Give gridded example cells their own country's area code

[lbm364dl]

What was wrong (mechanism)

Five cells in the gridded example = TRUE fixtures in R/toy_examples.R carried an area_code that contradicts their own lon/lat. Unlike the cases #417 fixed, these are valid FAOSTAT codes, so they resolve — to the wrong country, silently, in the output a reader sees first:

export	cell	fixture area_code	spatialize-country-grid says
read_luh2_landuse	9.25, 47.75	11 (Austria)	79 (Germany)
read_luh2_landuse	35.25, -1.25	79 (Germany)	114 (Kenya)
build_water_balance	9.25, 47.75 and 9.75, 47.75	11 (Austria)	79 (Germany)
build_water_balance	35.25, -1.25 and 35.75, -1.25	79 (Germany)	114 (Kenya)
get_soc_climate_drivers	9.25, 47.75	11 (Austria)	79 (Germany)

Verified against the deployed spatialize-country-grid pin (58,795 cells) — the majority cell-to-area assignment the real builders use. The sibling cells were already correct: Spain 203 (-3.25/-3.75, 40.25), Brazil 21 (-55.25/-55.75, -12.25), Greece 84 (26.25, 35.25), Argentina 9 (-64.25, -35.75).

Classification: mechanical. No published value changes — only documented example fixtures move, and only their area_code (plus the reporting-polity columns derived from it).

What changed

• .example_water_balance(), .example_soc_climate_drivers() and .example_luh2_landuse() now carry the code the grid assigns to each cell (11 → 79, 79 → 114), with a comment at each fixture naming the source and the codes.
• New checked-in extract tests/testthat/fixtures/country_grid_example_cells.csv: the pin's rows (revision 20260625T101041Z-8ff94) for exactly the 17 cells the example fixtures use. This is what makes the claim checkable offline — the suite must not read the pin.
• New guard in tests/testthat/test_exported_polity_columns.R: sweeps every example = TRUE output that has lon, lat and area_code and requires the fixture code to equal the grid's. An empty area_code in the extract means the pin has no row for that cell — the deliberately abstract toy cells around (0.25, 0.25) and (0.25, 50.25) are open water, so they make no claim about a place and are left alone (the second half of the issue's decision). A cell that is not listed at all fails the guard, so a new fixture cell must be looked up against the pin rather than invented.

Not touched: the synthetic cell-polity crosswalks inside test_water_balance.R / test_luh2_landuse.R, which are injected inputs to wiring tests rather than claims a reader sees in exported output.

Verification

New guard against the unfixed fixtures (git stash push R/toy_examples.R):

Failure ('test_exported_polity_columns.R:287:3'): gridded example cells carry the grid's own area code
`disagree` (`actual`) and character() (`expected`) don't have the same values.
* Only in `actual`: "build_water_balance (9.25, 47.75): fixture 11, grid 79",
  "build_water_balance (9.75, 47.75): fixture 11, grid 79",
  "build_water_balance (35.75, -1.25): fixture 79, grid 114",
  "build_water_balance (35.25, -1.25): fixture 79, grid 114",
  "get_soc_climate_drivers (9.25, 47.75): fixture 11, grid 79",
  "read_luh2_landuse (35.25, -1.25): fixture 79, grid 114",
  "read_luh2_landuse (9.25, 47.75): fixture 11, grid 79"
[ FAIL 1 | WARN 0 | SKIP 0 | PASS 51 ]

After the fix:

✔ |         52 | exported_polity_columns [6.2s]
[ FAIL 0 | WARN 0 | SKIP 0 | PASS 52 ]

Affected consumers of the changed fixtures:

✔ |      1 101 | carbon_balance [9.1s]
✔ |          9 | exported_examples [1.3s]
✔ |         84 | luh2_landuse [37.9s]
✔ |         81 | toy_examples [2.4s]
✔ |      1 139 | water_balance [9.1s]
[ FAIL 0 | WARN 0 | SKIP 2 | PASS 414 ]

air format . run; lintr clean on both changed files. No roxygen changed, so man/ and _pkgdown.yml are untouched.

Closes #686.

🤖 Generated with Claude Code