One world, many maps, zero setup

Maps with Python is an interactive field guide to map projections. Grab a globe, flatten it, and watch every projection bend the truth differently — with the geodesy computed live by NumPy running in your browser. This is how it came together, and why "no Python setup needed" was the whole point.

A map is an optimization problem

Every flat map of our 3D world forces a choice: every projection preserves something and changes something else. Mercator was designed in 1569 for navigation — a sailor needs a steady bearing, so directions are true and areas near the poles balloon. A reader comparing countries needs fair areas, which is the job of equal-area projections like Gall–Peters and Equal Earth. AuthaGraph abandons the flat rectangle entirely, folding the world onto a tetrahedron.

You can state that tradeoff. You can also let someone feel it: drag a real globe, flatten it, and watch the distortion move to wherever you're not looking. That's the site.

Follow the circles

The distortion overlay descends from Tissot's indicatrix — the idea of drawing the same small circle on Earth in many places and seeing what the projection does to it. A projection can enlarge the circle, shrink it, or stretch it into an ellipse. These marks make the tradeoffs visible in a way no table of projection parameters can.

One subtlety worth confessing in the UI itself: on the interactive globe, circles near the rim look flattered purely from perspective — you're seeing them at an angle. That's not flat-map distortion. The site separates the two so the indicatrix teaches the right lesson.

The stack: real Python, in the page

• Pyodide + NumPy — the projection mathematics runs on genuine NumPy inside the browser tab. No server, no install, no notebook to start. The same arrays and formulas you'd write in a script execute locally at interactive rates.
• Three.js — the grabbable globe, texture-mapped with NASA Blue Marble, with Natural Earth 110m and 50m vector data for coastlines, lakes, and place names.
• KaTeX — the mathematics is typeset properly, because projection formulas deserve better than ASCII approximations.
• Cited all the way down — Equal Earth follows Šavrič, Patterson & Jenny (2018/2019); the AuthaGraph formulation follows Narukawa's 2022 paper, which approximates the original hand-built tetrahedron with four cones at under 4% radial deviation; distortion metrics draw on the USGS projection manual and Goldberg & Gott's flexion and skewness work.

Why zero setup matters

A field guide that requires an environment is a field guide most people never open. Running the math in-browser means the site works on a phone, works offline once loaded, and never asks anyone to trust a server with their session. It also makes a quiet argument we care about: Python is a good language for explaining things, not just for building products. A few lines of NumPy can change how you see the world.

Why this matters now

Projection choice isn't a settled historical curiosity. In September 2026 the UN General Assembly adopted a resolution championed by Togo encouraging equal-area projections for general-reference world maps, and the African Union followed with a statement welcoming it. Mercator's 450-year run as the default world map is finally being examined by the people it has always misrepresented. A site about what projections optimize for turned out to be timely in a way we didn't plan.