https://github.com/forge-sql-orm/forge-sql-orm
Seamlessly integrate Drizzle ORM with Forge-SQL to enable type-safe database operations in your Atlassian Forge applications. Includes a custom driver, schema migration support, two levels of caching (local in-memory and global via @forge/kvs), optimistic locking, query analysis, and more.
https://github.com/forge-sql-orm/forge-sql-orm
atlassian-forge drizzle drizzle-framework drizzle-mysql2 drizzle-orm forge forge-app forge-sql forge-sql-orm migration-tool orm orm-framework orm-library rovo sql
Last synced: 3 months ago
JSON representation
Seamlessly integrate Drizzle ORM with Forge-SQL to enable type-safe database operations in your Atlassian Forge applications. Includes a custom driver, schema migration support, two levels of caching (local in-memory and global via @forge/kvs), optimistic locking, query analysis, and more.
- Host: GitHub
- URL: https://github.com/forge-sql-orm/forge-sql-orm
- Owner: forge-sql-orm
- License: mit
- Created: 2025-02-28T10:41:27.000Z (over 1 year ago)
- Default Branch: master
- Last Pushed: 2026-01-16T19:18:10.000Z (5 months ago)
- Last Synced: 2026-01-18T08:31:00.169Z (5 months ago)
- Topics: atlassian-forge, drizzle, drizzle-framework, drizzle-mysql2, drizzle-orm, forge, forge-app, forge-sql, forge-sql-orm, migration-tool, orm, orm-framework, orm-library, rovo, sql
- Language: TypeScript
- Homepage:
- Size: 30.4 MB
- Stars: 15
- Watchers: 1
- Forks: 0
- Open Issues: 3
-
Metadata Files:
- Readme: README.md
- Contributing: CONTRIBUTING.md
- Funding: .github/FUNDING.yml
- License: LICENSE
- Code of conduct: CODE_OF_CONDUCT.md
- Security: SECURITY.md
Awesome Lists containing this project
README
# Forge SQL ORM
[](https://www.npmjs.com/package/forge-sql-orm)
[](https://www.npmjs.com/package/forge-sql-orm)
[](https://www.npmjs.com/package/forge-sql-orm-cli)
[](https://www.npmjs.com/package/forge-sql-orm-cli)
[](https://sonarcloud.io/summary/new_code?id=forge-sql-orm_forge-sql-orm)
[](https://github.com/forge-sql-orm/forge-sql-orm/blob/master/LICENSE)
[](https://github.com/forge-sql-orm/forge-sql-orm/actions/workflows/node.js.yml)
[](https://deepscan.io/dashboard#view=project&tid=26652&pid=30920&bid=997203)
[](https://snyk.io/test/github/forge-sql-orm/forge-sql-orm)
[](https://sonarcloud.io/summary/new_code?id=forge-sql-orm_forge-sql-orm)
[](https://sonarcloud.io/summary/new_code?id=forge-sql-orm_forge-sql-orm)
[](https://sonarcloud.io/summary/new_code?id=forge-sql-orm_forge-sql-orm)
[](https://sonarcloud.io/summary/new_code?id=forge-sql-orm_forge-sql-orm)
[](https://sonarcloud.io/summary/new_code?id=forge-sql-orm_forge-sql-orm)
[](https://sonarcloud.io/summary/new_code?id=forge-sql-orm_forge-sql-orm)
[](https://qlty.sh/gh/forge-sql-orm/projects/forge-sql-orm)
**Forge-SQL-ORM** is an ORM designed for working with [@forge/sql](https://developer.atlassian.com/platform/forge/storage-reference/sql-tutorial/) in **Atlassian Forge**. It is built on top of [Drizzle ORM](https://orm.drizzle.team) and provides advanced capabilities for working with relational databases inside Forge.
## Key Features
- ✅ **Custom Drizzle Driver** for direct integration with @forge/sql
- ✅ **Local Cache System (Level 1)** for in-memory query optimization within single resolver invocation scope
- ✅ **Global Cache System (Level 2)** with cross-invocation caching, automatic cache invalidation and context-aware operations (using [@forge/kvs](https://developer.atlassian.com/platform/forge/storage-reference/storage-api-custom-entities/) )
- ✅ **Performance Monitoring**: Query execution metrics and analysis capabilities with automatic error analysis for timeout and OOM errors, scheduled slow query monitoring with execution plans, and async query degradation analysis for non-blocking performance monitoring
- ✅ **Type-Safe Query Building**: Write SQL queries with full TypeScript support
- ✅ **Supports complex SQL queries** with joins and filtering using Drizzle ORM
- ✅ **Advanced Query Methods**: `selectFrom()`, `selectDistinctFrom()`, `selectCacheableFrom()`, `selectDistinctCacheableFrom()` for all-column queries with field aliasing
- ✅ **Query Execution with Metadata**: `executeWithMetadata()` method for capturing detailed execution metrics including database execution time, response size, and query analysis capabilities with performance monitoring. Supports two modes for query plan printing: TopSlowest mode (default) and SummaryTable mode
- ✅ **Raw SQL Execution**: `execute()`, `executeCacheable()`, `executeDDL()`, and `executeDDLActions()` methods for direct SQL queries with local and global caching
- ✅ **Common Table Expressions (CTEs)**: `with()` method for complex queries with subqueries
- ✅ **Schema migration support**, allowing automatic schema evolution
- ✅ **Automatic entity generation** from MySQL/tidb databases
- ✅ **Automatic migration generation** from MySQL/tidb databases
- ✅ **Drop Migrations** Generate a migration to drop all tables and clear migrations history for subsequent schema recreation
- ✅ **Schema Fetching** Development-only web trigger to retrieve current database schema and generate SQL statements for schema recreation
- ✅ **Ready-to-use Migration Triggers** Built-in web triggers for applying migrations, dropping tables (development-only), and fetching schema (development-only) with proper error handling and security controls
- ✅ **Optimistic Locking** Ensures data consistency by preventing conflicts when multiple users update the same record
- ✅ **Query Plan Analysis**: Detailed execution plan analysis and optimization insights
- ✅ **Rovo Integration** Secure pattern for natural-language analytics with comprehensive security validations, Row-Level Security (RLS) support, and dynamic SQL query execution
## Table of Contents
### 🚀 Getting Started
- [Key Features](#key-features)
- [Usage Approaches](#usage-approaches)
- [Installation](#installation)
- [CLI Commands](#cli-commands) | [CLI Documentation](forge-sql-orm-cli/README.md)
- [Quick Start](#quick-start)
### 📖 Core Features
- [Field Name Collision Prevention](#field-name-collision-prevention-in-complex-queries)
- [Drizzle Usage with forge-sql-orm](#drizzle-usage-with-forge-sql-orm)
- [Direct Drizzle Usage with Custom Driver](#direct-drizzle-usage-with-custom-driver)
### 🗄️ Database Operations
- [Fetch Data](#fetch-data)
- [Modify Operations](#modify-operations)
- [SQL Utilities](#sql-utilities)
### ⚡ Caching System
- [Setting Up Caching with @forge/kvs](#setting-up-caching-with-forgekvs-optional)
- [Global Cache System (Level 2)](#global-cache-system-level-2)
- [Cache Context Operations](#cache-context-operations)
- [Local Cache Operations (Level 1)](#local-cache-operations-level-1)
- [Cache-Aware Query Operations](#cache-aware-query-operations)
- [Manual Cache Management](#manual-cache-management)
### 🔒 Advanced Features
- [Optimistic Locking](#optimistic-locking)
- [Rovo Integration](#rovo-integration) - Secure pattern for natural-language analytics with dynamic SQL queries
- [Query Analysis and Performance Optimization](#query-analysis-and-performance-optimization)
- [Automatic Error Analysis](#automatic-error-analysis) - Automatic timeout and OOM error detection with execution plans
- [Slow Query Monitoring](#slow-query-monitoring) - Scheduled monitoring of slow queries with execution plans
- [Date and Time Types](#date-and-time-types)
### 🛠️ Development Tools
- [CLI Commands](#cli-commands) | [CLI Documentation](forge-sql-orm-cli/README.md)
- [Web Triggers for Migrations](#web-triggers-for-migrations)
- [Step-by-Step Migration Workflow](#step-by-step-migration-workflow)
- [Drop Migrations](#drop-migrations)
### 📚 Examples
- [Simple Example](examples/forge-sql-orm-example-simple)
- [Drizzle Driver Example](examples/forge-sql-orm-example-drizzle-driver-simple)
- [Optimistic Locking Example](examples/forge-sql-orm-example-optimistic-locking)
- [Dynamic Queries Example](examples/forge-sql-orm-example-dynamic)
- [Query Analysis Example](examples/forge-sql-orm-example-query-analyses)
- [Organization Tracker Example](examples/forge-sql-orm-example-org-tracker)
- [Checklist Example](examples/forge-sql-orm-example-checklist)
- [Cache Example](examples/forge-sql-orm-example-cache) - Advanced caching capabilities with performance monitoring
- [Rovo Integration Example](https://github.com/vzakharchenko/Forge-Secure-Notes-for-Jira) - Real-world Rovo AI agent implementation with secure natural-language analytics
### 📚 Reference
- [ForgeSqlOrmOptions](#forgesqlormoptions)
- [Migration Guide](#migration-guide)
## 🚀 Quick Navigation
**New to Forge-SQL-ORM?** Start here:
- [Quick Start](#quick-start) - Get up and running in 5 minutes
- [Installation](#installation) - Complete setup guide
- [Basic Usage Examples](#fetch-data) - Simple query examples
**Looking for specific features?**
- [Global Cache System (Level 2)](#global-cache-system-level-2) - Cross-invocation persistent caching
- [Local Cache System (Level 1)](#local-cache-operations-level-1) - In-memory invocation caching
- [Optimistic Locking](#optimistic-locking) - Data consistency
- [Rovo Integration](#rovo-integration) - Secure natural-language analytics
- [Migration Tools](#web-triggers-for-migrations) - Database migrations
- [Query Analysis](#query-analysis-and-performance-optimization) - Performance optimization
**Looking for practical examples?**
- [Simple Example](examples/forge-sql-orm-example-simple) - Basic ORM usage
- [Optimistic Locking Example](examples/forge-sql-orm-example-optimistic-locking) - Real-world conflict handling
- [Organization Tracker Example](examples/forge-sql-orm-example-org-tracker) - Complex relationships
- [Checklist Example](examples/forge-sql-orm-example-checklist) - Jira integration
- [Cache Example](examples/forge-sql-orm-example-cache) - Advanced caching capabilities
- [Rovo Integration Example](https://github.com/vzakharchenko/Forge-Secure-Notes-for-Jira) - Real-world Rovo AI agent with secure analytics
## Usage Approaches
### 1. Full Forge-SQL-ORM Usage
```typescript
import ForgeSQL from "forge-sql-orm";
const forgeSQL = new ForgeSQL();
```
Best for: Advanced features like optimistic locking, automatic versioning, and automatic field name collision prevention in complex queries.
### 2. Direct Drizzle Usage
```typescript
import { drizzle } from "drizzle-orm/mysql-proxy";
import { forgeDriver } from "forge-sql-orm";
const db = drizzle(forgeDriver);
```
Best for: Simple Modify operations without optimistic locking. Note that you need to manually patch drizzle `patchDbWithSelectAliased` for select fields to prevent field name collisions in Atlassian Forge SQL.
