Creating Publication-Quality Figures Using PyX

Automating Diagram Generation in Python with PyX

Diagrams—flowcharts, circuit sketches, publication figures—are easier to produce consistently when automated. PyX is a Python library that combines powerful drawing primitives with LaTeX-quality text rendering, making it ideal for programmatic, repeatable diagram generation. This article shows how to set up PyX, design reusable drawing components, and build automated pipelines to generate diagrams at scale.

Why choose PyX for automation

• Vector output: Produces PDF/EPS/SVG suitable for editing and publication.
• Programmable: Diagrams are generated from code, enabling repeatability, parameterization, and batch creation.
• LaTeX-quality text: Uses TeX for labels, so fonts and math match documents.
• Composability: Strong primitives (paths, groups, pens, brushes) let you build higher-level diagram components.

Setup and installation

1. Install PyX and its dependencies:
   • Ensure a TeX distribution is installed (e.g., TeX Live or MiKTeX).
   • Install PyX:
   pip install pyx

2. Verify TeX is available on PATH so PyX can render text.

Basic PyX concepts (brief)

• Canvas: drawing surface (pyx.canvas.canvas).
• Path: vector shape (move, line, curve).
• Stroke/Fill: pens and brushes to style paths.
• Text: pyx.text.text or canvas.text uses LaTeX.
• Transformations: translate, rotate, scale via context.

Simple example: a labeled flow node

from pyx import canvas, path, text, color, trafo c = canvas.canvas()# Draw a rounded rectangle nodex, y, w, h = 0, 0, 4, 2rect = path.rect(x, y, w, h)c.fill(rect, [color.rgb.white])c.stroke(rect, [color.rgb.black])# Add label with TeXc.text(x + w/2, y + h/2, r”\large\textbf{Process}“, [text.halign.center, text.valign.middle])c.writePDFfile(“flow_node”)

Designing reusable components

Encapsulate repeated diagram elements as functions or classes to allow parameterization.

Example: reusable flow-node factory

from pyx import canvas, path, text, color def draw_node(c, x, y, w, h, label, fill=color.rgb.white): rect = path.rect(x, y, w, h) c.fill(rect, [fill]) c.stroke(rect, [color.rgb.black]) c.text(x + w/2, y + h/2, label, [text.halign.center, text.valign.middle]) c = canvas.canvas()draw_node(c, 0, 0, 4, 2, r”\textbf{Start}“)draw_node(c, 5, 0, 4, 2, r”\textbf{End}“, fill=color.rgb(0.9,0.9,1))c.writePDFfile(“nodes”)

Connecting components: anchors and ports

Define anchor points on components so connectors can be computed programmatically.

• Compute anchor coordinates relative to component position.
• Use straight lines or Bezier curves for connectors.
• Add arrowheads using short paths or built-in arrow styles.

Example: connect two nodes with an arrow

from pyx import path def connect(c, x1, y1, x2, y2): p = path.path(path.moveto(x1, y1), path.lineto(x2, y2)) c.stroke(p, [pyx.style.linewidth.THick, pyx.deco.earrow.size.normal])

Layout strategies for automation

• Grid layout: place nodes by row/column indices.
• Graph-based layout: use networkx to compute positions, then render with PyX.
• Constraint-based: simple rules (align, equal spacing) implemented in code.

Example: using networkx for position computation

• Compute positions with networkx.spring_layout or shell_layout.
• Scale and translate positions to your canvas coordinates, then draw nodes and edges.

Batch generation and parameter sweeps

• Create scripts to loop over datasets or parameter combinations and output numbered files (diagram_01.pdf, diagram_02.pdf).
• Embed metadata in filename or diagram (small text) for traceability.
• Use multiprocessing for parallel rendering of many diagrams.

Integrating with LaTeX and documents

• Export as PDF/EPS and include with \includegraphics.
• Because PyX uses TeX for text, fonts and math match the document seamlessly.

Tips for robust automation

• Separate data, layout, and rendering code.
• Use version control for diagram-generating scripts.
• Cache expensive layout computations.
• Validate that TeX and required fonts are available in the runtime environment (CI, servers).
• Produce both vector (PDF/SVG) and raster (PNG) outputs when needed.

Example project structure

• diagrams/
  • templates.py (component factories)