Skip to content

aggregate.md

Latest commit

 

History

History
65 lines (49 loc) · 1.73 KB

aggregate.md

File metadata and controls

65 lines (49 loc) · 1.73 KB
Raw

(cli-aggregate)=

aggregate - Aggregate Results

Aggregate multiple benchmark result files into a CSV file for performance trend tracking and downstream analysis.

Basic Syntax

benchbox aggregate --input-dir <path> --output-file <path> [OPTIONS]

Options

Required:

  • --input-dir PATH: Directory containing result JSON files (searched recursively)
  • --output-file PATH: Output CSV file path

Optional:

  • --benchmark TEXT: Filter results by benchmark name
  • --platform TEXT: Filter results by platform name

Output Format

The CSV output contains one row per result file with these columns:

Column Description
timestamp Benchmark execution timestamp
benchmark Benchmark name
platform Platform name
scale_factor Scale factor used
total_time_s Total execution time in seconds
geometric_mean_ms Geometric mean of query times (ms)
p50_ms Median query time (ms)
p95_ms 95th percentile query time (ms)
p99_ms 99th percentile query time (ms)
num_queries Number of successful queries
file Source result filename

Usage Examples

# Aggregate all results in directory
benchbox aggregate --input-dir benchmark_runs/ --output-file trends.csv

# Filter by benchmark
benchbox aggregate \
  --input-dir benchmark_runs/ \
  --output-file tpch_trends.csv \
  --benchmark tpch

# Filter by platform
benchbox aggregate \
  --input-dir benchmark_runs/ \
  --output-file duckdb_trends.csv \
  --platform duckdb

Related

  • report - Historical analysis and trend detection from the result database
  • results - View individual benchmark results