### 3. Local Cache Optimization
```typescript
import ForgeSQL from "forge-sql-orm";
const forgeSQL = new ForgeSQL();
// Optimize repeated queries within a single invocation
await forgeSQL.executeWithLocalContext(async () => {
// Multiple queries here will benefit from local caching
const users = await forgeSQL
.select({ id: users.id, name: users.name })
.from(users)
.where(eq(users.active, true));
// This query will use local cache (no database call)
const cachedUsers = await forgeSQL
.select({ id: users.id, name: users.name })
.from(users)
.where(eq(users.active, true));
// Using new methods for better performance
const usersFrom = await forgeSQL.selectFrom(users).where(eq(users.active, true));
// This will use local cache (no database call)
const cachedUsersFrom = await forgeSQL.selectFrom(users).where(eq(users.active, true));
// Raw SQL with local caching
const rawUsers = await forgeSQL.execute("SELECT id, name FROM users WHERE active = ?", [true]);
});
```
Best for: Performance optimization of repeated queries within resolvers or single invocation contexts.
## Field Name Collision Prevention in Complex Queries
When working with complex queries involving multiple tables (joins, inner joins, etc.), Atlassian Forge SQL has a specific behavior where fields with the same name from different tables get collapsed into a single field with a null value. This is not a Drizzle ORM issue but rather a characteristic of Atlassian Forge SQL's behavior.
Forge-SQL-ORM provides two ways to handle this:
### Using Forge-SQL-ORM
```typescript
import ForgeSQL from "forge-sql-orm";
const forgeSQL = new ForgeSQL();
// Automatic field name collision prevention
await forgeSQL
.select({ user: users, order: orders })
.from(orders)
.innerJoin(users, eq(orders.userId, users.id));
```
### Using Direct Drizzle
```typescript
import { drizzle } from "drizzle-orm/mysql-proxy";
import { forgeDriver, patchDbWithSelectAliased } from "forge-sql-orm";
const db = patchDbWithSelectAliased(drizzle(forgeDriver));
// Manual field name collision prevention
await db
.selectAliased({ user: users, order: orders })
.from(orders)
.innerJoin(users, eq(orders.userId, users.id));
```
### Important Notes
- This is a specific behavior of Atlassian Forge SQL, not Drizzle ORM
- For complex queries involving multiple tables, it's recommended to always specify select fields and avoid using `select()` without field selection
- The solution automatically creates unique aliases for each field by prefixing them with the table name
- This ensures that fields with the same name from different tables remain distinct in the query results
## Installation
Forge-SQL-ORM is designed to work with @forge/sql and requires some additional setup to ensure compatibility within Atlassian Forge.
✅ Step 1: Install Dependencies
**Basic installation (without caching):**
```sh
npm install forge-sql-orm @forge/sql drizzle-orm -S
```
**With caching support:**
```sh
npm install forge-sql-orm @forge/sql @forge/kvs drizzle-orm -S
```
**⚠️ Important for UI-Kit projects:**
If you're installing `forge-sql-orm` in a UI-Kit project (projects using `@forge/react`), you may encounter peer dependency conflicts with `@types/react`. This is due to a conflict between `@types/react@18` (required by `@forge/react`) and `@types/react@19` (optional peer dependency from `drizzle-orm` via `bun-types`).
To resolve this, use the `--legacy-peer-deps` flag:
```sh
# Basic installation for UI-Kit projects
npm install forge-sql-orm @forge/sql drizzle-orm -S --legacy-peer-deps
# With caching support for UI-Kit projects
npm install forge-sql-orm @forge/sql @forge/kvs drizzle-orm -S --legacy-peer-deps
```
**Note:** The `--legacy-peer-deps` flag tells npm to ignore peer dependency conflicts. This is safe in this case because `bun-types` is an optional peer dependency and doesn't affect the functionality of `forge-sql-orm` in Forge environments.
This will:
- Install Forge-SQL-ORM (the ORM for @forge/sql)
- Install @forge/sql, the Forge database layer
- Install @forge/kvs, the Forge Key-Value Store for caching (optional, only needed for caching features)
- Install Drizzle ORM and its MySQL driver
- Install TypeScript types for MySQL
- Install forge-sql-orm-cli A command-line interface tool for managing Atlassian Forge SQL migrations and model generation with Drizzle ORM integration.
## Quick Start
### 1. Basic Setup
```typescript
import ForgeSQL from "forge-sql-orm";
// Initialize ForgeSQL
const forgeSQL = new ForgeSQL();
// Simple query
const users = await forgeSQL.select().from(users);
```
### 2. With Caching (Optional)
```typescript
import ForgeSQL from "forge-sql-orm";
// Initialize with caching
const forgeSQL = new ForgeSQL({
cacheEntityName: "cache",
cacheTTL: 300,
});
// Cached query
const users = await forgeSQL
.selectCacheable({ id: users.id, name: users.name })
.from(users)
.where(eq(users.active, true));
```
### 3. Local Cache Optimization
```typescript
// Optimize repeated queries within a single invocation
await forgeSQL.executeWithLocalContext(async () => {
const users = await forgeSQL
.select({ id: users.id, name: users.name })
.from(users)
.where(eq(users.active, true));
// This query will use local cache (no database call)
const cachedUsers = await forgeSQL
.select({ id: users.id, name: users.name })
.from(users)
.where(eq(users.active, true));
// Using new methods for better performance
const usersFrom = await forgeSQL.selectFrom(users).where(eq(users.active, true));
// Raw SQL with local caching
const rawUsers = await forgeSQL.execute("SELECT id, name FROM users WHERE active = ?", [true]);
});
```
### 4. Resolver Performance Monitoring
```typescript
// Resolver with performance monitoring
resolver.define("fetch", async (req: Request) => {
try {
return await forgeSQL.executeWithMetadata(
async () => {
// Resolver logic with multiple queries
const users = await forgeSQL.selectFrom(demoUsers);
const orders = await forgeSQL
.selectFrom(demoOrders)
.where(eq(demoOrders.userId, demoUsers.id));
return { users, orders };
},
async (totalDbExecutionTime, totalResponseSize, printQueriesWithPlan) => {
const threshold = 500; // ms baseline for this resolver
if (totalDbExecutionTime > threshold * 1.5) {
console.warn(
`[Performance Warning fetch] Resolver exceeded DB time: ${totalDbExecutionTime} ms`,
);
await printQueriesWithPlan(); // Optionally log or capture diagnostics for further analysis
} else if (totalDbExecutionTime > threshold) {
console.debug(`[Performance Debug fetch] High DB time: ${totalDbExecutionTime} ms`);
}
},
{
// Optional: Configure query plan printing behavior
mode: "TopSlowest", // Print top slowest queries (default)
topQueries: 3, // Print top 3 slowest queries
},
);
} catch (e) {
const error = e?.cause?.debug?.sqlMessage ?? e?.cause;
console.error(error, e);
throw error;
}
});
```
**Query Plan Printing Options:**
The `printQueriesWithPlan` function supports two modes:
1. **TopSlowest Mode (default)**: Prints execution plans for the slowest queries from the current resolver invocation
- `mode`: Set to `'TopSlowest'` (default)
- `topQueries`: Number of top slowest queries to analyze (default: 1)
2. **SummaryTable Mode**: Uses `CLUSTER_STATEMENTS_SUMMARY` for query analysis
- `mode`: Set to `'SummaryTable'`
- `summaryTableWindowTime`: Time window in milliseconds (default: 15000ms)
- Only works if queries are executed within the specified time window
### 5. Rovo Integration (Secure Analytics)
```typescript
// Secure dynamic SQL queries for natural-language analytics
const rovo = forgeSQL.rovo();
const settings = await rovo
.rovoSettingBuilder(usersTable, accountId)
.addContextParameter(":currentUserId", accountId)
.useRLS()
.addRlsColumn(usersTable.id)
.addRlsWherePart((alias) => `${alias}.${usersTable.id.name} = '${accountId}'`)
.finish()
.build();
const result = await rovo.dynamicIsolatedQuery(
"SELECT id, name FROM users WHERE status = 'active' AND userId = :currentUserId",
settings,
);
```
### 6. Next Steps
- [Full Installation Guide](#installation) - Complete setup instructions
- [Core Features](#core-features) - Learn about key capabilities
- [Global Cache System (Level 2)](#global-cache-system-level-2) - Cross-invocation caching features
- [Local Cache System (Level 1)](#local-cache-operations-level-1) - In-memory caching features
- [Rovo Integration](#rovo-integration) - Secure natural-language analytics
- [API Reference](#reference) - Complete API documentation
## Drizzle Usage with forge-sql-orm
If you prefer to use Drizzle ORM with the additional features of Forge-SQL-ORM (like optimistic locking and caching), you can use the enhanced API:
```typescript
import ForgeSQL from "forge-sql-orm";
const forgeSQL = new ForgeSQL();
// Versioned operations with cache management (recommended)
await forgeSQL.modifyWithVersioningAndEvictCache().insert(Users, [userData]);
await forgeSQL.modifyWithVersioningAndEvictCache().updateById(updateData, Users);
// Versioned operations without cache management
await forgeSQL.modifyWithVersioning().insert(Users, [userData]);
await forgeSQL.modifyWithVersioning().updateById(updateData, Users);
// Non-versioned operations with cache management
await forgeSQL.insertAndEvictCache(Users).values(userData);
await forgeSQL.updateAndEvictCache(Users).set(updateData).where(eq(Users.id, 1));
// Basic Drizzle operations (cache context aware)
await forgeSQL.insert(Users).values(userData);
await forgeSQL.update(Users).set(updateData).where(eq(Users.id, 1));
// Direct Drizzle access
const db = forgeSQL.getDrizzleQueryBuilder();
const users = await db.select().from(users);
// Using new methods for enhanced functionality
const usersFrom = await forgeSQL.selectFrom(users).where(eq(users.active, true));
const usersDistinct = await forgeSQL.selectDistinctFrom(users).where(eq(users.active, true));
const usersCacheable = await forgeSQL.selectCacheableFrom(users).where(eq(users.active, true));
// Raw SQL execution
const rawUsers = await forgeSQL.execute("SELECT * FROM users WHERE active = ?", [true]);
// Raw SQL with caching
// ⚠️ IMPORTANT: When using executeCacheable(), all table names must be wrapped with backticks (`)
const cachedRawUsers = await forgeSQL.executeCacheable(
"SELECT * FROM `users` WHERE active = ?",
[true],
300,
);
// Raw SQL with execution metadata and performance monitoring
const usersWithMetadata = await forgeSQL.executeWithMetadata(
async () => {
const users = await forgeSQL.selectFrom(usersTable);
const orders = await forgeSQL
.selectFrom(ordersTable)
.where(eq(ordersTable.userId, usersTable.id));
return { users, orders };
},
(totalDbExecutionTime, totalResponseSize, printQueriesWithPlan) => {
const threshold = 500; // ms baseline for this resolver
if (totalDbExecutionTime > threshold * 1.5) {
console.warn(`[Performance Warning] Resolver exceeded DB time: ${totalDbExecutionTime} ms`);
await printQueriesWithPlan(); // Analyze and print query execution plans
} else if (totalDbExecutionTime > threshold) {
console.debug(`[Performance Debug] High DB time: ${totalDbExecutionTime} ms`);
}
console.log(`DB response size: ${totalResponseSize} bytes`);
},
{
// Optional: Configure query plan printing
mode: "TopSlowest", // Print top slowest queries (default)
topQueries: 2, // Print top 2 slowest queries
},
);
// DDL operations for schema modifications
await forgeSQL.executeDDL(`
CREATE TABLE users (
id INT PRIMARY KEY AUTO_INCREMENT,
name VARCHAR(255) NOT NULL,
email VARCHAR(255) UNIQUE
)
`);
// Execute regular SQL queries in DDL context for performance monitoring
await forgeSQL.executeDDLActions(async () => {
// Execute regular SQL queries in DDL context for monitoring
const slowQueries = await forgeSQL.execute(`
SELECT * FROM INFORMATION_SCHEMA.STATEMENTS_SUMMARY
WHERE AVG_LATENCY > 1000000
`);
// Execute complex analysis queries in DDL context
const performanceData = await forgeSQL.execute(`
SELECT * FROM INFORMATION_SCHEMA.CLUSTER_STATEMENTS_SUMMARY_HISTORY
WHERE SUMMARY_END_TIME > DATE_SUB(NOW(), INTERVAL 1 HOUR)
`);
return { slowQueries, performanceData };
});
// Common Table Expressions (CTEs)
const userStats = await forgeSQL
.with(
forgeSQL.selectFrom(users).where(eq(users.active, true)).as("activeUsers"),
forgeSQL.selectFrom(orders).where(eq(orders.status, "completed")).as("completedOrders"),
)
.select({
totalActiveUsers: sql`COUNT(au.id)`,
totalCompletedOrders: sql`COUNT(co.id)`,
})
.from(sql`activeUsers au`)
.leftJoin(sql`completedOrders co`, eq(sql`au.id`, sql`co.userId`));
// Rovo Integration for secure dynamic SQL queries
const rovo = forgeSQL.rovo();
const settings = await rovo
.rovoSettingBuilder(usersTable, accountId)
.addContextParameter(":currentUserId", accountId)
.useRLS()
.addRlsColumn(usersTable.id)
.addRlsWherePart((alias) => `${alias}.${usersTable.id.name} = '${accountId}'`)
.finish()
.build();
const rovoResult = await rovo.dynamicIsolatedQuery(
"SELECT id, name FROM users WHERE status = 'active' AND userId = :currentUserId",
settings,
);
```
This approach gives you direct access to all Drizzle ORM features while still using the @forge/sql backend with enhanced caching and versioning capabilities.
