Notebook Conventions

Setup Pattern

All notebooks should start with the standard setup cell:

using Revise
using Linear_Algebra

LAlatex.set_backend!(:symbolics)
LAlatex.reset_display_defaults!()

LAlatex, BlockArrays, and LaTeXStrings are re-exported by Linear_Algebra, so no separate using is needed. In particular, the L"..." string macro is available immediately after using Linear_Algebra. set_backend!(:symbolics) is required because this package uses Symbolics.jl; the default :latexify backend gives worse output for symbolic expressions. reset_display_defaults!() ensures a clean display state on every kernel restart.

Notebook Guidelines

• Notebooks are for exploration and study, not tested in CI

• The notebooks/ environment is not a workspace member — run pkg> dev .. + pkg> instantiate once after cloning. See julia-coding-conventions skill for the full pattern. Do not commit notebooks/Manifest.toml.

• Use println() for output to make results clear when re-running cells

• Include explanatory markdown cells between code cells

• Use Unicode variable names consistent with the source code (e.g., v₁, θ, λ)

• Reference the documentation sections being studied

Plotting in Notebooks

• Plots render inline in Jupyter notebooks

• No need for headless mode configuration

• Use the same plotting functions from the package (plot_param_line, etc.)

LAlatex Display

LAlatex.jl (by ea42gh) is re-exported by Linear_Algebra and provides clean LaTeX rendering of linear algebra objects in notebooks. Use it instead of raw println or display whenever presenting mathematical results.

Core functions

Common options for l_show

l_show(L"A = ", A; arraystyle=:bmatrix)          # bracket style: :bmatrix, :pmatrix, :vmatrix, :array
l_show(expr; symopts=(expand=true,))              # symbolic transformations: expand, factor, collect
l_show(A; number_formatter=x -> round_value(x,2)) # custom number formatting
l_show(A; number_formatter=percentage_formatter)  # percentage display
l_show(aligned(...); inline=false, tag="1")        # numbered display equation (no label — see below)
with_display_defaults(arraystyle=:bmatrix) do ... end  # scoped defaults

Block-partitioned matrices

Wrap a plain matrix in BlockArray to get visual partition lines (horizontal \hline and vertical | separators) between blocks:

B = BlockArray([1 2 4; 3 4 5], [1, 1], [2, 1])   # row sizes [1,1], col sizes [2,1]
display(l_show(L"B = ", B))

KaTeX constraint: no \label

Jupyter notebooks render math with KaTeX, which supports \tag{...} but not \label{...}. Never pass the label= keyword to l_show in a notebook — it will produce a parse error. Use tag= only:

# ✅ correct
display(l_show(aligned(...); inline=false, tag="1"))
# ❌ will fail in Jupyter
display(l_show(aligned(...); inline=false, tag="1", label="eq:foo"))

label= is only valid when rendering to full LaTeX/MathJax (e.g. Documenter.jl docs).

Reference

• LAlatex.jl docs

• Demo notebook (upstream)

• Interactive Binder demo

Julia Kernel Gotchas

See julia-coding-conventions skill (stale variable slot, building matrix from row vectors).