Customising Jupyter with an engineering DSL

A Jupyter notebook cell is Python — but it doesn't have to look like Python. With an import hook you can rewrite the source of every cell before the interpreter sees it, which means the notation you use on paper can become real, executable code.

I've been building this out as Engineering DSL, a package that turns a notebook into something close to an engineer's worksheet. One import activates it:

from utils.Engineer import *

From then on, cells accept units, := assignment, · multiplication, ‖ for parallel resistors, √, superscript powers, and subscript indexing:

V_in  := 12.0 V
R_top := 4.7 kΩ
R_bot := 10. kΩ
V_out := V_in · R_bot/(R_top + R_bot)     # 8.2 V

R_eq := 100. Ω ‖ 220. Ω ‖ 470. Ω          # parallel resistors
τ := 10. kΩ · 100. nF                     # → 1.0 ms, units simplify
hyp := √(30.cm² + 40.cm²)                 # → 50 cm

The trick is a source-transform hook (via ideas) that rewrites each cell before Python compiles it — no string parsing, no magic cells. Units ride on forallpeople, so 12 V / 3 A really is 4 Ω and results auto-scale their SI prefix. Numeric literals carry their significant figures through arithmetic, matrices are live sympy objects with LaTeX rendering, and plot() reads the units off the data to label its own axes.

The repo includes notebook manuals and a gallery of one-cell examples — electrical, mechanical, fluids, symbolic solving, radix and Roman-numeral literals, even DKK currency conversion with live rates. If you write calculations for a living, the difference between V_out = V_in * R_bot / (R_top + R_bot) and the version above is the difference between transcribing your work and just writing it.

Code and notebooks: github.com/RichardThulstrup/engineering-dsl