## Direct Drizzle Usage with Custom Driver
If you prefer to use Drizzle ORM directly without the additional features of Forge-SQL-ORM (like optimistic locking), you can use the custom driver:
```typescript
import { drizzle } from "drizzle-orm/mysql-proxy";
import { forgeDriver, patchDbWithSelectAliased } from "forge-sql-orm";
// Initialize drizzle with the custom driver and patch it for aliased selects
const db = patchDbWithSelectAliased(drizzle(forgeDriver));
// Use drizzle directly
const users = await db.select().from(users);
const users = await db.selectAliased(getTableColumns(users)).from(users);
const users = await db.selectAliasedDistinct(getTableColumns(users)).from(users);
await db.insert(users)...;
await db.update(users)...;
await db.delete(users)...;
// Use drizzle with kvs cache
const users = await db.selectAliasedCacheable(getTableColumns(users)).from(users);
const users = await db.selectAliasedDistinctCacheable(getTableColumns(users)).from(users);
await db.insertAndEvictCache(users)...;
await db.updateAndEvictCache(users)...;
await db.deleteAndEvictCache(users)...;
// Use drizzle with kvs cache context
await forgeSQL.executeWithCacheContext(async () => {
await db.insertWithCacheContext(users)...;
await db.updateWithCacheContext(users)...;
await db.deleteWithCacheContext(users)...;
// invoke without cache
const users = await db.selectAliasedCacheable(getTableColumns(users)).from(users);
// Cache is cleared only once at the end for all affected tables
});
// Using new methods with direct drizzle
const usersFrom = await forgeSQL.selectFrom(users)
.where(eq(users.active, true));
const usersDistinct = await forgeSQL.selectDistinctFrom(users)
.where(eq(users.active, true));
const usersCacheable = await forgeSQL.selectCacheableFrom(users)
.where(eq(users.active, true));
// Raw SQL execution
const rawUsers = await forgeSQL.execute(
"SELECT * FROM users WHERE active = ?",
[true]
);
// Raw SQL with caching
// ⚠️ IMPORTANT: When using executeCacheable(), all table names must be wrapped with backticks (`)
const cachedRawUsers = await forgeSQL.executeCacheable(
"SELECT * FROM `users` WHERE active = ?",
[true],
300
);
// Raw SQL with execution metadata and performance monitoring
const usersWithMetadata = await forgeSQL.executeWithMetadata(
async () => {
const users = await forgeSQL.selectFrom(usersTable);
const orders = await forgeSQL.selectFrom(ordersTable).where(eq(ordersTable.userId, usersTable.id));
return { users, orders };
},
(totalDbExecutionTime, totalResponseSize, printQueriesWithPlan) => {
const threshold = 500; // ms baseline for this resolver
if (totalDbExecutionTime > threshold * 1.5) {
console.warn(`[Performance Warning] Resolver exceeded DB time: ${totalDbExecutionTime} ms`);
await printQueriesWithPlan(); // Analyze and print query execution plans
} else if (totalDbExecutionTime > threshold) {
console.debug(`[Performance Debug] High DB time: ${totalDbExecutionTime} ms`);
}
console.log(`DB response size: ${totalResponseSize} bytes`);
},
{
// Optional: Configure query plan printing
mode: 'TopSlowest', // Print top slowest queries (default)
topQueries: 1, // Print top slowest query
},
);
```
## Setting Up Caching with @forge/kvs (Optional)
The caching system is optional and only needed if you want to use cache-related features. To enable the caching system, you need to install the required dependency and configure your manifest.
### How Caching Works
To use caching, you need to use Forge-SQL-ORM methods that support cache management:
**Methods that perform cache eviction after execution and in cache context (batch eviction):**
- `forgeSQL.insertAndEvictCache()`
- `forgeSQL.updateAndEvictCache()`
- `forgeSQL.deleteAndEvictCache()`
- `forgeSQL.modifyWithVersioningAndEvictCache()`
- `forgeSQL.getDrizzleQueryBuilder().insertAndEvictCache()`
- `forgeSQL.getDrizzleQueryBuilder().updateAndEvictCache()`
- `forgeSQL.getDrizzleQueryBuilder().deleteAndEvictCache()`
**Methods that participate in cache context only (batch eviction):**
- All methods except the default Drizzle methods:
- `forgeSQL.insert()`
- `forgeSQL.update()`
- `forgeSQL.delete()`
- `forgeSQL.modifyWithVersioning()`
- `forgeSQL.getDrizzleQueryBuilder().insertWithCacheContext()`
- `forgeSQL.getDrizzleQueryBuilder().updateWithCacheContext()`
- `forgeSQL.getDrizzleQueryBuilder().deleteWithCacheContext()`
**Methods do not do evict cache, better do not use with cache feature:**
- `forgeSQL.getDrizzleQueryBuilder().insert()`
- `forgeSQL.getDrizzleQueryBuilder().update()`
- `forgeSQL.getDrizzleQueryBuilder().delete()`
**Cacheable methods:**
- `forgeSQL.selectCacheable()`
- `forgeSQL.selectDistinctCacheable()`
- `forgeSQL.getDrizzleQueryBuilder().selectAliasedCacheable()`
- `forgeSQL.getDrizzleQueryBuilder().selectAliasedDistinctCacheable()`
**Cache context example:**
```typescript
await forgeSQL.executeWithCacheContext(async () => {
// These methods participate in batch cache clearing
await forgeSQL.insert(Users).values(userData);
await forgeSQL.update(Users).set(updateData).where(eq(Users.id, 1));
await forgeSQL.delete(Users).where(eq(Users.id, 1));
// Cache is cleared only once at the end for all affected tables
});
```
The diagram below shows the lifecycle of a cacheable query in Forge-SQL-ORM:
1. Resolver calls forge-sql-orm with a SQL query and parameters.
2. forge-sql-orm generates a cache key = hash(sql, parameters).
3. It asks @forge/kvs for an existing cached result.
- Cache hit → result is returned immediately.
- Cache miss / expired → query is executed against @forge/sql.
4. Fresh result is stored in @forge/kvs with TTL and returned to the caller.
The diagram below shows how Evict Cache works in Forge-SQL-ORM:
1. **Data modification** is executed through `@forge/sql` (e.g., `UPDATE users ...`).
2. After a successful update, **forge-sql-orm** queries the `cache` entity by using the **`sql` field** with `filter.contains("users")` to find affected cached queries.
3. The returned cache entries are deleted in **batches** (up to 25 per transaction).
4. Once eviction is complete, the update result is returned to the resolver.
5. **Note:** Expired entries are not processed here — they are cleaned up separately by the scheduled cache cleanup trigger using the `expiration` index.
The diagram below shows how Scheduled Expiration Cleanup works:
**Note:** forge-sql-orm uses Forge KVS TTL feature (`{ ttl: { unit: "SECONDS", value: number } }`) to mark entries as expired. However, **actual deletion is asynchronous and may take up to 48 hours**. During this window, read operations may still return expired results. The scheduler trigger proactively cleans up expired entries to prevent cache growth from impacting INSERT/UPDATE performance.
1. A periodic scheduler (Forge trigger) runs cache cleanup independently of data modifications.
2. forge-sql-orm queries the cache entity by the expiration index to find entries with expiration < now.
3. Entries are deleted in batches (up to 25 per transaction) until the page is empty; pagination is done with a cursor (e.g., 100 per page).
4. This keeps the cache footprint small and prevents stale data accumulation, especially important when cache size impacts data modification performance.
The diagram below shows how Cache Context works:
`executeWithCacheContext(fn)` lets you group multiple data modifications and perform **one consolidated cache eviction** at the end:
1. The context starts with an empty `affectedTables` set.
2. Each successful `INSERT/UPDATE/DELETE` inside the context registers its table name in `affectedTables`.
3. **Reads inside the same context** that target tables present in `affectedTables` will **bypass the cache** (read-through to SQL) to avoid serving stale data. These reads also **do not write** back to cache until eviction completes.
4. On context completion, `affectedTables` is de-duplicated and used to build **one combined KVS query** over the `sql` field with
`filter.or(filter.contains(""), filter.contains(""), ...)`, returning all impacted cache entries in a single scan (paged by cursor, e.g., 100/page).
5. Matching cache entries are deleted in **batches** (≤25 per transaction) until the page is exhausted; then the next page is fetched via the cursor.
6. Expiration is handled separately by the scheduled cleanup and is **not part of** the context flow.
### Important Considerations
**@forge/kvs Limits:**
Please review the [official @forge/kvs quotas and limits](https://developer.atlassian.com/platform/forge/platform-quotas-and-limits/#kvs-and-custom-entity-store-quotas) before implementing caching.
**TTL Limitations:**
- **Maximum TTL**: The maximum supported TTL is 1 year from the time the expiry is set.
- **Asynchronous deletion**: Expired data is not removed immediately upon expiry. Deletion may take up to 48 hours. During this window, read operations may still return expired results.
- **Performance impact**: If cache grows large, expired entries can impact INSERT/UPDATE performance. Use the Clear Cache Scheduler Trigger to proactively clean up expired entries.
**Caching Guidelines:**
- Don't cache everything - be selective about what to cache
- Don't cache simple and fast queries - sometimes direct query is faster than cache
- Consider data size and frequency of changes
- Monitor cache usage to stay within quotas
- Use appropriate TTL values
- If cache growth impacts performance, configure the Clear Cache Scheduler Trigger
**⚠️ Important Cache Limitations:**
- **Table names starting with `a_`**: Tables whose names start with `a_` (case-insensitive) are automatically ignored in cache operations. KVS Cache will not work with such tables, and they will be excluded from cache invalidation and cache key generation.
