rustypyxl 0.3.0

A fast, Rust-powered Excel xlsx library for Python with openpyxl-compatible API

rustypyxl

A Rust-powered Excel (XLSX) library for Python with an openpyxl-compatible API.

Installation

pip install rustypyxl

Usage

import rustypyxl

# Load a workbook
wb = rustypyxl.load_workbook('input.xlsx')
ws = wb.active

# Read values
value = wb.get_cell_value('Sheet1', 1, 1)

# Write values
wb.set_cell_value('Sheet1', 1, 1, 'Hello')
wb.set_cell_value('Sheet1', 2, 1, 42.5)
wb.set_cell_value('Sheet1', 3, 1, '=SUM(A1:A2)')

# Bulk operations
wb.write_rows('Sheet1', [
    ['Name', 'Age', 'Score'],
    ['Alice', 30, 95.5],
    ['Bob', 25, 87.3],
])

data = wb.read_rows('Sheet1', min_row=1, max_row=100)

# Save
wb.save('output.xlsx')

Features

• openpyxl-compatible API: Familiar patterns for easy migration
• Read and write support: Full round-trip capability
• Cell values: Strings, numbers, booleans, dates, formulas
• Formatting: Fonts, alignment, fills, borders, number formats
• Workbook features: Comments, hyperlinks, named ranges, merged cells
• Sheet features: Protection, data validation, column/row dimensions
• Parquet import/export: Direct Parquet ↔ Excel conversion (bypasses Python FFI)
• S3 support: Works with boto3 via bytes I/O
• Bytes I/O: Load from bytes or file-like objects, save to bytes
• Configurable compression: Trade off speed vs file size

Parquet Import

Import large Parquet files directly into Excel worksheets. Data flows from Parquet → Rust → Excel without crossing the Python FFI boundary, making it very fast for large datasets.

import rustypyxl

wb = rustypyxl.Workbook()
wb.create_sheet("Data")

# Import parquet file into sheet
result = wb.insert_from_parquet(
    sheet_name="Data",
    path="large_dataset.parquet",
    start_row=1,
    start_col=1,
    include_headers=True,
    column_renames={"old_name": "new_name"},  # optional
    columns=["col1", "col2", "col3"],  # optional: select specific columns
)

print(f"Imported {result['rows_imported']} rows")
print(f"Data range: {result['range_with_headers']}")

wb.save("output.xlsx")

Performance: ~4 seconds for 1M rows × 20 columns on M1 MacBook Pro.

Parquet Export

Export worksheet data to Parquet format with automatic type inference:

import rustypyxl

wb = rustypyxl.load_workbook("data.xlsx")

# Export entire sheet
result = wb.export_to_parquet(
    sheet_name="Sheet1",
    path="output.parquet",
    has_headers=True,              # first row contains headers
    compression="snappy",          # snappy, zstd, gzip, lz4, none
    column_renames={"old": "new"}, # optional: rename columns
    column_types={"date_col": "datetime"},  # optional: force column types
)

print(f"Exported {result['rows_exported']} rows")
print(f"File size: {result['file_size']} bytes")

# Export specific range
result = wb.export_range_to_parquet(
    sheet_name="Sheet1",
    path="subset.parquet",
    min_row=1, min_col=1,
    max_row=1000, max_col=5,
)

Supported column type hints: string, float64, int64, boolean, date, datetime, auto.

Loading from Bytes or File-like Objects

Load workbooks from in-memory bytes or file-like objects:

import rustypyxl
import io

# From bytes
with open("file.xlsx", "rb") as f:
    data = f.read()
wb = rustypyxl.load_workbook(data)

# From file-like object (e.g., BytesIO, HTTP response)
wb = rustypyxl.load_workbook(io.BytesIO(data))

# Save to bytes (for HTTP responses, S3, etc.)
output_bytes = wb.save_to_bytes()

S3 Support

Use save_to_bytes() and load_workbook(bytes) with boto3 for S3 integration:

import boto3
import rustypyxl

s3 = boto3.client("s3")

# Load from S3
response = s3.get_object(Bucket="my-bucket", Key="path/to/file.xlsx")
wb = rustypyxl.load_workbook(response["Body"].read())

# Save to S3
data = wb.save_to_bytes()
s3.put_object(Bucket="my-bucket", Key="path/to/output.xlsx", Body=data)

This works with any S3-compatible service and uses boto3's credential handling (IAM roles, environment variables, etc.).

Streaming Writes (Low Memory)

For very large files, use WriteOnlyWorkbook which streams rows directly to disk:

import rustypyxl

wb = rustypyxl.WriteOnlyWorkbook("large_output.xlsx")
wb.create_sheet("Data")

for i in range(1_000_000):
    wb.append_row([f"Row {i}", i, i * 1.5, i % 2 == 0])

wb.close()  # Must call close() to finalize the file

This uses minimal memory regardless of file size, similar to openpyxl's write_only=True mode.

Benchmarks

Micro benchmarks on M1 MacBook Pro. Your results may vary depending on data characteristics and hardware.

Write Performance (1M rows × 20 columns)

Library	Time
rustypyxl	~10s
openpyxl	~200s

Read Performance

Dataset	rustypyxl	calamine	openpyxl
10k × 20 (numeric)	0.08s	0.10s	0.51s
10k × 20 (strings)	0.10s	0.12s	1.23s
100k × 20 (numeric)	0.97s	1.03s	4.74s
100k × 20 (mixed)	1.20s	1.28s	12.1s

calamine is a Rust Excel reader with Python bindings via python-calamine (read-only).

Memory Usage (Read)

Dataset	rustypyxl	calamine	openpyxl
10k × 20	29 MB	9 MB	11 MB
50k × 20	58 MB	48 MB	53 MB
100k × 20	95 MB	95 MB	106 MB

Memory Usage (Write)

Dataset	rustypyxl	WriteOnlyWorkbook	openpyxl (write_only)
10k × 20	10 MB	~0 MB	0.4 MB
50k × 20	50 MB	~0 MB	0.4 MB
100k × 20	99 MB	~0 MB	0.4 MB

WriteOnlyWorkbook streams rows directly to disk like openpyxl's write_only mode.

Building from Source

# Install Rust and maturin
curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh
pip install maturin

# Build
maturin develop --release

License

MIT