test(printplay): #204 extract recto-verso back reorder to pure method + 9 tests

[jsboige]

What & why

The Print & Play PDF (PrintAndPlayDocument) prints card backs on the reverse of each sheet so a sheet can be flipped to read the back. To make each back line up behind its matching front when the sheet is turned along its horizontal edge, the backs of each grid row are reversed before rendering — a per-row reversal (not a full-array mirror), because a horizontal flip swaps left/right within each row while preserving row order.

This fragile alignment contract was inlined in Compose() with zero unit coverage:

var backCardsArray = pageBackImages.ToJaggedArray(nbColumns)
    .Select(row => row.Reverse().ToArray()).ToArray().Flatten();

It could only be exercised by rendering a full QuestPDF document and visually inspecting the back alignment. A regression here (e.g. reversing the whole array instead of per-row, or flipping rows too) produces printed sheets whose backs mismatch their fronts — unusable sheets, only caught at print time.

Change

Output-neutral extraction + additive test coverage. No behavior change.

1. Extracted the inline composition into a pure, deterministic, generic static method:

   public static T[] ReorderBacksForRectoVerso<T>(IList<T> backs, int nbColumns)
       => backs.ToJaggedArray(nbColumns)
               .Select(row => row.Reverse().ToArray())
               .ToArray()
               .Flatten();

   The call site (Compose()) now reads ReorderBacksForRectoVerso(pageBackImages, nbColumns) — identical computation, just unit-testable.

2. Added PrintAndPlayRectoVersoContractTests (9 tests) pinning the contract:

   • Headline — full rows: 6 backs / 3 cols → [B2,B1,B0,B5,B4,B3] (per-row reversal, row order preserved; explicitly rejects a full-array mirror [B5..B0]).
   • Trailing short row — the lone card on a partial row reverses to itself and stays at the tail.
   • Single column — horizontal flip is identity (1-wide row reverses to itself).
   • One short row / width-2 / empty / single-back — degenerate cases (no crash, correct identity).
   • byte[] grounding — proves the generic works for the production element type (card image bytes), not just string labels.
   • Cell-wise alignment semantic — rebuilds input + output grids and asserts output[row][col] == input[row][cols-1-col] for every cell (the precise within-row mirror).

The building blocks (ToJaggedArray, Flatten) are already pinned by UtilityExtensionsLayoutTests; this PR pins the Print&Play-specific composition — the gap that test file does not cover.

Verification

• New class: 9/9 pass.
• Full suite on clean master base: 291 passed / 0 failed / 5 skipped (5 skips = Freeplane GUI + others requiring interactive session) — no regression.
• Diff: +1 pure static method (output-neutral) + 1 test file. No CSV, no config, no RowsetNb/rscount changes, no workflow/rules touched.

Scope

Dispatch #204 γ — test-coverage expansion on fragile pipeline contracts. Pivot sanctioned by dispatch qbw8vq ("sinon pivote PdfManager layout math"). Additive only; no production behavior change.

🤖 Worker po-2024

[clusterManager-Myia]

[NanoClaw]

test(printplay): extract recto-verso back reorder to pure generic method + 8 tests.

Extraction: The inline backs.ToJaggedArray(nbColumns).Select(row => row.Reverse().ToArray()).ToArray().Flatten() is pulled into a generic ReorderBacksForRectoVerso<T> — output-neutral, call site preserves exact computation.

Tests: 8 Fact tests covering: full evenly-divisible rows, trailing short row, single column, short row within larger column count, even width-2, empty backs, single back, production byte[] element type, and a cell-wise alignment contract (output[row][col] == input[row][cols-1-col]). Well-documented doc comments explain the per-row reversal semantics vs full mirror.

No secrets, no issues. LGTM.