### Step 1: Install Dependencies
```bash
npm install @forge/kvs -S
```
### Step 2: Configure Manifest
Add the storage entity configuration to your `manifest.yml`. The `scheduledTrigger` is **optional** - only configure it if your cache grows large and impacts INSERT/UPDATE performance:
```yaml
modules:
# Optional: Only needed if cache growth impacts INSERT/UPDATE performance
scheduledTrigger:
- key: clear-cache-trigger
function: clearCache
interval: fiveMinute
storage:
entities:
- name: cache
attributes:
sql:
type: string
expiration:
type: integer
data:
type: string
indexes:
- sql
- expiration
sql:
- key: main
engine: mysql
function:
- key: clearCache
handler: index.clearCache
```
```typescript
// Example usage in your Forge app
import { clearCacheSchedulerTrigger } from "forge-sql-orm";
export const clearCache = () => {
return clearCacheSchedulerTrigger({
cacheEntityName: "cache",
});
};
```
### Step 3: Configure ORM Options
Set the cache entity name in your ForgeSQL configuration:
```typescript
const options = {
cacheEntityName: "cache", // Must match the entity name in manifest.yml
cacheTTL: 300, // Default cache TTL in seconds (5 minutes)
cacheWrapTable: true, // Wrap table names with backticks in cache keys
// ... other options
};
const forgeSQL = new ForgeSQL(options);
```
**Important Notes:**
- The `cacheEntityName` must exactly match the `name` in your manifest storage entities
- The entity attributes (`sql`, `expiration`, `data`) are required for proper cache functionality
- Indexes on `sql` and `expiration` improve cache lookup performance
- Cache data uses Forge KVS TTL for expiration (deletion is asynchronous, may take up to 48 hours)
- No additional permissions are required beyond standard Forge app permissions
### Complete Setup Examples
**Basic setup (without caching):**
**package.json:**
```shell
npm install forge-sql-orm @forge/sql drizzle-orm -S
# For UI-Kit projects, use: npm install forge-sql-orm @forge/sql drizzle-orm -S --legacy-peer-deps
```
**manifest.yml:**
```yaml
modules:
sql:
- key: main
engine: mysql
```
**index.ts:**
```typescript
import ForgeSQL from "forge-sql-orm";
const forgeSQL = new ForgeSQL();
// simple insert
await forgeSQL.insert(Users, [userData]);
// Use versioned operations without caching
await forgeSQL.modifyWithVersioning().insert(Users, [userData]);
const users = await forgeSQL.select({ id: Users.id });
```
**With caching support:**
```shell
npm install forge-sql-orm @forge/sql @forge/kvs drizzle-orm -S
# For UI-Kit projects, use: npm install forge-sql-orm @forge/sql @forge/kvs drizzle-orm -S --legacy-peer-deps
```
**manifest.yml:**
```yaml
modules:
# Optional: Only needed if cache growth impacts INSERT/UPDATE performance
scheduledTrigger:
- key: clear-cache-trigger
function: clearCache
interval: fiveMinute
storage:
entities:
- name: cache
attributes:
sql:
type: string
expiration:
type: integer
data:
type: string
indexes:
- sql
- expiration
sql:
- key: main
engine: mysql
function:
- key: clearCache
handler: index.clearCache
```
**index.ts:**
```typescript
import ForgeSQL from "forge-sql-orm";
const forgeSQL = new ForgeSQL({
cacheEntityName: "cache",
});
import { clearCacheSchedulerTrigger } from "forge-sql-orm";
import { getTableColumns } from "drizzle-orm";
export const clearCache = () => {
return clearCacheSchedulerTrigger({
cacheEntityName: "cache",
});
};
// Now you can use caching features
const usersData = await forgeSQL
.selectCacheable(getTableColumns(users))
.from(users)
.where(eq(users.active, true));
// simple insert
await forgeSQL.insertAndEvictCache(users, [userData]);
// Use versioned operations with caching
await forgeSQL.modifyWithVersioningAndEvictCache().insert(users, [userData]);
// use Cache Context
const data = await forgeSQL.executeWithCacheContextAndReturnValue(async () => {
// after insert mark users to evict
await forgeSQL.insert(users, [userData]);
// after insertAndEvictCache mark orders to evict
await forgeSQL.insertAndEvictCache(orders, [order1, order2]);
// execute query and put result to local cache
await forgeSQL
.selectCacheable({
userId: users.id,
userName: users.name,
orderId: orders.id,
orderName: orders.name,
})
.from(users)
.innerJoin(orders, eq(orders.userId, users.id))
.where(eq(users.active, true));
// use local cache without @forge/kvs and @forge/sql
return await forgeSQL
.selectCacheable({
userId: users.id,
userName: users.name,
orderId: orders.id,
orderName: orders.name,
})
.from(users)
.innerJoin(orders, eq(orders.userId, users.id))
.where(eq(users.active, true));
});
// execute query and put result to kvs cache
await forgeSQL
.selectCacheable({
userId: users.id,
userName: users.name,
orderId: orders.id,
orderName: orders.name,
})
.from(users)
.innerJoin(orders, eq(orders.userId, users.id))
.where(eq(users.active, true));
// get result from @foge/kvs cache without real @forge/sql call
await forgeSQL
.selectCacheable({
userId: users.id,
userName: users.name,
orderId: orders.id,
orderName: orders.name,
})
.from(users)
.innerJoin(orders, eq(orders.userId, users.id))
.where(eq(users.active, true));
// use Local Cache for performance optimization
const optimizedData = await forgeSQL.executeWithLocalCacheContextAndReturnValue(async () => {
// First query - hits database and caches result
const users = await forgeSQL
.select({ id: users.id, name: users.name })
.from(users)
.where(eq(users.active, true));
// Second query - uses local cache (no database call)
const cachedUsers = await forgeSQL
.select({ id: users.id, name: users.name })
.from(users)
.where(eq(users.active, true));
// Using new methods for better performance
const usersFrom = await forgeSQL.selectFrom(users).where(eq(users.active, true));
// This will use local cache (no database call)
const cachedUsersFrom = await forgeSQL.selectFrom(users).where(eq(users.active, true));
// Raw SQL with local caching
const rawUsers = await forgeSQL.execute("SELECT id, name FROM users WHERE active = ?", [true]);
// Insert operation - evicts local cache
await forgeSQL.insert(users).values({ name: "New User", active: true });
// Third query - hits database again and caches new result
const updatedUsers = await forgeSQL
.select({ id: users.id, name: users.name })
.from(users)
.where(eq(users.active, true));
return { users, cachedUsers, updatedUsers, usersFrom, cachedUsersFrom, rawUsers };
});
```
## Choosing the Right Method - ForgeSQL ORM
### When to Use Each Approach
| Method | Use Case | Versioning | Cache Management |
| ------------------------------------- | ----------------------------------------------------------- | ---------- | -------------------- |
| `modifyWithVersioningAndEvictCache()` | High-concurrency scenarios with Cache support | ✅ Yes | ✅ Yes |
| `modifyWithVersioning()` | High-concurrency scenarios | ✅ Yes | Cache Context |
| `insertAndEvictCache()` | Simple inserts | ❌ No | ✅ Yes |
| `updateAndEvictCache()` | Simple updates | ❌ No | ✅ Yes |
| `deleteAndEvictCache()` | Simple deletes | ❌ No | ✅ Yes |
| `insert/update/delete` | Basic Drizzle operations | ❌ No | Cache Context |
| `selectFrom()` | All-column queries with field aliasing | ❌ No | Local Cache |
| `selectDistinctFrom()` | Distinct all-column queries with field aliasing | ❌ No | Local Cache |
| `selectCacheableFrom()` | All-column queries with field aliasing and caching | ❌ No | Local + Global Cache |
| `selectDistinctCacheableFrom()` | Distinct all-column queries with field aliasing and caching | ❌ No | Local + Global Cache |
| `execute()` | Raw SQL queries with local caching | ❌ No | Local Cache |
| `executeCacheable()` | Raw SQL queries with local and global caching | ❌ No | Local + Global Cache |
| `executeDDL()` | DDL operations (CREATE, ALTER, DROP, etc.) | ❌ No | No Caching |
| `executeDDLActions()` | Execute regular SQL queries in DDL operation context | ❌ No | No Caching |
| `with()` | Common Table Expressions (CTEs) | ❌ No | Local Cache |
## Choosing the Right Method - Direct Drizzle
### When to Use Each Approach
| Method | Use Case | Versioning | Cache Management |
| ---------------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------- | ---------- | -------------------- |
| `insertWithCacheContext/insertWithCacheContext/updateWithCacheContext` | Basic Drizzle operations | ❌ No | Cache Context |
| `insertAndEvictCache()` | Simple inserts without conflicts | ❌ No | ✅ Yes |
| `updateAndEvictCache()` | Simple updates without conflicts | ❌ No | ✅ Yes |
| `deleteAndEvictCache()` | Simple deletes without conflicts | ❌ No | ✅ Yes |
| `insert/update/delete` | Basic Drizzle operations | ❌ No | ❌ No |
| `selectFrom()` | All-column queries with field aliasing | ❌ No | Local Cache |
| `selectDistinctFrom()` | Distinct all-column queries with field aliasing | ❌ No | Local Cache |
| `selectCacheableFrom()` | All-column queries with field aliasing and caching | ❌ No | Local + Global Cache |
| `selectDistinctCacheableFrom()` | Distinct all-column queries with field aliasing and caching | ❌ No | Local + Global Cache |
| `execute()` | Raw SQL queries with local caching | ❌ No | Local Cache |
| `executeCacheable()` | Raw SQL queries with local and global caching | ❌ No | Local + Global Cache |
| `executeWithMetadata()` | Resolver-level profiling with execution metrics and configurable query plan printing (TopSlowest or SummaryTable mode) | ❌ No | Local Cache |
| `executeDDL()` | DDL operations (CREATE, ALTER, DROP, etc.) | ❌ No | No Caching |
| `executeDDLActions()` | Execute regular SQL queries in DDL operation context | ❌ No | No Caching |
| `with()` | Common Table Expressions (CTEs) | ❌ No | Local Cache |
where Cache context - allows you to batch cache invalidation events and bypass cache reads for affected tables.
## Step-by-Step Migration Workflow
1. **Install CLI and setup scripts**
```bash
npm install forge-sql-orm-cli -D
npm pkg set scripts.models:create="forge-sql-orm-cli generate:model --output src/entities --saveEnv"
npm pkg set scripts.migration:create="forge-sql-orm-cli migrations:create --force --output src/migration --entitiesPath src/entities"
npm pkg set scripts.migration:update="forge-sql-orm-cli migrations:update --entitiesPath src/entities --output src/migration"
```
_(This is done only once when setting up the project)_
2. **Generate initial schema from an existing database**
```sh
npm run models:create
```
_(This will prompt for database credentials on first run and save them to `.env` file)_
3. **Create the first migration**
```sh
npm run migration:create
```
_(This initializes the database migration structure, also done once)_
4. **Deploy to Forge and verify that migrations work**
- Deploy your **Forge app** with migrations.
- Run migrations using a **Forge web trigger** or **Forge scheduler**.
5. **Modify the database (e.g., add a new column, index, etc.)**
- Use **DbSchema** or manually alter the database schema.
6. **Update the migration**
```sh
npm run migration:update
```
- ⚠️ **Do NOT update schema before this step!**
- If schema is updated first, the migration will be empty!
7. **Deploy to Forge and verify that the migration runs without issues**
- Run the updated migration on Forge.
