size.by & opacity.by parameters for jitter/scatter plots

[j-andrews7]

To encode yet another piece of data in the plot.

[j-andrews7]

This can already be done with size, but it's only mentioned in the details rather than the parameter itself. And doesn't work with do.letter. Using do.hover with it also results in a single combined legend for color.by and size, which is functionally uninterpretable.

set.seed(1)
library(palmerpenguins)
df <- as.data.frame(penguins[!apply(penguins, 1, function(x) { any(is.na(x)) }), ])

rownames(df) <- paste0("row", rownames(df))

# Dimensionality reduction
pca <- prcomp(df[, c("bill_length_mm", "bill_depth_mm", "flipper_length_mm", "body_mass_g")])
df <- cbind(df, pca$x)

# Additional 'random' observations
df$groups <- sample(c("A","B","C","D","E"), nrow(df), replace = TRUE)
df$age <- sample(c("1","2","3","4"), nrow(df), replace = TRUE)
df$number <- as.numeric(seq_len(nrow(df)))

scatterPlot(
        df, x.by = "PC1", y.by = "PC2",
        color.by = "island",
        size = "number", do.hover = TRUE)

Versus this with do.hover = FALSE:

The do.letter error just requires a few checks in .add_letters to overcome, but the do.hover issue is inherent to plotly (see plotly/plotly.R#2163) and unlikely to be fixed until the underlying plotly.js version used by the R package is updated (and this bug is fixed plotly/plotly.js#7023).

[j-andrews7]

Added this same functionality to opacity in #19, and cleaned up the docs a bit to make it more clear.