Normalize all source files to NFC

[michakraus]

Follow-up to the review of #110, which turned up an encoding inconsistency that had been in the tree for a long time.

The problem

Accented identifiers in this repository had drifted into both of their legal spellings:

character	composed (NFC)	decomposed (NFD)
ū	U+016B, 2 bytes	u + U+0304 COMBINING MACRON, 3 bytes
ḡ	U+1E21, 3 bytes	g + U+0304, 3 bytes
ṗ	U+1E57, 3 bytes	p + U+0307, 3 bytes
ñ é	U+00F1 / U+00E9, 2 bytes	base + U+0303 / U+0301, 3 bytes

Julia's parser NFC-normalizes identifiers, so the two spellings are the same binding. Verified rather than assumed (with Ñ, whose two forms are the easiest to write out):

julia> nfd = "Ñ"; nfc = "Ñ"          # "4ecc83" vs "c391"
julia> eval(Meta.parse(nfd * "_probe = 7"))   # assign through the decomposed spelling
julia> eval(Meta.parse(nfc * "_probe"))       # read through the composed one
7
julia> filter(n -> occursin("_probe", string(n)), names(Main, all=true))
1-element Vector{Symbol}: :Ñ_probe            # one binding, not two

So a file mixing them compiles and runs identically, and no test of behaviour can distinguish them — which is exactly what let this sit. What they are not is the same text:

• grep ū typed one way silently finds nothing while the identifier sits plainly on screen;
• an exact-string edit fails for no visible reason;
• a diff shows the line as modified while rendering byte-for-byte identically, so encoding churn is indistinguishable from a real edit.

Files drift between the forms without anyone touching a character, because macOS filesystem APIs hand back decomposed text and some editors recompose on save.

The change

Seven files, 31 lines, 44 combining marks folded away:

file	characters
src/harmonic_oscillator.jl	ū ḡ ṗ
src/lotka_volterra_2d_common.jl	ū ḡ ṗ
src/lotka_volterra_2d_equations.jl	ū ḡ
src/lotka_volterra_2d_symmetric.jl	ñ
src/lotka_volterra_4d.jl	ṗ
README.md, docs/make.jl	é (in Hénon)

src/toda_lattice.jl was the eighth file until #110 landed: hand-writing the Toda vector fields rewrote its Ñ lines into NFC as a side effect, so this branch was rebased onto that and Ñ drops off the list. The rebase conflicted on exactly those seven lines and resolved in favour of #110's version, which is byte-identical to what normalization would have produced.

Every changed line is canonically equivalent to the one it replaces. Checked mechanically, not by eye: for each file, NFC(before) == after, and NFD(before) == NFD(after), so nothing but the normal form changed.

Only 15 bytes are removed rather than 44, because the composed forms of ṗ, ḡ and ẋ are themselves three-byte codepoints and cost exactly what their decomposed spellings did.

The guard

test/unicode_normalization_tests.jl keeps it this way. It asserts the normal form of each file rather than any particular character, so there is no list of accented identifiers to maintain, and it reports the offending paths rather than a count, so a failure says what to run normalize over.

Characters with no precomposed form pass untouched: q̇ is q + U+0307 and has no single codepoint, so it is already in normal form and no normalization pass can change it. obsolete/ is excluded, being kept for reference and not edited — its six files are the only decomposed text left in the tree.

Unicode is now a declared dependency in test/Project.toml, for the third time the same reason applies after Logging and LinearAlgebra: an undeclared stdlib resolves under julia --project=test but not inside the sandbox environment Pkg.test builds from that file, where using Unicode fails with "Package Unicode not found in current path". The pre-push hook caught this, which is the whole point of it.

Full Pkg.test() passes, as does the pre-push hook suite.

🤖 Generated with Claude Code

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[codecov]

Codecov Report

❌ Patch coverage is 27.27273% with 16 lines in your changes missing coverage. Please review.
✅ Project coverage is 75.22%. Comparing base (23f8c6d) to head (8a38e26).

Files with missing lines	Patch %	Lines
src/lotka_volterra_2d_common.jl	0.00%	9 Missing ⚠️
src/harmonic_oscillator.jl	0.00%	7 Missing ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #111   +/-   ##
=======================================
  Coverage   75.22%   75.22%           
=======================================
  Files          43       43           
  Lines        2442     2442           
=======================================
  Hits         1837     1837           
  Misses        605      605           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.