upcoming sf will break nhdR (when adopting s2)

[edzer]

Hi! please see r-spatial/sf#1649 - please also consider commenting there when you run into troubles or need advice or help. Best,

[edzer]

Seeing this output:

══ Failed tests ════════════════════════════════════════════════════════════════
── Error (test-utils.R:10:3): find_vpu works with character vpus ───────────────
Error: Evaluation error: Found 2 features with invalid spherical geometry.
[6] Loop 18 is not valid: Edge 1 crosses edge 3
[22] Loop 49 is not valid: Edge 0 crosses edge 2.
Backtrace:
     █
  1. ├─testthat::expect_equal(nhdR:::find_vpu(pnt), "03S") test-utils.R:10:2
  2. │ └─testthat::quasi_label(enquo(object), label, arg = "object")
  3. │   └─rlang::eval_bare(expr, quo_get_env(quo))
  4. └─nhdR:::find_vpu(pnt)
  5.   ├─base::suppressMessages(st_join(sf::st_sf(pnt), vpu)$UnitID)
  6.   │ └─base::withCallingHandlers(...)
  7.   ├─sf::st_join(sf::st_sf(pnt), vpu)
  8.   └─sf:::st_join.sf(sf::st_sf(pnt), vpu)
  9.     ├─sf:::join(x, y, ...)
 10.     └─sf:::st_intersects.sf(x, y, ...)
 11.       └─sf:::st_geos_binop(...)
 12.         └─s2:::fn(x, y, s2::s2_options(model = s2_model, ...))
 13.           ├─s2:::cpp_s2_intersects_matrix(...)
 14.           ├─s2::as_s2_geography(y)
 15.           └─sf:::as_s2_geography.sf(y)
 16.             └─sf:::st_as_s2.sf(x, ..., oriented = oriented)
 17.               ├─sf::st_as_s2(st_geometry(x), ...)
 18.               └─sf:::st_as_s2.sfc(st_geometry(x), ...)
 19.                 ├─s2::as_s2_geography(st_as_binary(x), ..., oriented = oriented)
 20.                 └─s2:::as_s2_geography.WKB(st_as_binary(x), ..., oriented = oriented)
 21.                   ├─s2:::new_s2_xptr(...)
 22.                   └─s2:::s2_geography_from_wkb(x, oriented = oriented, check = check)
── Error (test-utils.R:16:3): find_vpu works with numeric vpus ─────────────────
Error: Evaluation error: Found 2 features with invalid spherical geometry.
6] Loop 18 is not valid: Edge 1 crosses edge 3
[22] Loop 49 is not valid: Edge 0 crosses edge 2.
Backtrace:
     █
  1. ├─testthat::expect_equal(nhdR:::find_vpu(pnt), "01") test-utils.R:16:2
  2. │ └─testthat::quasi_label(enquo(object), label, arg = "object")
  3. │   └─rlang::eval_bare(expr, quo_get_env(quo))
  4. └─nhdR:::find_vpu(pnt)
  5.   ├─base::suppressMessages(st_join(sf::st_sf(pnt), vpu)$UnitID)
  6.   │ └─base::withCallingHandlers(...)
  7.   ├─sf::st_join(sf::st_sf(pnt), vpu)
  8.   └─sf:::st_join.sf(sf::st_sf(pnt), vpu)
  9.     ├─sf:::join(x, y, ...)
 10.     └─sf:::st_intersects.sf(x, y, ...)
 11.       └─sf:::st_geos_binop(...)
 12.         └─s2:::fn(x, y, s2::s2_options(model = s2_model, ...))
 13.           ├─s2:::cpp_s2_intersects_matrix(...)
 14.           ├─s2::as_s2_geography(y)
 15.           └─sf:::as_s2_geography.sf(y)
 16.             └─sf:::st_as_s2.sf(x, ..., oriented = oriented)
 17.               ├─sf::st_as_s2(st_geometry(x), ...)
 18.               └─sf:::st_as_s2.sfc(st_geometry(x), ...)
 19.                 ├─s2::as_s2_geography(st_as_binary(x), ..., oriented = oriented)
 20.                 └─s2:::as_s2_geography.WKB(st_as_binary(x), ..., oriented = oriented)
 21.                   ├─s2:::new_s2_xptr(...)
 22.                   └─s2:::s2_geography_from_wkb(x, oriented = oriented, check = check)

[ FAIL 2 | WARN 0 | SKIP 14 | PASS 3 ]

so it seems there are some geometries not valid on the sphere. Let me know if we can help. Testing with branch sf_1_0 from the sf repo, and s2 from CRAN (1.0.5).

[jsta]

I think I solved the issue. sf::st_make_valid was not creating valid s2 geometries for some internal package data. The somewhat complex solution at r-spatial/s2#99 (comment) seems to work.

[edzer]

Hi Jem! Great work - I looked into it yesterday, and didn't succeed - I couldn't find that issue you refer to back! Maybe this is also of interest to @dblodgett-usgs who might be fighting with the same weird polygons.