Fix #96

[atsyplenkov]

Solves the issue of plotting invalid geometries.

[atsyplenkov]

#96

[paleolimbot]

Thank you!

This look good on principle...is checking validity and then simplfying actually faster, though? (i.e., should we just use the topology-preserving simplify always?)

[atsyplenkov]

Okay, so I might be going down a bit of a rabbit hole here, but I tried to check out the time complexity of geos_is_valid, geos_simplify, and geos_simplify_preserve_topology. I just used some data that was easy to load, so my benchmarks aren’t super reproducible, since I cannot share the data in public! But I hope my point is clear as is.

I ran all three operations 5 times on polygons with different numbers of vertices, from 4 to around 600.

So, here’s the deal: both geos_is_valid and geos_simplify have a time complexity of $O(1)$, which is awesome. On the other hand, geos_simplify_preserve_topology looks like it has $O(\log n)$ at best, but honestly, I think it’s more like $O(n)$. Both $O(n)$ and $O(\log n)$ aren’t terrible, but I think it’s probably better to avoid using them unnecessarily.

  library(sf)
#> Linking to GEOS 3.12.1, GDAL 3.8.4, PROJ 9.3.1; sf_use_s2() is TRUE
library(geos)
library(ggplot2)

# Load as geos_geometry
d2021_p <- "/raw/vector/poly_debris_2021/poly_debris_2021.shp"
d2021 <- sf::st_read(d2021_p) |> st_zm() |> as_geos_geometry()

# Sort polygons
d2021_sorted <- d2021[order(geos_num_coordinates(d2021))]

length(d2021_sorted)
#> [1] 1250
summary(geos_num_coordinates(d2021_sorted))
#>    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
#>    4.00   39.00   58.00   78.01   95.00  595.00

# Benchmark
df <- list()

for (i in seq_along(d2021_sorted)) {
  df[[i]] <- 
    bench::mark(
      geos_simplify = geos_simplify(d2021_sorted[i], 1),
      geos_simplify_preserve_topology = geos_simplify_preserve_topology(d2021_sorted[i], 1),
      geos_is_valid = geos_is_valid(d2021_sorted[i]),
    iterations = 5L,
    check = FALSE,
    time_unit = "ms"
  )
}



df |> 
  dplyr::bind_rows(.id = "id") |> 
  dplyr::mutate(
    n = sort(
      rep(
        geos::geos_num_coordinates(d2021_sorted),
        each = 3
      )
    )
  ) |>
  ggplot(
    aes(
      x = n,
      y = median,
      group = as.character(expression),
      color = as.character(expression)
    )
  ) +
  geom_point(alpha = 0.05) +
  geom_smooth(se = F) +
  labs(x = "Number of coordinates", y = "Time (ms)", color = "") +
  theme_minimal() +
  theme(legend.position = "bottom")
#> `geom_smooth()` using method = 'gam' and formula = 'y ~ s(x, bs = "cs")'

^{Created on 2024-11-20 with reprex v2.1.1}