Home GitHub

Architecture

Last updated

This page shows in one overview diagram which layers make up MySQLPool and how they connect.

graph TB
    App[Application] --> Init["init()"]
    Init --> Cfg[Config Resolution<br/>object / env vars]
    Cfg --> RP[(Read Pool)]
    Cfg --> WP[(Write Pool)]
    App --> QB[Chainable Query Builder<br/>static state]
    App --> Raw["read() / write()"]
    QB --> Exec["query() Executor<br/>slow query log"]
    Raw --> Exec
    Exec --> RP
    Exec --> WP
    Sig[SIGINT / SIGTERM] --> Close["close()"]
    Close --> RP
    Close --> WP

Layers

Layer Location Responsibility
Config resolution init(), correctConfig() (src/MySQLPool.ts:40-97) Picks read and write configs by priority, fills defaults, creates and verifies mysql2/promise pools
Query builder table() through increase() (src/MySQLPool.ts:117-227) Accumulates columns, conditions, joins, sorting, and paging in static fields
SQL assembly get(), insert(), update(), upsert() (src/MySQLPool.ts:229-360) Turns the accumulated state into SQL and bindings
Executor query() (src/MySQLPool.ts:362-392) Picks the pool, checks out a connection, runs, times, and releases it
Raw entry points read(), write() (src/MySQLPool.ts:394-406) Run hand-written SQL on a fixed pool
Lifecycle close() and signal handlers (src/MySQLPool.ts:99-115, 409-417) Close the pools; on a signal, close and exit the process

Cross-Cutting Principles

Principle Detail
Single database exit Every SQL statement passes through query(), so slow query logging and connection release live in one place
Values bound, identifiers inlined Data values use ? placeholders; column names, operators, and LIMIT/OFFSET numbers are written into SQL
Process-wide singleton MySQLPool has only static members and a private constructor, so the whole process shares one pair of pools and one builder state

Further Reading

中文