addition/interpolation, projection/prediction, and limiting/offsetting rule layers

[corybrunson]

This PR addresses half of #4, which should remain open. It is feature-complete and (i believe) only unit tests remain to be written.

[corybrunson]

All tests pass and examples work as expected.

However, GeomAxis has become quite complicated, with an absurd number of conditionals based on whether the received data include limit and offset aesthetics. I believe it should be split into GeomAxis (for unlimited lines, with or without offsets) and GeomRule (for limited segments, with or without offsets). Then StatRule will pair by default with GeomRule, which will expect axis,lower,upper and welcome (optionally) yintercept,xintercept. Meanwhile, StatRule can also pair with GeomAxis, which will welcome both pairs of aesthetics.

Importantly, the behavior of layer(stat = "rule", geom = "axis") should be different than current if lower,upper are provided: The axes should still be drawn as unlimited lines, but the lower and upper limits should be included in $setup_data() so that the plotting window includes them. If lower,upper are not provided, then the current behavior of widening the window only to include the offset vectors should remain.

[corybrunson]

I believe this is ready to merge. It resolves all of #4 (!), where much of the thinking is documented, as well as #58.

[corybrunson]

I almost forgot to make StatProjection compatible with ord_aes() i.e. to enable plotting projections (from full ordination space onto biplot space) of segments (projecting inherited data to reference data) as well as segments projecting projections (from full to biplot) of inherited data to projections of reference data. The iris data example will be revised to illustrate this functionality. Since old code will be deleted, the reprex below (with the revised code implemented as StatProjection2) shows that the refactoring is actually more time- and memory-efficient, likely due largely to more clever use of matrix operations.

library(devtools)
#> Loading required package: usethis
load_all("~/Documents/software/ordr/")
#> ℹ Loading ordr
#> Loading required package: ggplot2
iris_pca <- ordinate(iris, cols = 1:4, model = prcomp)
bench::mark(
  `1` = {
    p <- ggbiplot(iris_pca, aes(color = Species, label = name)) +
      stat_rows_projection()
    layer_data(p)
  },
  `2` = {
    p <- ggbiplot(iris_pca, aes(color = Species, label = name)) +
      stat_rows_projection2()
    layer_data(p)
  },
  check = FALSE
)
#> # A tibble: 2 × 6
#>   expression      min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 1            32.7ms   33.3ms      29.7    9.61MB     7.42
#> 2 2            27.6ms   28.4ms      35.0    2.25MB    10.8
bench::mark(
  `1` = {
    p <- ggbiplot(iris_pca, ord_aes(iris_pca, color = Species, label = name)) +
      stat_rows_projection()
    layer_data(p)
  },
  `2` = {
    p <- ggbiplot(iris_pca, ord_aes(iris_pca, color = Species, label = name)) +
      stat_rows_projection2()
    layer_data(p)
  },
  check = FALSE
)
#> # A tibble: 2 × 6
#>   expression      min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 1            33.6ms     35ms      28.1    1.64MB    12.5 
#> 2 2            29.3ms   29.9ms      33.0    1.25MB     7.07

^{Created on 2024-12-25 with reprex v2.1.1}