8. **Update the schema**
```sh
npm run models:create
```
9. **Repeat steps 5-8 as needed**
**⚠️ WARNING:**
- **Do NOT swap steps 7 and 5!** If you update schema before generating a migration, the migration will be empty!
- Always generate the **migration first**, then update the **schema**.
## Drop Migrations
The Drop Migrations feature allows you to completely reset your database schema in Atlassian Forge SQL. This is useful when you need to:
- Start fresh with a new schema
- Reset all tables and their data
- Clear migration history
- Ensure your local schema matches the deployed database
### Important Requirements
Before using Drop Migrations, ensure that:
1. Your local schema exactly matches the current database schema deployed in Atlassian Forge SQL
2. You have a backup of your data if needed
3. You understand that this operation will delete all tables and data
### Usage
1. First, ensure your local schema matches the deployed database:
```bash
npm run models:create
```
2. Generate the drop migration:
```bash
npm run migration:drop
```
_(Add this script to your package.json: `npm pkg set scripts.migration:drop="forge-sql-orm-cli migrations:drop --entitiesPath src/entities --output src/migration"`)_
3. Deploy and run the migration in your Forge app:
```js
import migrationRunner from "./database/migration";
import { MigrationRunner } from "@forge/sql/out/migration";
const runner = new MigrationRunner();
await migrationRunner(runner);
await runner.run();
```
4. After dropping all tables, you can create a new migration to recreate the schema:
```bash
npm run migration:create
```
The `--force` parameter is already included in the script to allow creating migrations after dropping all tables.
### Example Migration Output
The generated drop migration will look like this:
```js
import { MigrationRunner } from "@forge/sql/out/migration";
export default (migrationRunner: MigrationRunner): MigrationRunner => {
return migrationRunner
.enqueue("v1_MIGRATION0", "ALTER TABLE `orders` DROP FOREIGN KEY `fk_orders_users`")
.enqueue("v1_MIGRATION1", "DROP INDEX `idx_orders_user_id` ON `orders`")
.enqueue("v1_MIGRATION2", "DROP TABLE IF EXISTS `orders`")
.enqueue("v1_MIGRATION3", "DROP TABLE IF EXISTS `users`")
.enqueue("MIGRATION_V1_1234567890", "DELETE FROM __migrations");
};
```
### ⚠️ Important Notes
- This operation is **irreversible** - all data will be lost
- Make sure your local schema is up-to-date with the deployed database
- Consider backing up your data before running drop migrations
- The migration will clear the `__migrations` table to allow for fresh migration history
- Drop operations are performed in the correct order: first foreign keys, then indexes, then tables
---
## Date and Time Types
When working with date and time fields in your models, you should use the custom types provided by Forge-SQL-ORM to ensure proper handling of date/time values. This is necessary because Forge SQL has specific format requirements for date/time values:
| Date type | Required Format | Example |
| --------- | ------------------------------ | -------------------------- |
| DATE | YYYY-MM-DD | 2024-09-19 |
| TIME | HH:MM:SS[.fraction] | 06:40:34 |
| TIMESTAMP | YYYY-MM-DD HH:MM:SS[.fraction] | 2024-09-19 06:40:34.999999 |
```typescript
// ❌ Don't use standard Drizzle date/time types
export const testEntityTimeStampVersion = mysqlTable("test_entity", {
id: int("id").primaryKey().autoincrement(),
time_stamp: timestamp("times_tamp").notNull(),
date_time: datetime("date_time").notNull(),
time: time("time").notNull(),
date: date("date").notNull(),
});
// ✅ Use Forge-SQL-ORM custom types instead
import {
forgeDateTimeString,
forgeDateString,
forgeTimestampString,
forgeTimeString,
} from "forge-sql-orm";
export const testEntityTimeStampVersion = mysqlTable("test_entity", {
id: int("id").primaryKey().autoincrement(),
time_stamp: forgeTimestampString("times_tamp").notNull(),
date_time: forgeDateTimeString("date_time").notNull(),
time: forgeTimeString("time").notNull(),
date: forgeDateString("date").notNull(),
});
```
### Why Custom Types?
The custom types in Forge-SQL-ORM handle the conversion between JavaScript Date objects and Forge SQL's required string formats automatically. Without these custom types, you would need to manually format dates like this:
```typescript
// Without custom types, you'd need to do this manually:
const date = moment().format("YYYY-MM-DD");
const time = moment().format("HH:mm:ss.SSS");
const timestamp = moment().format("YYYY-MM-DDTHH:mm:ss.SSS");
```
Our custom types provide:
- Automatic conversion between JavaScript Date objects and Forge SQL's required string formats
- Consistent date/time handling across your application
- Type safety for date/time fields
- Proper handling of timezone conversions
- Support for all Forge SQL date/time types (datetime, timestamp, date, time)
### Available Custom Types
- `forgeDateTimeString` - For datetime fields (YYYY-MM-DD HH:MM:SS[.fraction])
- `forgeTimestampString` - For timestamp fields (YYYY-MM-DD HH:MM:SS[.fraction])
- `forgeDateString` - For date fields (YYYY-MM-DD)
- `forgeTimeString` - For time fields (HH:MM:SS[.fraction])
Each type ensures that the data is properly formatted according to Forge SQL's requirements while providing a clean, type-safe interface for your application code.
# Connection to ORM
```js
import ForgeSQL from "forge-sql-orm";
const forgeSQL = new ForgeSQL();
```
or
```typescript
import { drizzle } from "drizzle-orm/mysql-proxy";
import { forgeDriver } from "forge-sql-orm";
// Initialize drizzle with the custom driver
const db = drizzle(forgeDriver);
// Use drizzle directly
const users = await db.select().from(users);
```
## Fetch Data
### Basic Fetch Operations
```js
// Using forgeSQL.select()
const user = await forgeSQL.select({ user: users }).from(users);
// Using forgeSQL.selectDistinct()
const user = await forgeSQL.selectDistinct({ user: users }).from(users);
// Using forgeSQL.selectCacheable()
const user = await forgeSQL.selectCacheable({ user: users }).from(users);
// Using forgeSQL.selectFrom() - Select all columns with field aliasing
const user = await forgeSQL.selectFrom(users).where(eq(users.id, 1));
// Using forgeSQL.selectDistinctFrom() - Select distinct all columns with field aliasing
const user = await forgeSQL.selectDistinctFrom(users).where(eq(users.id, 1));
// Using forgeSQL.selectCacheableFrom() - Select all columns with field aliasing and caching
const user = await forgeSQL.selectCacheableFrom(users).where(eq(users.id, 1));
// Using forgeSQL.selectDistinctCacheableFrom() - Select distinct all columns with field aliasing and caching
const user = await forgeSQL.selectDistinctCacheableFrom(users).where(eq(users.id, 1));
// Using forgeSQL.execute() - Execute raw SQL with local caching
const user = await forgeSQL.execute("SELECT * FROM users WHERE id = ?", [1]);
// Using forgeSQL.executeCacheable() - Execute raw SQL with local and global caching
// ⚠️ IMPORTANT: When using executeCacheable(), all table names in SQL queries must be wrapped with backticks (`)
// Example: SELECT * FROM `users` WHERE id = ? (NOT: SELECT * FROM users WHERE id = ?)
const user = await forgeSQL.executeCacheable("SELECT * FROM `users` WHERE id = ?", [1], 300);
// Using forgeSQL.getDrizzleQueryBuilder()
const user = await forgeSQL.getDrizzleQueryBuilder().select().from(Users).where(eq(Users.id, 1));
// OR using direct drizzle with custom driver
const db = drizzle(forgeDriver);
const user = await db.select().from(Users).where(eq(Users.id, 1));
// Returns: { id: 1, name: "John Doe" }
// Using executeQueryOnlyOne for single result with error handling
const user = await forgeSQL
.fetch()
.executeQueryOnlyOne(
forgeSQL.getDrizzleQueryBuilder().select().from(Users).where(eq(Users.id, 1)),
);
// Returns: { id: 1, name: "John Doe" }
// Throws error if multiple records found
// Returns undefined if no records found
// Using with aliases
// With forgeSQL
const usersAlias = alias(Users, "u");
const result = await forgeSQL
.getDrizzleQueryBuilder()
.select({
userId: sql < string > `${usersAlias.id} as \`userId\``,
userName: sql < string > `${usersAlias.name} as \`userName\``,
})
.from(usersAlias);
// OR with direct drizzle
const db = drizzle(forgeDriver);
const result = await db
.select({
userId: sql < string > `${usersAlias.id} as \`userId\``,
userName: sql < string > `${usersAlias.name} as \`userName\``,
})
.from(usersAlias);
// Returns: { userId: 1, userName: "John Doe" }
```
### Complex Queries
```js
// Using joins with automatic field name collision prevention
// With forgeSQL
const orderWithUser = await forgeSQL
.select({ user: users, order: orders })
.from(orders)
.innerJoin(users, eq(orders.userId, users.id));
// Using new selectFrom methods with joins
const orderWithUser = await forgeSQL
.selectFrom(orders)
.innerJoin(users, eq(orders.userId, users.id))
.where(eq(orders.id, 1));
// Using selectCacheableFrom with joins and caching
const orderWithUser = await forgeSQL
.selectCacheableFrom(orders)
.innerJoin(users, eq(orders.userId, users.id))
.where(eq(orders.id, 1));
// Using with() for Common Table Expressions (CTEs)
const userStats = await forgeSQL
.with(
forgeSQL.selectFrom(users).where(eq(users.active, true)).as("activeUsers"),
forgeSQL.selectFrom(orders).where(eq(orders.status, "completed")).as("completedOrders"),
)
.select({
totalActiveUsers: sql`COUNT(au.id)`,
totalCompletedOrders: sql`COUNT(co.id)`,
})
.from(sql`activeUsers au`)
.leftJoin(sql`completedOrders co`, eq(sql`au.id`, sql`co.userId`));
// OR with direct drizzle
const db = patchDbWithSelectAliased(drizzle(forgeDriver));
const orderWithUser = await db
.selectAliased({ user: users, order: orders })
.from(orders)
.innerJoin(users, eq(orders.userId, users.id));
// Returns: {
// user_id: 1,
// user_name: "John Doe",
// order_id: 1,
// order_product: "Product 1"
// }
// Using distinct with aliases
const uniqueUsers = await db.selectAliasedDistinct({ user: users }).from(users);
// Returns unique users with aliased fields
// Using executeQueryOnlyOne for unique results
const userStats = await forgeSQL.fetch().executeQueryOnlyOne(
forgeSQL
.getDrizzleQueryBuilder()
.select({
totalUsers: sql`COUNT(*) as \`totalUsers\``,
uniqueNames: sql`COUNT(DISTINCT name) as \`uniqueNames\``,
})
.from(Users),
);
// Returns: { totalUsers: 100, uniqueNames: 80 }
// Throws error if multiple records found
```
### Raw SQL Queries
```js
// Using executeRawSQL for direct SQL queries
const users = await forgeSQL
.fetch()
.executeRawSQL("SELECT * FROM users");
// Using execute() for raw SQL with local caching
const users = await forgeSQL
.execute("SELECT * FROM users WHERE active = ?", [true]);
// Using executeCacheable() for raw SQL with local and global caching
// ⚠️ IMPORTANT: When using executeCacheable(), all table names in SQL queries must be wrapped with backticks (`)
// Example: SELECT * FROM `users` WHERE active = ? (NOT: SELECT * FROM users WHERE active = ?)
