LayMesh

Reproducible scientific figures and precise physical layouts from readable code. Compose native plots, images, vector shapes, text and formulas in one editable .lay source; export SVG, PDF and PNG with a native Rust engine.

中文文档 · English documentation · Examples · Source

Install

Requires Python 3.10+. Version 0.3.0a2 is an alpha prerelease:

python -m pip install --pre laymesh
python -m laymesh --version

Platform wheels bundle one native Rust executable and formula fonts. Rendering requires no Rust compiler, Node.js or TeX installation and downloads no engine or fonts. Body fonts come from the operating system or user-provided files; missing glyphs warn and display vector boxes. Supply font files with a layout for reproducible text across machines.

Wheel targets are Windows x64, macOS 14+ on Intel and Apple Silicon, and Linux x64 / arm64. Linux requires glibc at least as new as the wheel's manylinux_2_XX tag. Availability depends on uploaded wheels; Alpine/musl, 32-bit platforms and Windows ARM64 have no wheels. Native plots work without Matplotlib.

First figure: CLI

Save this complete source as figure.lay:

page = canvas(size=(120 mm, 90 mm), background="#ffffff")
p = plot(size=(110 mm, 80 mm),
         x=axis(label="Time (s)"), y=axis(label="Signal"))
p.line(x=[0, 1, 2, 3], y=[1, 3, 2, 4], label="Experiment")
p.legend(position="top_left")
page.add(p, offset=(5 mm, 5 mm))

laymesh validate figure.lay
laymesh inspect figure.lay --json
laymesh render figure.lay -o figure.svg
laymesh render figure.lay -o figure.pdf
laymesh render figure.lay -o figure.png --dpi 300

python -m laymesh invokes the same CLI. Geometry defaults to mm; typography and line widths default to pt. PNG DPI changes output pixels while preserving physical dimensions. .lay is a restricted standalone language, separate from Python.

Python API

from laymesh import render_source

source = '''
page = canvas(size=(120 mm, 90 mm), background="#ffffff")
p = plot(size=(110 mm, 80 mm),
         x=axis(label="Time (s)"), y=axis(label="Signal"))
p.line(x=[0, 1, 2, 3], y=[1, 3, 2, 4], label="Experiment")
p.legend(position="top_left")
page.add(p, offset=(5 mm, 5 mm))
'''
result = render_source(source, output="python-figure.svg", save_source="python-figure.lay")
print(result.output)

Use render_file("figure.lay", output="figure.pdf") to render existing source. Pass Python variables via namespace={"values": values} and refer to them with an unquoted {{values}} placeholder in .lay. Lists, dictionaries, NumPy arrays and pandas DataFrames can feed native plots. Matplotlib Figures can be placed as image assets. Keep calculations in Python; placeholders accept variable names, not expressions.

save_source saves expanded source and a sibling .assets/ directory for independent CLI rendering. Preserve that directory and any referenced modules, stylesheets or fonts with the .lay file. Python API and data examples.

Jupyter

Install LayMesh in the active kernel's environment. First run:

%load_ext laymesh.ipython
values = [1, 3, 2, 4]

Then run this separate cell:

%%laymesh -o notebook.pdf --save-source notebook.lay
page = canvas(size=(120 mm, 90 mm))
p = plot(size=(110 mm, 80 mm))
p.line(x=[0, 1, 2, 3], y={{values}})
page.add(p, offset=(5 mm, 5 mm))

The magic displays an SVG preview and writes the requested export. %laymesh figure.lay -o figure.png --dpi 300 renders an existing file. Both magics support --warnings show|hide. Notebook guide.

Scope

LayMesh provides physical units, anchors, cropping, paths, reusable modules, LCSS styles and native scientific plots including statistical, polar and radar layers. Layouts are single-page; panels use explicit positioning. Finish plot layers and decorations before placing a chart. Fixed plot areas keep their physical dimensions; decoration overflow emits a warning. Language reference.

中文

LayMesh 用一个可编辑的 .lay 文件组织科研图表、图片、文字与公式，导出 SVG、PDF、PNG。需要 Python 3.10+；预发布版本使用 python -m pip install --pre laymesh，数据和 Matplotlib 支持可选 laymesh[data,plot]。平台 wheel 内置原生 Rust 引擎和公式字体，正文使用系统或用户字体。支持 laymesh 命令、Python API 和 Jupyter Magic；保存源码与资源后可独立重新导出。中文入门。

LayMesh is MIT licensed. Bundled native dependencies, formula fonts and colormap data retain their own license texts inside the installed package. _vendor/manifest.json records engine, platform and build hashes; _vendor/licenses/manifest.json records dependency license provenance.