Skip to content

Benchmarks

This page documents performance benchmarks for sheets.

The goal is to evaluate:

  • Parsing speed
  • Writing speed
  • Memory efficiency

and compare against Python’s built-in csv module.

Benchmark Setup

Benchmarks were run using a synthetic CSV file with:

  • 1,000,000 rows
  • 20 columns per row
  • Mixed field types:
  • integers
  • floats
  • strings
  • quoted strings
  • Approximate file size: 200–300 MB

Example row structure:

1
2
id,count,price,name,quoted,city,...
1,2,0.0100,user_1,"quoted,value",city_1,...

Test Operations

Three main benchmarks were measured.

1. Parse Benchmark

Measures:

  • Reading CSV file
  • Parsing all rows
  • Fully loading into memory

Equivalent to:

1
local rows = csv.read_csv("benchmark_1m.csv")

2. Write Benchmark

Measures:

  • Serializing rows into CSV text

Equivalent to:

1
local output = csv.write(rows)

Repeated 100,000 times for stable measurements.

3. Memory Benchmark

Measures:

  • Memory usage before parse
  • Memory usage after full parse

This represents total in-memory footprint after loading all rows.

Benchmark Results

Metric sheets Python csv
Parse Time ~9.2s ~6.7s
Memory Usage ~644 MB ~1261 MB
Write Time ~0.17s ~0.59s

Benchmark Analysis

Parsing Performance

Library Parse Time
Python csv ~6.7s
sheets ~9.2s

Python’s built-in csv module parses faster due to highly optimized C internals and decades of optimization.

However, sheets remains competitive while offering excellent memory efficiency.

Memory Efficiency

Library Memory Usage
Python csv ~1261 MB
sheets ~644 MB

Result

sheets uses approximately:

  • ~49% less memory
  • Nearly 2× lower memory usage

This is one of the strongest advantages of sheets.

For large datasets, lower memory usage can significantly improve stability and scalability.

Writing Performance

Library Write Time
Python csv ~0.59s
sheets ~0.17s

Result

sheets achieves approximately:

  • ~3.5× faster writing performance

This makes it particularly effective for:

  • Data export pipelines
  • CSV generation workloads
  • Reporting systems

Key Takeaways

sheets provides:

  • Competitive parsing speed
  • Excellent memory efficiency
  • Outstanding write performance

Main strengths:

  • ~2× lower memory usage than Python csv
  • ~3.5× faster writing
  • High-performance C backend using libcsv

Benchmark Scripts

Benchmark scripts are included in the repository.

Lua benchmark:

1
examples/comparison.lua

Python benchmark:

1
test/python_csv_bench.py

Notes

Benchmark results may vary depending on:

  • CPU
  • RAM
  • OS
  • Lua version
  • Compiler optimization flags

These benchmarks are intended to provide practical comparative insight rather than absolute performance guarantees.