const users = await forgeSQL
.executeCacheable("SELECT * FROM `users` WHERE active = ?", [true], 300);
// Using executeWithMetadata() for capturing execution metrics and performance monitoring
const usersWithMetadata = await forgeSQL.executeWithMetadata(
async () => {
const users = await forgeSQL.selectFrom(usersTable);
const orders = await forgeSQL.selectFrom(ordersTable).where(eq(ordersTable.userId, usersTable.id));
return { users, orders };
},
(totalDbExecutionTime, totalResponseSize, printQueriesWithPlan) => {
const threshold = 500; // ms baseline for this resolver
if (totalDbExecutionTime > threshold * 1.5) {
console.warn(`[Performance Warning] Resolver exceeded DB time: ${totalDbExecutionTime} ms`);
await printQueriesWithPlan(); // Analyze and print query execution plans
} else if (totalDbExecutionTime > threshold) {
console.debug(`[Performance Debug] High DB time: ${totalDbExecutionTime} ms`);
}
console.log(`DB response size: ${totalResponseSize} bytes`);
},
{
// Optional: Configure query plan printing
mode: 'TopSlowest', // Print top slowest queries (default)
topQueries: 1, // Print top slowest query
},
);
// Using executeDDL() for DDL operations (CREATE, ALTER, DROP, etc.)
await forgeSQL.executeDDL(`
CREATE TABLE users (
id INT PRIMARY KEY AUTO_INCREMENT,
name VARCHAR(255) NOT NULL,
email VARCHAR(255) UNIQUE
)
`);
await forgeSQL.executeDDL(sql`
ALTER TABLE users
ADD COLUMN created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP
`);
await forgeSQL.executeDDL("DROP TABLE IF EXISTS old_users");
// Using executeDDLActions() for executing regular SQL queries in DDL context
// This method executes a series of actions within a DDL operation context for monitoring
await forgeSQL.executeDDLActions(async () => {
// Execute regular SQL queries in DDL context for performance monitoring
const slowQueries = await forgeSQL.execute(`
SELECT * FROM INFORMATION_SCHEMA.STATEMENTS_SUMMARY
WHERE AVG_LATENCY > 1000000
`);
// Execute complex analysis queries in DDL context
const performanceData = await forgeSQL.execute(`
SELECT * FROM INFORMATION_SCHEMA.CLUSTER_STATEMENTS_SUMMARY_HISTORY
WHERE SUMMARY_END_TIME > DATE_SUB(NOW(), INTERVAL 1 HOUR)
`);
return { slowQueries, performanceData };
});
// Using execute() with complex queries
const userStats = await forgeSQL
.execute(`
SELECT
u.id,
u.name,
COUNT(o.id) as order_count,
SUM(o.amount) as total_amount
FROM users u
LEFT JOIN orders o ON u.id = o.user_id
WHERE u.active = ?
GROUP BY u.id, u.name
`, [true]);
```
## Modify Operations
Forge-SQL-ORM provides multiple approaches for Modify operations, each with different characteristics:
### 1. Basic Drizzle Operations (Cache Context Aware)
These operations work like standard Drizzle methods but participate in cache context when used within `executeWithCacheContext()`:
```js
// Basic insert (participates in cache context when used within executeWithCacheContext)
await forgeSQL.insert(Users).values({ id: 1, name: "Smith" });
// Basic update (participates in cache context when used within executeWithCacheContext)
await forgeSQL.update(Users).set({ name: "Smith Updated" }).where(eq(Users.id, 1));
// Basic delete (participates in cache context when used within executeWithCacheContext)
await forgeSQL.delete(Users).where(eq(Users.id, 1));
```
### 2. Non-Versioned Operations with Cache Management
These operations don't use optimistic locking but provide cache invalidation:
```js
// Insert without versioning but with cache invalidation
await forgeSQL.insertAndEvictCache(Users).values({ id: 1, name: "Smith" });
// Update without versioning but with cache invalidation
await forgeSQL.updateAndEvictCache(Users).set({ name: "Smith Updated" }).where(eq(Users.id, 1));
// Delete without versioning but with cache invalidation
await forgeSQL.deleteAndEvictCache(Users).where(eq(Users.id, 1));
```
### 3. Versioned Operations with Cache Management (Recommended)
These operations use optimistic locking and automatic cache invalidation:
```js
// Insert with versioning and cache management
const userId = await forgeSQL
.modifyWithVersioningAndEvictCache()
.insert(Users, [{ id: 1, name: "Smith" }]);
// Bulk insert with versioning
await forgeSQL.modifyWithVersioningAndEvictCache().insert(Users, [
{ id: 2, name: "Smith" },
{ id: 3, name: "Vasyl" },
]);
// Update by ID with optimistic locking and cache invalidation
await forgeSQL
.modifyWithVersioningAndEvictCache()
.updateById({ id: 1, name: "Smith Updated" }, Users);
// Delete by ID with versioning and cache invalidation
await forgeSQL.modifyWithVersioningAndEvictCache().deleteById(1, Users);
```
### 4. Versioned Operations without Cache Management
These operations use optimistic locking but don't manage cache:
```js
// Insert with versioning only (no cache management)
const userId = await forgeSQL.modifyWithVersioning().insert(Users, [{ id: 1, name: "Smith" }]);
// Update with versioning only
await forgeSQL.modifyWithVersioning().updateById({ id: 1, name: "Smith Updated" }, Users);
// Delete with versioning only
await forgeSQL.modifyWithVersioning().deleteById(1, Users);
```
### 5. Legacy Modify Operations (Removed in 2.1.x)
⚠️ **BREAKING CHANGE**: The `crud()` and `modify()` methods have been completely removed in version 2.1.x.
```js
// ❌ These methods no longer exist in 2.1.x
// const userId = await forgeSQL.crud().insert(Users, [{ id: 1, name: "Smith" }]);
// await forgeSQL.crud().updateById({ id: 1, name: "Smith Updated" }, Users);
// await forgeSQL.crud().deleteById(1, Users);
// ✅ Use the new methods instead
const userId = await forgeSQL.modifyWithVersioning().insert(Users, [{ id: 1, name: "Smith" }]);
await forgeSQL.modifyWithVersioning().updateById({ id: 1, name: "Smith Updated" }, Users);
await forgeSQL.modifyWithVersioning().deleteById(1, Users);
```
### Advanced Operations
```js
// Insert with sequence (nextVal)
import { nextVal } from "forge-sql-orm";
const user = {
id: nextVal("user_id_seq"),
name: "user test",
organization_id: 1,
};
const id = await forgeSQL.modifyWithVersioning().insert(appUser, [user]);
// Update with custom WHERE condition
await forgeSQL
.modifyWithVersioning()
.updateFields({ name: "New Name", age: 35 }, Users, eq(Users.email, "smith@example.com"));
// Insert with duplicate handling
await forgeSQL.modifyWithVersioning().insert(
Users,
[
{ id: 4, name: "Smith" },
{ id: 4, name: "Vasyl" },
],
true,
);
```
## SQL Utilities
### formatLimitOffset
The `formatLimitOffset` utility function is used to safely insert numeric values directly into SQL queries for LIMIT and OFFSET clauses. This is necessary because Atlassian Forge SQL doesn't support parameterized queries for these clauses.
```typescript
import { formatLimitOffset } from "forge-sql-orm";
// Example usage in a query
const result = await forgeSQL
.select()
.from(orderItem)
.orderBy(asc(orderItem.createdAt))
.limit(formatLimitOffset(10))
.offset(formatLimitOffset(350000));
// The generated SQL will be:
// SELECT * FROM order_item
// ORDER BY created_at ASC
// LIMIT 10
// OFFSET 350000
```
**Important Notes:**
- The function performs type checking to prevent SQL injection
- It throws an error if the input is not a valid number
- Use this function instead of direct parameter binding for LIMIT and OFFSET clauses
- The function is specifically designed to work with Atlassian Forge SQL's limitations
**Security Considerations:**
- The function includes validation to ensure the input is a valid number
- This prevents SQL injection by ensuring only numeric values are inserted
- Always use this function instead of string concatenation for LIMIT and OFFSET values
## Global Cache System (Level 2)
[↑ Back to Top](#table-of-contents)
Forge-SQL-ORM includes a sophisticated global caching system that provides **cross-invocation caching** - the ability to share cached data between different resolver invocations. The global cache system is built on top of [@forge/kvs Custom entity store](https://developer.atlassian.com/platform/forge/storage-reference/storage-api-custom-entities/) and provides persistent cross-invocation caching with automatic serialization/deserialization of complex data structures.
### Cache Levels Overview
Forge-SQL-ORM implements a two-level caching architecture:
- **Level 1 (Local Cache)**: In-memory caching within a single resolver invocation scope
- **Level 2 (Global Cache)**: Cross-invocation persistent caching using KVS storage
This multi-level approach provides optimal performance by checking the fastest cache first, then falling back to cross-invocation persistent storage.
### Cache Configuration
The caching system uses Atlassian Forge's Custom entity store to persist cache data. Each cache entry is stored as a custom entity with TTL management via Forge KVS. Note that expired data deletion is asynchronous and may take up to 48 hours. If cache growth impacts INSERT/UPDATE performance, use the Clear Cache Scheduler Trigger for proactive cleanup.
```typescript
const options = {
cacheEntityName: "cache", // KVS Custom entity name for cache storage
cacheTTL: 300, // Default cache TTL in seconds (5 minutes)
cacheWrapTable: true, // Wrap table names with backticks in cache keys
additionalMetadata: {
users: {
tableName: "users",
versionField: {
fieldName: "updatedAt",
},
},
},
};
const forgeSQL = new ForgeSQL(options);
```
### How Caching Works with @forge/kvs
The caching system leverages Forge's Custom entity store to provide:
- **Persistent Storage**: Cache data survives app restarts and deployments
- **TTL Support**: Uses Forge KVS TTL feature for expiration (deletion is asynchronous, may take up to 48 hours)
- **Efficient Retrieval**: Fast key-based lookups using Forge's optimized storage
- **Data Serialization**: Automatic handling of complex objects and query results
- **Batch Operations**: Efficient bulk cache operations for better performance
```typescript
// Cache entries are stored as custom entities in Forge's KVS
// Example cache key structure:
// Key: "CachedQuery_8d74bdd9d85064b72fb2ee072ca948e5"
// Value: { data: [...], expiration: 1234567890, sql: "select * from 1" }
```
### Cache Context Operations
The cache context allows you to batch cache invalidation events and bypass cache reads for affected tables:
```typescript
// Execute operations within a cache context
await forgeSQL.executeWithCacheContext(async () => {
// All cache invalidation events are collected and executed in batch
await forgeSQL.modifyWithVersioningAndEvictCache().insert(Users, [userData]);
await forgeSQL.modifyWithVersioningAndEvictCache().updateById(updateData, Users);
// Cache is cleared only once at the end for all affected tables
});
// Execute with return value
const result = await forgeSQL.executeWithCacheContextAndReturnValue(async () => {
const user = await forgeSQL.modifyWithVersioningAndEvictCache().insert(Users, [userData]);
return user;
});
// Basic operations also participate in cache context
await forgeSQL.executeWithCacheContext(async () => {
// These operations will participate in batch cache clearing
await forgeSQL.insert(Users).values(userData);
await forgeSQL.update(Users).set(updateData).where(eq(Users.id, 1));
await forgeSQL.delete(Users).where(eq(Users.id, 1));
// Cache is cleared only once at the end for all affected tables
});
```
### Local Cache Operations (Level 1)
Forge-SQL-ORM provides a local cache system (Level 1 cache) that stores query results in memory for the duration of a single resolver invocation. This is particularly useful for optimizing repeated queries within the same execution context(resolver invocation).
