Benchmarks Guide

This guide covers running and creating performance benchmarks for AG Grid.

Overview

Performance benchmarks help detect regressions and validate optimizations.

Two types of performance testing exist in AG Grid:

1. Behavioural Benchmarks (Vitest) - Located in testing/behavioural/ - micro-benchmarks for specific operations
2. Performance E2E Tests (Playwright) - Located in testing/performance/e2e/ - end-to-end performance scenarios

Running Benchmarks

# Run behavioural benchmarks
yarn nx benchmark ag-behavioural-testing

# Run specific benchmark file
yarn nx benchmark ag-behavioural-testing -- src/tree-data/datapath/benchmarks/tree-data-path.bench.ts

# Watch mode for development
yarn nx benchmark ag-behavioural-testing --configuration=watch

# Update benchmark baselines
yarn nx benchmark ag-behavioural-testing --configuration=update

Key Performance Areas

Grid Rendering

• Initial render time
• Virtual scrolling performance
• Row/column update speed

Data Operations

• Sorting performance
• Filtering speed
• Grouping operations

Memory Usage

• Memory footprint with large datasets
• Memory cleanup after destroy

Writing Benchmarks

When creating new benchmarks:

1. Focus on realistic scenarios
2. Use consistent data sizes
3. Measure both time and memory
4. Document expected baseline values
5. Run multiple iterations for accuracy

Debugging Performance

For watch mode during development:

yarn nx benchmark ag-behavioural-testing --configuration=watch

Best Practices

1. Run benchmarks before and after changes
2. Use production builds for accurate results
3. Close other applications during benchmarking
4. Document any environmental factors
5. Track trends over time, not just absolute values