FastXLSX

A lightweight, high-performance Python library for blazing-fast XLSX I/O operations.
Powered by Rust's Calamine (reading) and Rust-XlsxWriter (writing), with seamless Python integration via PyO3.

Docs available Here.

✨ Key Features

✅ Supported Capabilities

• Data Types: Native support for bool, int, float, date, datetime, and str.
• Data Operations: Scalars, rows, columns, matrices, and batch processing.
• Coordinate Systems: Dual support for A1 (e.g., B2) and R1C1 (e.g., (2, 3)) notation.
• Parallel Processing: Multi-threaded read/write operations for massive datasets.
• Type Safety: Full type hints and IDE-friendly documentation.
• Blasting Performance: 5-10x faster compared to openpyxl.

🚫 Current Limitations

• File Formats: Only XLSX (no XLS/XLSB support).
• Formulas & Styling: Cell formulas, merged cells, and formatting not supported.
• Modifications: Append/update operations on existing files unavailable.
• Advanced Features: Charts, images, and other advanced features not supported.

🏆 Performance Benchmarks

Tested on AMD Ryzen 7 5600X @ 3.7GHz (Ubuntu 24.04 VM) using pytest-benchmark.
Full details could be obtained from benchmarks.

📝 Writing Performance (Lower is Better)

library	Mixed Data (ms)	5000x10 Matrix (ms)	Batch Write (ms)
fastxlsx	0.97(1.00x)	62.06(1.00x)	7.77(1.00x)
pyexcelerate	2.65(2.73x)	256.89(4.14x)	50.33(6.48x)
xlsxwriter	5.03(5.19x)	297.14(4.79x)	61.25(7.89x)
openpyxl(write_only)	5.91(6.09x)	422.22(6.80x)	83.89(10.80x)
openpyxl	6.25(6.44x)	737.30(11.88x)	83.65(10.77x)

📖 Reading Performance (Lower is Better)

library	Mixed Data (ms)	5000x10 Matrix (ms)	Batch Write (ms)
fastxlsx	0.24(1.00x)	24.22(1.00x)	3.14(1.00x)
python-calamine	0.32(1.30x)	33.51(1.38x)	28.25(8.99x)
openpyxl	3.93(16.07x)	330.63(13.65x)	62.71(19.96x)

⚠️ Windows Users Note: Batch operations use multiprocessing.Pool, which may underperform due to spawn method limitations.

🛠️ Installation

PyPI Install

pip install fastxlsx

Source Build (Requires Rust Toolchain)

git clone https://github.com/shuangluoxss/fastxlsx.git
cd fastxlsx
pip install .

🚀 Quick Start Guide

Writing

import datetime
import numpy as np
from fastxlsx import DType, WriteOnlyWorkbook, WriteOnlyWorksheet, write_many

# Initialize workbook
wb = WriteOnlyWorkbook()
ws = wb.create_sheet("sheet1")

ws.write_cell((0, 0), "Hello World!")
ws.write_cell((1, 0), True, dtype=DType.Bool)
ws.write_cell("B1", datetime.datetime.now(), dtype=DType.DateTime)
ws.write_row((4, 2), ["var_a", "var_b", "var_c"], dtype=DType.Str)
ws.write_column((4, 0), [2.5, "xyz", datetime.date.today()], dtype=DType.Any)
# If `dtype` is one of [DType.Bool, DType.Int, DType.Float], must pass a numpy array
ws.write_matrix((5, 2), np.random.random((3, 3)), dtype=DType.Float)

# Save to file
wb.save("./example.xlsx")

# Write multiple files in parallel
workbooks_to_write = {}
for i_workbook in range(10):
    ws_list = []
    for i_sheet in range(6):
        ws = WriteOnlyWorksheet(f"Sheet{i_sheet}")
        ws.write_cell("A1", 10 * i_workbook + i_sheet, dtype=DType.Int)
        ws.write_matrix((1, 1), np.random.random((3, 3)), dtype=DType.Float)
        ws_list.append(ws)
    workbooks_to_write[f"example_{i_workbook:02d}.xlsx"] = ws_list
write_many(workbooks_to_write)

Reading

from fastxlsx import DShape, DType, RangeInfo, ReadOnlyWorkbook, read_many

# Load xlsx file
wb = ReadOnlyWorkbook("./example.xlsx")
# List all sheet names
wb.sheetnames
# Get a worksheet by index or name
ws = wb.get_by_idx(0)
# Read a single cell, notice the index is 0-based
print(ws.cell_value((0, 0)))
print(ws.cell_value("B1", dtype=DType.DateTime))
# Read a column with `read_value` and `RangeInfo`
print(ws.read_value(RangeInfo((4, 0), DShape.Column(3), dtype=DType.Any)))
print(
    ws.read_values(
        {
            "var_a": RangeInfo((5, 2), DShape.Column(3), dtype=DType.Float),
            "matrix": RangeInfo((5, 2), DShape.Matrix(3, 3), dtype=DType.Float),
        }
    )
)

# Read multiple sheets
print(wb.read_worksheets({"sheet1": [RangeInfo((2, 2), DShape.Scalar())]}))
# Read multiple files in parallel
print(
    read_many(
        {
            f"./example_{i_workbook:02d}.xlsx": {
                f"Sheet{i_sheet}": [
                    RangeInfo((0, 0), DShape.Scalar()),
                    RangeInfo((1, 1), DShape.Matrix(3, 3)),
                ]
                for i_sheet in range(6)
            }
            for i_workbook in range(10)
        }
    )
)

For more details, see Advanced Usage.

📖 Motivation

As is well known, Excel is not a good format for performance, but due to its widely used nature, sometimes we have to handle massive XLSX datasets. When I do some postprocessing work in Python, a lot of time is wasted on reading and writing; and when I tried to speed it up by parallelization, the spawn feature in Windows disturb me again. Therefore, I decided to develop a xlsx read-write library with Rust+PyO3 to solve that.

Thanks to the high performance of calamine and rust_xlsxwriter, as well as the great work of PyO3 and maturin, it is possible to do that by just binding them together with Python. Also thanks to the help of Deepseek enable me, a Rust beginner, could finish that.

📌 Future Plans

• Add support for formula and cell formatting
  rust_xlsxwriter supports formula and cell formatting well so that is not too hard to implent them into fastxlsx. But personally, when I export a large amount of data, format is usually not important, so the priority of this item is not high.
• Improve error handling