serverpod-offline-sync/legacy/drift_offline_sync/benchmark/README.md

Benchmarks

This folder contains benchmarks to measure the performance impact of using the drift_offline_sync package on database operations.

What is Measured

The benchmarks measure two key metrics:

1. Insert/Update/Delete Performance: Time required to insert/update/delete a large number of rows with many columns.
2. Storage Overhead: Database file size after inserting a large number of rows with many columns.

The benchmarks are run in a separate isolate to avoid interference with the main process. A baseline benchmark is run first to establish a reference point and then a CRDT-enabled benchmark is run to measure the performance impact of using the drift_offline_sync package.

Running the Benchmarks

To run the benchmarks:

dart benchmark/run.dart

Accessing Benchmark Results

When running from the CI, benchmark results are automatically stored as git notes on commits. If running on a pull-request, it is also pasted as a comment that is kept up to date on each push. To access historical benchmark data:

# Fetch benchmark notes from remote (if not already available)
git fetch origin refs/notes/benchmarks:refs/notes/benchmarks

# View benchmark results for the latest commit
git notes --ref=benchmarks show

# View benchmark results for a specific commit
git notes --ref=benchmarks show <commit-sha>

# View all commits with their benchmark results
git log --show-notes=benchmark

Benchmark Details

Table Structure

The benchmarks use TableWithEveryColumnType which includes:

• Boolean (aBool)
• DateTime (aDateTime)
• Text (aText)
• Integer (anInt)
• Int64 (anInt64)
• Real/Double (aReal)
• Blob (aBlob)
• Integer Enum (anIntEnum)
• Text with Converter (aTextWithConverter)
• UUID (aUuid)

This comprehensive table structure ensures the benchmarks reflect realistic use cases with various data types.