raoulg/atlas

What a program is, beneath the language it is written in.
See it ·
Install ·
For agents
Two programs can show you the same screen. One is sloppy code with lots of shortcuts and
inconsistent structure. The other is an elegant orchestration of logic. Atlas shows which is which.
The same program. The same figure.
Fetch sunspot counts from the web, keep the CSV, clean it, smooth it, draw a figure. Written twice in
Python: once as a sloppy script, once typed and elegant. Atlas draws what each one does.
What both make: the monthly mean sunspot number since 1749 (SILSO, Royal Observatory of Belgium), smoothed over 13 months.
A sloppy script
Typed and elegant
Both reach the same places: the web, the raw CSV, the figure. Inside they part. In the sloppy script
every step reads and changes one global, Global(sunspots.df); nothing is handed from step to step,
and get_data downloads on every run and reads back the CSV it has just written, so the file is
only what it leaves behind. In the typed one download runs only when no CSV is kept, in a frame
that says so (only when not raw.exists()), and load reads what is there, so the file stands
between the two; a DataFrame goes from load to smooth to Trend.save, and the command names
every file. A red dot is a question Atlas asks the author: why do get_data and plot pick their
own files, and what do the four functions give back? Click a picture to open it in Atlas.
Ported to Rust. A picture alike.
Code gets rewritten in another language: for speed, for one binary to ship, for a team, and more
and more often by an agent. Tests tell you the port gives the same figure. They don't tell you
whether the design survived: the same steps, the same values handed between them, the same places.
Atlas draws every language in the same terms, so the original and the port can lie side by side:
the same shape where the design survived, and differences where it did not, or where the language
works otherwise.
Python, typed
Rust, ported
The shape is the same: the web, the raw CSV and the figure, and between them download, load,
smooth and Trend.save, open here. The differences are the interesting part. A step gone or a
place moved would be drift, and Atlas would show where. Here it is the language speaking:
| | Python | Rust |
|---|---|---|
| **The values** | one pandas DataFrame, in and out of smooth | the program's own records: Month · series in, Smoothed · smoothed out |
| **The figure** | Trend.draw gives a Figure, and save writes it | save opens a DrawingArea on the file first, and draw draws straight into it |
| **Defaults** | in the signature | a function, Settings.default: Rust has no default arguments |
The mathematics of your code
Mathematics is a language that allows you to make things consistent. Best practices in coding are
consistent, so it makes sense to view your computer code through a mathematical perspective.
Because we can turn the mathematics of the code into a visual, you can literally see the
consistency of your code, if you followed best practices. This is independent of the language you
use: every program is drawn by the same mathematical rules.
**What is inside:**
*Category theory*
- **A category of types.** Types are objects, functions are arrows, call order is composition. A