#### What is Local Cache?
Local cache is an in-memory caching layer that operates within a single resolver invocation scope. Unlike the global KVS cache, local cache:
- **Stores data in memory** using Node.js `AsyncLocalStorage`
- **Automatically clears** when the invocation completes (Resolver call)
- **Provides instant access** to previously executed queries in resolver invocation
- **Reduces database load** for repeated operations within the same invocation
- **Works alongside** the global KVS cache system
#### Key Features of Local Cache
- **In-Memory Storage**: Query results are cached in memory using Node.js `AsyncLocalStorage`
- **Invocation-Scoped**: Cache is automatically cleared when the invocation completes
- **Automatic Eviction**: Cache is cleared when insert/update/delete operations are performed
- **No Persistence**: Data is not stored between Invocations (unlike global KVS cache)
- **Performance Optimization**: Reduces database queries for repeated operations
- **Simple Configuration**: Works out of the box with simple setup
#### Usage Examples
##### Basic Local Cache Usage
```typescript
// Execute operations within a local cache context
await forgeSQL.executeWithLocalContext(async () => {
// First call - executes query and caches result
const users = await forgeSQL
.select({ id: users.id, name: users.name })
.from(users)
.where(eq(users.active, true));
// Second call - gets result from local cache (no database query)
const cachedUsers = await forgeSQL
.select({ id: users.id, name: users.name })
.from(users)
.where(eq(users.active, true));
// Using new selectFrom methods with local caching
const usersFrom = await forgeSQL.selectFrom(users).where(eq(users.active, true));
// This will use local cache (no database call)
const cachedUsersFrom = await forgeSQL.selectFrom(users).where(eq(users.active, true));
// Using execute() with local caching
const rawUsers = await forgeSQL.execute("SELECT id, name FROM users WHERE active = ?", [true]);
// This will use local cache (no database call)
const cachedRawUsers = await forgeSQL.execute("SELECT id, name FROM users WHERE active = ?", [
true,
]);
// Raw SQL with execution metadata and performance monitoring
const usersWithMetadata = await forgeSQL.executeWithMetadata(
async () => {
const users = await forgeSQL.selectFrom(usersTable);
const orders = await forgeSQL
.selectFrom(ordersTable)
.where(eq(ordersTable.userId, usersTable.id));
return { users, orders };
},
(totalDbExecutionTime, totalResponseSize, printQueriesWithPlan) => {
const threshold = 500; // ms baseline for this resolver
if (totalDbExecutionTime > threshold * 1.5) {
console.warn(`[Performance Warning] Resolver exceeded DB time: ${totalDbExecutionTime} ms`);
await printQueriesWithPlan(); // Analyze and print query execution plans
} else if (totalDbExecutionTime > threshold) {
console.debug(`[Performance Debug] High DB time: ${totalDbExecutionTime} ms`);
}
console.log(`DB response size: ${totalResponseSize} bytes`);
},
{
// Optional: Configure query plan printing
topQueries: 1, // Print top slowest query (default)
mode: "TopSlowest", // Print top slowest queries (default)
},
);
// Insert operation - evicts local cache for users table
await forgeSQL.insert(users).values({ name: "New User", active: true });
// Third call - executes query again and caches new result
const updatedUsers = await forgeSQL
.select({ id: users.id, name: users.name })
.from(users)
.where(eq(users.active, true));
});
// Execute with return value
const result = await forgeSQL.executeWithLocalCacheContextAndReturnValue(async () => {
// First call - executes query and caches result
const users = await forgeSQL
.select({ id: users.id, name: users.name })
.from(users)
.where(eq(users.active, true));
// Second call - gets result from local cache (no database query)
const cachedUsers = await forgeSQL
.select({ id: users.id, name: users.name })
.from(users)
.where(eq(users.active, true));
return { users, cachedUsers };
});
```
##### Real-World Resolver Example
```typescript
// Atlassian forge resolver with local cache optimization
const userResolver = async (req) => {
return await forgeSQL.executeWithLocalCacheContextAndReturnValue(async () => {
// Get user details using selectFrom (all columns with field aliasing)
const user = await forgeSQL.selectFrom(users).where(eq(users.id, args.userId));
// Get user's orders using selectCacheableFrom (with caching)
const orders = await forgeSQL.selectCacheableFrom(orders).where(eq(orders.userId, args.userId));
// Get user's profile using raw SQL with execute()
const profile = await forgeSQL.execute(
"SELECT id, bio, avatar FROM profiles WHERE user_id = ?",
[args.userId],
);
// Get user statistics using complex raw SQL
const stats = await forgeSQL.execute(
`
SELECT
COUNT(o.id) as total_orders,
SUM(o.amount) as total_spent,
AVG(o.amount) as avg_order_value
FROM orders o
WHERE o.user_id = ? AND o.status = 'completed'
`,
[args.userId],
);
// If any of these queries are repeated within the same resolver,
// they will use the local cache instead of hitting the database
return {
...user[0],
orders,
profile: profile[0],
stats: stats[0],
};
});
};
```
#### Local Cache (Level 1) vs Global Cache (Level 2)
| Feature | Local Cache (Level 1) | Global Cache (Level 2) |
| ------------------ | ------------------------------------- | ------------------------------------------------------------------- |
| **Storage** | In-memory (Node.js process) | Persistent (KVS Custom Entities) |
| **Scope** | Single forge invocation | Cross-invocation (between calls) |
| **Persistence** | No (cleared on invocation end) | Yes (survives app redeploy) |
| **Performance** | Very fast (memory access) | Fast (KVS optimized storage) |
| **Memory Usage** | Low (invocation-scoped) | Higher (persistent storage) |
| **Use Case** | Invocation optimization | Cross-invocation data sharing |
| **Configuration** | None required | Requires KVS setup |
| **TTL Support** | No (invocation-scoped) | Yes (TTL via Forge KVS, async deletion up to 48h) |
| **Cache Eviction** | Automatic on DML operations | Manual or scheduled cleanup (optional if cache impacts performance) |
| **Best For** | Repeated queries in single invocation | Frequently accessed data across invocations |
#### Integration with Global Cache (Level 2)
Local cache (Level 1) works alongside the global cache (Level 2) system:
```typescript
// Multi-level cache checking: Level 1 → Level 2 → Database
await forgeSQL.executeWithLocalContext(async () => {
// This will check:
// 1. Local cache (Level 1 - in-memory)
// 2. Global cache (Level 2 - KVS)
// 3. Database query
const users = await forgeSQL
.selectCacheable({ id: users.id, name: users.name })
.from(users)
.where(eq(users.active, true));
// Using new methods with multi-level caching
const usersFrom = await forgeSQL.selectCacheableFrom(users).where(eq(users.active, true));
// Raw SQL with multi-level caching
// ⚠️ IMPORTANT: When using executeCacheable(), all table names must be wrapped with backticks (`)
const rawUsers = await forgeSQL.executeCacheable(
"SELECT id, name FROM `users` WHERE active = ?",
[true],
300, // TTL in seconds
);
});
```
#### Local Cache Flow Diagram
The diagram below shows how local cache works in Forge-SQL-ORM:
1. **Request Start**: Local cache context is initialized with empty cache
2. **First Query**: Cache miss → Global cache miss → Database query → Save to local cache
3. **Repeated Query**: Cache hit → Return cached result (no database call)
4. **Data Modification**: Insert/Update/Delete → Evict local cache for affected table
5. **Query After Modification**: Cache miss (was evicted) → Database query → Save to local cache
6. **Request End**: Local cache context is destroyed, all data cleared
### Cache-Aware Query Operations
```typescript
// Execute queries with caching
const users = await forgeSQL.modifyWithVersioningAndEvictCache().executeQuery(
forgeSQL.select().from(Users).where(eq(Users.active, true)),
600, // Custom TTL in seconds
);
// Execute single result queries with caching
const user = await forgeSQL
.modifyWithVersioningAndEvictCache()
.executeQueryOnlyOne(forgeSQL.select().from(Users).where(eq(Users.id, 1)));
// Execute raw SQL with caching
const results = await forgeSQL.modifyWithVersioningAndEvictCache().executeRawSQL(
"SELECT * FROM users WHERE active = ?",
[true],
300, // TTL in seconds
);
// Using new methods for cache-aware operations
const usersFrom = await forgeSQL.selectCacheableFrom(Users).where(eq(Users.active, true));
const usersDistinct = await forgeSQL
.selectDistinctCacheableFrom(Users)
.where(eq(Users.active, true));
// Raw SQL with local and global caching
// ⚠️ IMPORTANT: When using executeCacheable(), all table names must be wrapped with backticks (`)
const rawUsers = await forgeSQL.executeCacheable(
"SELECT * FROM `users` WHERE active = ?",
[true],
300, // TTL in seconds
);
// Using with() for Common Table Expressions with caching
const userStats = await forgeSQL
.with(
forgeSQL.selectFrom(users).where(eq(users.active, true)).as("activeUsers"),
forgeSQL.selectFrom(orders).where(eq(orders.status, "completed")).as("completedOrders"),
)
.select({
totalActiveUsers: sql`COUNT(au.id)`,
totalCompletedOrders: sql`COUNT(co.id)`,
})
.from(sql`activeUsers au`)
.leftJoin(sql`completedOrders co`, eq(sql`au.id`, sql`co.userId`));
// Using executeWithMetadata() for capturing execution metrics with performance monitoring
const usersWithMetadata = await forgeSQL.executeWithMetadata(
async () => {
const users = await forgeSQL.selectFrom(usersTable);
const orders = await forgeSQL
.selectFrom(ordersTable)
.where(eq(ordersTable.userId, usersTable.id));
return { users, orders };
},
(totalDbExecutionTime, totalResponseSize, printQueriesWithPlan) => {
const threshold = 500; // ms baseline for this resolver
if (totalDbExecutionTime > threshold * 1.5) {
console.warn(`[Performance Warning] Resolver exceeded DB time: ${totalDbExecutionTime} ms`);
await printQueriesWithPlan(); // Analyze and print query execution plans
} else if (totalDbExecutionTime > threshold) {
console.debug(`[Performance Debug] High DB time: ${totalDbExecutionTime} ms`);
}
console.log(`DB response size: ${totalResponseSize} bytes`);
},
{
// Optional: Configure query plan printing
mode: "TopSlowest", // Print top slowest queries (default)
topQueries: 1, // Print top slowest query
},
);
```
### Manual Cache Management
```typescript
// Clear cache for specific tables
await forgeSQL.modifyWithVersioningAndEvictCache().evictCache(["users", "orders"]);
// Clear cache for specific entities
await forgeSQL.modifyWithVersioningAndEvictCache().evictCacheEntities([Users, Orders]);
```
## Optimistic Locking
[↑ Back to Top](#table-of-contents)
Optimistic locking is a concurrency control mechanism that prevents data conflicts when multiple transactions attempt to update the same record concurrently. Instead of using locks, this technique relies on a version field in your entity models.
### Supported Version Field Types
- `datetime` - Timestamp-based versioning
- `timestamp` - Timestamp-based versioning
- `integer` - Numeric version increment
- `decimal` - Numeric version increment
### Configuration
```typescript
const options = {
additionalMetadata: {
users: {
tableName: "users",
versionField: {
fieldName: "updatedAt",
},
},
},
};
const forgeSQL = new ForgeSQL(options);
```
### Example Usage
```typescript
// The version field will be automatically handled
await forgeSQL.modifyWithVersioning().updateById(
{
id: 1,
name: "Updated Name",
updatedAt: new Date(), // Will be automatically set if not provided
},
Users,
);
```
or with cache support
```typescript
// The version field will be automatically handled
await forgeSQL.modifyWithVersioningAndEvictCache().updateById(
{
id: 1,
name: "Updated Name",
updatedAt: new Date(), // Will be automatically set if not provided
},
Users,
);
```
## Rovo Integration
[↑ Back to Top](#table-of-contents)
Rovo is a secure pattern for natural-language analytics in Forge apps. It enables safe execution of dynamic SQL queries with comprehensive security validations, making it ideal for AI-powered analytics features where users can query data using natural language.
