Simple perfect hash generator, driven by a YAML config file.
Runs as an EA: reads your key/value map from YAML, builds a perfect hash with PerfectHashByLeo, and emits a ready-to-include .mqh with the tables and lookup function already written.

---

Main Features

• YAML-driven, no code required: you don't write the perfect hash tables by hand — you list your keys (and optionally values) in a YAML file, run the EA once, and it generates the .mqh for you.
• Two swappable hash stages, matching how PerfectHashByLeo works internally (bucket hash + final displacement hash):
  • Bucket hash (hash.type): converts each string key into a ulong before bucketing. Supported: TSN_PHASH_TYPE_FNV1A_64 (FNV-1a 64-bit, with configurable basis/prime) and TSN_PHASH_TYPE_XXHASH_64 (xxHash64, with a configurable seed).
  • Final hash (final_hash): the per-bucket displacement hash, applied to the already-bucketed ulong key plus a seed. Supported: HASH_UL1_MUR_MUR (MurMur-style finalizer), HASH_UL1_SPLITMIX (SplitMix64), HASH_UL1_FMIX (MurmurHash3 fmix64 finalizer), HASH_UL1_FIBBO (Fibonacci hashing).
• Two output modes: either a table of final values (map_use_value: true — the generated function returns your mapped value directly) or a table of indices (map_use_value: false — the function returns a position into your own array, which is useful when the value is too complex to inline).
• Optional binary export: alongside the .mqh, you can dump the seeds/hashes/values/meta as raw .bin files (useful if you want to load the table at runtime instead of compiling it in).
• Generates a complete, working lookup function — both a string-keyed overload and a pre-hashed ulong-keyed overload, so callers who already computed the hash elsewhere can skip re-hashing.

YAML config parameters

Parameter	Description
max_attepms	Maximum displacement attempts (seed values 0, 1, 2, ...) tried per bucket before giving up.
namespace	Namespace the generated tables/function are wrapped in (leave empty/null to skip).
file_name_out	Full path of the .mqh file to generate.
table_prefix	Prefix used for all generated table names (<prefix>_seeds, <prefix>_hashes, <prefix>_values/<prefix>_tindex).
copyright	Text written into the generated file's #property copyright.
link	Text written into the generated file's #property link.
func_name	Name of the generated lookup function.
def_name_table_size	Name of the #define for the final table size.
def_bucket_size_name	Name of the #define for the bucket count.
invalid_value	Value returned by the generated function when a key isn't found (also used to fill unused slots).
comment_funct	If true, the generated function body is wrapped in /* */ (useful when you only want the tables and will write the function yourself).
guard_name	Optional #ifndef/#define include guard name for the generated file.
bin	Optional object with seeds, values, hash and meta file paths — if present, the raw tables are also written as binary files.
hash.type	Bucket hash algorithm: TSN_PHASH_TYPE_FNV1A_64 or TSN_PHASH_TYPE_XXHASH_64.
hash.config.basis / hash.config.prime	FNV-1a offset basis / prime (use def for the standard FNV-1a-64 constants).
hash.config.seed	xxHash64 seed (only used when hash.type is TSN_PHASH_TYPE_XXHASH_64).
final_hash	Final displacement hash: HASH_UL1_MUR_MUR, HASH_UL1_SPLITMIX, HASH_UL1_FMIX or HASH_UL1_FIBBO.
perfect_hash.elements_por_bucket	Target average number of keys per bucket (lower = more buckets, generally faster to find a working seed).
perfect_hash.load_factor	Final table load factor (keys / final table size) — lower leaves more free slots, easing displacement at the cost of memory.
map_use_value	true = generate a values table (function returns the value directly); false = generate an indices table (function returns a position for you to index your own array with).
map	The key set itself: a YAML object (KEY: value) when map_use_value: true, or an array of keys otherwise.

Usage example

# config.yaml
max_attepms: 15000
namespace: TSN
file_name_out: C:\Path\To\Out.mqh
table_prefix: g_table_test
copyright: Copyright 2026, Niquel Mendoza
link: https://www.mql5.com/
func_name: HashCustom
invalid_value: -1

hash:
  type: TSN_PHASH_TYPE_FNV1A_64
  config:
    basis: def
    prime: def

final_hash: HASH_UL1_MUR_MUR

perfect_hash:
  elements_por_bucket: 3
  load_factor: 0.81

map_use_value: true
map: { UNO: 1, DOS: 2, TRES: 3, CUATRO: 4, CINCO: 5 }

Run the EA (EA.mq5) pointing InpYamlFileName to this config, and it generates Out.mqh with the seed/hash/value tables plus a ready HashCustom(const string& key) / HashCustom(const ulong key_hash) function.

---

Repository Structure

SimPHash/
└── Src/   # Generator, hash strategies (Hash/), YAML enum registry (EnumReg/), and test config/output

---

⚠️ License Notice

This repository requires access to two private repositories (YamlParserByLeo, ExtraCodes) to compile. See Requirements above for how to obtain access.

---

Requirements

See dependencies.json for the full list.

---

Installation

cd "C:\Users\YOUR_USER\AppData\Roaming\MetaQuotes\Terminal\YOUR_ID\MQL5\Shared Projects"
tsndep install "https://forge.mql5.io/nique_372/SimPHash.git"

Requires the tsndep package, available on PyPI. It automatically downloads and installs all declared dependencies.

---

Quick Start

1. Download the latest EA from the releases section of the repository: https://forge.mql5.io/nique_372/SimPHash/releases

2. Write your YAML config (see parameters and example above).

3. Point EA.mq5 at it and run it:

input string InpYamlFileName = "C:\\...\\config.yaml";

4. Include the generated .mqh in your project and call the generated function.

---

License

Read Full License By downloading or using this repository, you accept the license terms.

---

Contact

• Platform: MQL5 Community
• Profile: https://www.mql5.com/es/users/nique_372
• Articles: https://www.mql5.com/es/users/nique_372/publications