Examples

Every example here starts from one ordinary Julia file carrying #~ metaLines in its comments. The smallest examples pair each input with the one rendered share-view it derives; the richest derive several outputs from a single source — each with its own view of what is shown, what stays hidden, and what runs. Two kinds of control are on display: control of rendering — which content each derived view carries — and control of execution — which code a notebook or a docs build actually runs.

The smallest examples are the repository's committed corpus: seven input/output pairs under examples/ that render byte-identically, pinned by the shipped 1140-test suite at this release. Verify every pair yourself from a repository checkout:

julia --startup-file=no --project=. -e 'using Pkg; Pkg.instantiate()'   # once
julia --startup-file=no --project=. run_examples.jl

The notebooks pipeline

One source, many outputs. notebooks_from_source.jl, at the repository root, derives a whole notebook family from ONE marked Julia source file — purely from its #~ marks. From notebooks/src/montecarlo.jl — an ordinary, runnable Julia file (include it) — it emits four notebook editions; a standard jupyter nbconvert --execute run then produces the executed twin:

• full — every surviving cell, with legend-driven tags (parameters, skip-execution, solution, deep-dive);

• student — the worksheet: the solution cell is absent, a scaffold cell stands in, and instruction cells are locked;

• report — the reading view: every code input folded;

• slides — a slide type per cell, scraped from the source's depth marks; jupyter nbconvert --to slides turns it into a reveal.js deck;

• executed — the one edition with real outputs, from a standard jupyter nbconvert --execute run.

Each line's fate comes from the engine's own evaluated verdicts — the :visib rows of GoMeta.altValues_evals — never re-derived from the render. A verdict-hidden line is never visible in a rendered view, and never lost: it travels with the notebook in source form (an HTML comment in markdown cells, gometa cell metadata in code cells), so the derivation is a format transformation, not a projection. A discarded line travels nowhere. Hidden lines do not execute in notebooks.

A second source ships the same way: notebooks/src/quickstart.jl runs as a script from the checkout root and as its derived notebook — the onboarding artifact demonstrates the pipeline it teaches.

Every generated artifact regenerates byte-identically (the executed twin, which carries real kernel outputs, is validated separately, outside that byte gate):

julia --startup-file=no --project=. notebooks_from_source.jl --check

Documenter pages as views

The same marks in the same source also derive a pair of Documenter manual pages — docs/montecarlo-full.md and docs/montecarlo-reader.md. The pages are views: verdict-hidden lines are simply not on them (the one exception is Documenter's own executed blocks — # hide-marked @example lines and @setup bodies — which the docs build runs and display-hides while they remain present in the raw page markdown), and metaLines do not ride them. The #~ marks additionally decide which jldoctest blocks each page carries — the docs build runs exactly the tests the marks kept. Rendering control and execution control come from the same metadata.

Jupyter notebook example

Coming at launch

A full worked Jupyter example lands here at launch — it is being finalized now. In the meantime, the complete notebook family ships in the repository under notebooks/, with a README that explains every artifact and how to regenerate it.

Documenter.jl example

Coming at launch

A full worked Documenter.jl example lands here at launch — it is being finalized now. In the meantime, the derived manual-page pair ships in the repository under docs/, generated from notebooks/src/montecarlo.jl by the shipped generator at the repository root.

New to the syntax? Start with What is GoMeta? and metaLines, or install via Getting started.