zs2fast 0.3.0

Fast Rust/PyO3 parser for ZS2 files with Parquet export

zs2fast

A fast Python extension for converting Zwick ZS2 test machine files to Parquet format, written in Rust/PyO3.

Features

• Fast gzip decompression and binary parsing of .zs2 files
• Extracts channel time-series data with automatic unit resolution
• Extracts evaluated parameters and test results with units
• Converts to Parquet format using Apache Arrow
• Built with PyO3 for seamless Python integration
• Supports Python 3.9–3.12 on Linux, macOS, and Windows

Installation

From PyPI

pip install zs2fast

From Source

Requires Rust and Python 3.9+.

# Install maturin
pip install maturin

# Clone and build
git clone https://github.com/weltidom/zs2fast.git
cd zs2fast
maturin develop --release

Usage

import zs2fast
import polars as pl

# Convert a .zs2 file to raw long-format Parquet
# Useful for exploring binary structure and raw values
zs2fast.zs2_to_parquet("input.zs2", "output.parquet", include_u32=False)
# columns: series, subtype, index, value

# Export channel time-series with semantic names and units
zs2fast.zs2_channels_to_parquet("input.zs2", "channels.parquet")
# columns: sample_idx, channel_idx, channel_name, unit, timepoint, value, data_type
channels = pl.read_parquet("channels.parquet")
print(channels.filter(pl.col('unit').is_not_null()).head())

# Export calculated test results (Bruchbild, elongation, max force, etc.)
# Units are resolved from EinheitName → QS_ValSetting blob → UnitTables
zs2fast.zs2_evaluated_params_to_parquet("input.zs2", "evaluated_params.parquet")
# columns: sample_idx, param_id, short_name, param_name, unit, value, value_text
params = pl.read_parquet("evaluated_params.parquet")
print(params.filter(pl.col('value').is_not_null()).head())

# Extract per-sample test settings and parameters (not calculated results)
zs2fast.zs2_parameterliste_results_to_parquet("input.zs2", "sample_params.parquet")
# columns: sample_id, result_id, result_name, unit, value_text, value
sample_params = pl.read_parquet("sample_params.parquet")
print(sample_params.head())

Unit Resolution

The parser automatically resolves measurement units using a 3-tier strategy:

1. Direct unit name (EinheitName field) — Direct unit symbol (mm, N, MPa, etc.)
2. Unit table key (from QS_ValSetting binary blob) — References /UnitTables section (UT_Length → mm, UT_Force → N, etc.)
3. Fallback inference — Attempts to infer units from table key name

Channels typically use tier 1–2, while evaluated parameters primarily use tiers 2–3. This ensures maximum unit coverage across different .zs2 file structures.

Development

Build

maturin build --release

Lint

cargo fmt
cargo clippy

Test

cargo test

License

See LICENSE file for details.