**📖 Real-World Example**: See [Forge-Secure-Notes-for-Jira](https://github.com/vzakharchenko/Forge-Secure-Notes-for-Jira) for a complete implementation of Rovo AI agent with secure natural-language analytics.
### Key Features
- **Security-First Design**: Multiple layers of security validations to prevent SQL injection and unauthorized data access
- **Single Table Isolation**: Queries are restricted to a single table to prevent cross-table data access
- **Row-Level Security (RLS)**: Built-in support for data isolation based on user context
- **Comprehensive Validation**: Blocks JOINs, subqueries, window functions, and other potentially unsafe operations
- **Post-Execution Validation**: Verifies query results to ensure security fields are present and come from the correct table
- **Type-Safe Configuration**: Uses Drizzle ORM table objects for type-safe column references
### Security Validations
Rovo performs multiple security checks before and after query execution:
1. **Query Type Validation**: Only SELECT queries are allowed
2. **Table Restriction**: Queries must target only the specified table
3. **JOIN Detection**: JOINs are blocked using EXPLAIN analysis
4. **Subquery Detection**: Scalar subqueries in SELECT columns are blocked
5. **Window Function Detection**: Window functions are blocked for security
6. **Execution Plan Validation**: Verifies that only the expected table is accessed
7. **RLS Field Validation**: Ensures required security fields are present in results
8. **Post-Execution Validation**: Verifies all fields come from the correct table
### Basic Usage
```typescript
import ForgeSQL from "forge-sql-orm";
const forgeSQL = new ForgeSQL();
// Get Rovo instance
const rovo = forgeSQL.rovo();
// Create settings builder using Drizzle table object
const settings = await rovo
.rovoSettingBuilder(usersTable, accountId)
.addContextParameter(":currentUserId", accountId)
.useRLS()
.addRlsColumn(usersTable.id)
.addRlsWherePart((alias) => `${alias}.${usersTable.id.name} = '${accountId}'`)
.finish()
.build();
// Execute dynamic SQL query
const result = await rovo.dynamicIsolatedQuery(
"SELECT id, name FROM users WHERE status = 'active' AND userId = :currentUserId",
settings,
);
console.log(result.rows); // Query results
console.log(result.metadata); // Query metadata
```
### Row-Level Security (RLS) Configuration
RLS allows you to filter data based on user context, ensuring users can only access their own data:
```typescript
const rovo = forgeSQL.rovo();
// Configure RLS with conditional activation and multiple security fields
const settings = await rovo
.rovoSettingBuilder(securityNotesTable, accountId)
.addContextParameter(":currentUserId", accountId)
.addContextParameter(":currentProjectKey", projectKey)
.addContextParameter(":currentIssueKey", issueKey)
.useRLS()
.addRlsCondition(async () => {
// Conditionally enable RLS based on user role
const userService = getUserService();
return !(await userService.isAdmin()); // Only apply RLS for non-admin users
})
.addRlsColumn(securityNotesTable.createdBy) // Required field for RLS validation
.addRlsColumn(securityNotesTable.targetUserId) // Additional security field
.addRlsWherePart(
(alias) =>
`${alias}.${securityNotesTable.createdBy.name} = '${accountId}' OR ${alias}.${securityNotesTable.targetUserId.name} = '${accountId}'`,
) // RLS filter with OR condition
.finish()
.build();
// The query will automatically be wrapped with RLS filtering:
// SELECT * FROM (original_query) AS t WHERE (t.createdBy = 'accountId' OR t.targetUserId = 'accountId')
```
### Context Parameters
You can use context parameters for query substitution. Parameters use the `:parameterName` format (colon prefix, not double braces):
```typescript
const rovo = forgeSQL.rovo();
const settings = await rovo
.rovoSettingBuilder(usersTable, accountId)
.addContextParameter(":currentUserId", accountId)
.addContextParameter(":projectKey", "PROJ-123")
.addContextParameter(":status", "active")
.useRLS()
.addRlsColumn(usersTable.id)
.addRlsWherePart((alias) => `${alias}.${usersTable.userId.name} = '${accountId}'`)
.finish()
.build();
// In the SQL query, parameters are replaced:
const result = await rovo.dynamicIsolatedQuery(
"SELECT * FROM users WHERE projectKey = :projectKey AND status = :status AND userId = :currentUserId",
settings,
);
// Becomes: SELECT * FROM users WHERE projectKey = 'PROJ-123' AND status = 'active' AND userId = 'accountId'
```
### Using Raw Table Names
You can use `rovoRawSettingBuilder` with raw table name string:
```typescript
const rovo = forgeSQL.rovo();
// Using rovoRawSettingBuilder with raw table name
const settings = await rovo
.rovoRawSettingBuilder("users", accountId)
.addContextParameter(":currentUserId", accountId)
.useRLS()
.addRlsColumnName("id")
.addRlsWherePart((alias) => `${alias}.id = '${accountId}'`)
.finish()
.build();
const result = await rovo.dynamicIsolatedQuery(
"SELECT id, name FROM users WHERE status = 'active' AND userId = :currentUserId",
settings,
);
```
### Security Restrictions
Rovo blocks the following operations for security:
- **Data Modification**: Only SELECT queries are allowed
- **JOINs**: JOIN operations are detected and blocked
- **Subqueries**: Scalar subqueries in SELECT columns are blocked
- **Window Functions**: Window functions (e.g., `COUNT(*) OVER(...)`) are blocked
- **Multiple Tables**: Queries referencing multiple tables are blocked
- **Table Aliases**: Post-execution validation ensures fields come from the correct table
### Error Handling
Rovo provides detailed error messages when security violations are detected:
```typescript
try {
const result = await rovo.dynamicIsolatedQuery(
"SELECT * FROM users u JOIN orders o ON u.id = o.userId",
settings,
);
} catch (error) {
// Error: "Security violation: JOIN operations are not allowed..."
console.error(error.message);
}
```
### Example: Real-World Function Implementation
> **💡 Full Example**: See the complete implementation in [Forge-Secure-Notes-for-Jira](https://github.com/vzakharchenko/Forge-Secure-Notes-for-Jira) repository.
```typescript
import ForgeSQL from "forge-sql-orm";
import { Result } from "@forge/sql";
const FORGE_SQL_ORM = new ForgeSQL();
export async function runSecurityNotesQuery(
event: {
sql: string;
context: {
jira: {
issueKey: string;
projectKey: string;
};
};
},
context: { principal: { accountId: string } },
): Promise> {
const rovoIntegration = FORGE_SQL_ORM.rovo();
const accountId = context.principal.accountId;
const settings = await rovoIntegration
.rovoSettingBuilder(securityNotesTable, accountId)
.addContextParameter(":currentUserId", accountId)
.addContextParameter(":currentProjectKey", event.context?.jira?.projectKey ?? "")
.addContextParameter(":currentIssueKey", event.context?.jira?.issueKey ?? "")
.useRLS()
.addRlsCondition(async () => {
// Conditionally disable RLS for admin users
const userService = getUserService();
return !(await userService.isAdmin());
})
.addRlsColumn(securityNotesTable.createdBy)
.addRlsColumn(securityNotesTable.targetUserId)
.addRlsWherePart(
(alias: string) =>
`${alias}.${securityNotesTable.createdBy.name} = '${accountId}' OR ${alias}.${securityNotesTable.targetUserId.name} = '${accountId}'`,
)
.finish()
.build();
return await rovoIntegration.dynamicIsolatedQuery(event.sql, settings);
}
```
## ForgeSqlOrmOptions
The `ForgeSqlOrmOptions` object allows customization of ORM behavior:
| Option | Type | Description |
| -------------------------- | --------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
| `logRawSqlQuery` | `boolean` | Enables logging of raw SQL queries in the Atlassian Forge Developer Console. Useful for debugging and monitoring. Defaults to `false`. |
| `logCache` | `boolean` | Enables logging of cache operations (hits, misses, evictions) in the Atlassian Forge Developer Console. Useful for debugging caching issues. Defaults to `false`. |
| `disableOptimisticLocking` | `boolean` | Disables optimistic locking. When set to `true`, no additional condition (e.g., a version check) is added during record updates, which can improve performance. However, this may lead to conflicts when multiple transactions attempt to update the same record concurrently. |
| `additionalMetadata` | `object` | Allows adding custom metadata to all entities. This is useful for tracking common fields across all tables (e.g., `createdAt`, `updatedAt`, `createdBy`, etc.). The metadata will be automatically added to all generated entities. |
| `cacheEntityName` | `string` | KVS Custom entity name for cache storage. Must match the `name` in your `manifest.yml` storage entities configuration. Required for caching functionality. Defaults to `"cache"`. |
| `cacheTTL` | `number` | Default cache TTL in seconds. Defaults to `120` (2 minutes). |
| `cacheWrapTable` | `boolean` | Whether to wrap table names with backticks in cache keys. Defaults to `true`. |
| `hints` | `object` | SQL hints for query optimization. Optional configuration for advanced query tuning. |
## CLI Commands
Forge-SQL-ORM provides a command-line interface for managing database migrations and model generation.
**📖 [Full CLI Documentation](forge-sql-orm-cli/README.md)** - Complete CLI reference with all commands and options.
### Quick CLI Reference
The CLI tool provides the following main commands:
- `generate:model` - Generate Drizzle ORM models from your database schema
- `migrations:create` - Create new migration files
- `migrations:update` - Update existing migrations with schema changes
- `migrations:drop` - Create migration to drop tables
### Installation
The CLI tool must be installed as a local dependency and used via npm scripts in your `package.json`:
```bash
npm install forge-sql-orm-cli -D
```
### Setup npm Scripts
Add the following scripts to your `package.json`:
```bash
npm pkg set scripts.models:create="forge-sql-orm-cli generate:model --output src/entities --saveEnv"
npm pkg set scripts.migration:create="forge-sql-orm-cli migrations:create --force --output src/migration --entitiesPath src/entities"
npm pkg set scripts.migration:update="forge-sql-orm-cli migrations:update --entitiesPath src/entities --output src/migration"
```
### Basic Usage
After setting up the scripts, use them via npm:
```bash
# Generate models from database
npm run models:create
# Create migration
npm run migration:create
# Update migration
npm run migration:update
```
**Note:** The CLI tool is designed to work as a local dependency through npm scripts. Configuration is saved to `.env` file using the `--saveEnv` flag, so you only need to provide database credentials once.
For detailed information about all available options and advanced usage, see the [Full CLI Documentation](forge-sql-orm-cli/README.md).
## Web Triggers for Migrations
Forge-SQL-ORM provides web triggers for managing database migrations in Atlassian Forge:
### 1. Apply Migrations Trigger
This trigger allows you to apply database migrations through a web endpoint. It's useful for:
- Manually triggering migrations
- Running migrations as part of your deployment process
- Testing migrations in different environments
```typescript
// Example usage in your Forge app
import { applySchemaMigrations } from "forge-sql-orm";
import migration from "./migration";
export const handlerMigration = async () => {
return applySchemaMigrations(migration);
};
```
Configure in `manifest.yml`:
```yaml
webtrigger:
- key: invoke-schema-migration
function: runSchemaMigration
security:
egress:
allowDataEgress: false
allowedResponses: