension type is serialized into a byte sequence before hashing:

Type	Serialization
int, bigint, smallint	Native binary representation
text, varchar	UTF-8 bytes
uuid	16 raw bytes
date	Julian day number
boolean	1 byte (0 or 1)

4. NULL Handling

NULL dimension values are serialized as a distinct sentinel byte, ensuring that (warehouse=1, lot=NULL) and (warehouse=1, lot='') produce different hashes.

SQL Sources

• sql/02_hash.sql — Hash function definitions

Related Tests

• test/sql/13_multiple_dimensions.sql — Multi-dimension hashing, NULL handling, collision avoidance