RS-21803: Header Visibility and Cell Formatting test

[SurreyHughesDisplayr]

Summary

Adds unit test coverage for header visibility and cell formatting in CreateCustomTable (R/createcustomtable.R) — 18 test blocks covering the plan's 19 scenarios.

• show.row.headers / show.col.headers off, including the force-disable path when row or column names are absent, each with a positive render control so a block cannot pass on empty output.
• Cell class assignment pinned as whole contiguous rows, e.g. <tr><td class="celldefault1">1</td><td class="celldefault5">5</td><td class="celldefault9">9</td></tr>, so inserting or reordering cells fails the test.
• row.classes / col.classes applied without index shift — asserted positively on the data cells and negatively on the row-header cell, which is the actual index-shift pin, with a row-count guard so the extraction cannot pass vacuously.
• Named vs positional col.classes list access produce identical body HTML, compared across the whole post-</thead> block plus a positive control.
• font.size, alignment padding and per-cell fill rules pinned as exact counts and exact declarations.

Plan corrections

Three of the plan's assertions could not be used as written and were corrected after probing real output:

• "Assert <th does not appear" is unusable — <thead> substring-matches <th>. Anchored to <th class=" instead (0 with headers off, 4 with them on).
• "Assert no 2px" is unusable — 2px legitimately appears in border-spacing: 2px 2px. Narrowed to font-size: 2px absent versus font-size: 2em present.
• Whole-document comparison for named vs positional access is not byte-stable, because each call generates a fresh random container-name stem. Compared the post-</thead> body region, which is stem-free.

Note on cell.inline.styl

createcustomtable.R contains cell.inline.styl <- rbind(...), assigning to a variable that is never read anywhere in the package — a dead store. It looks like a typo for the inline-styles matrix, but renaming it does not fix anything: patching the body to use the apparently-intended name makes every call fail with arguments cannot be recycled to the same length, because the styles matrix never gains a header row. The correct disposition is deletion of the line, and current emitted output is correct. The test therefore pins the real behaviour — the sig.change.fills inline style appears only on the flagged body cell and never in <thead> — rather than describing this as a behavioural defect.

Jira: https://numbers.atlassian.net/browse/RS-21803

Stacked PR

Based on RS-21803-bandedrows (#69), not master — these plans all add tests to tests/testthat/test-createcustomtable.R and share its helpers. Merge order: #66, #67, #68, #69, then this. The diff against that base is this plan's work only.

Test plan

Rscript -e "devtools::load_all('.'); testthat::test_file('tests/testthat/test-createcustomtable.R')"

212 passing, 0 failures, 0 warnings.

🤖 Generated with Claude Code