`fct_infreq()` should support weights

[krlmlr]

similarly to fct_lump() .

library(forcats)

x <- letters[c(1:3, 2)]
w <- c(4, 1, 3, 1)

x
#> [1] "a" "b" "c" "b"
w
#> [1] 4 1 3 1

fct_infreq(x)
#> [1] a b c b
#> Levels: b a c

fct_infreq(x, w = w)
#> Error in fct_infreq(x, w = w): unused argument (w = w)

# Should behave like:
fct_reorder(x, w, sum, .desc = TRUE)
#> [1] a b c b
#> Levels: a c b

^{Created on 2020-04-27 by the reprex package (v0.3.0)}

[twest820]

Thanks @krlmlr. I just hit this with geom_bar(aes(y = fct_infreq(<factor>), weight = <weight>)), which doesn't sort the discrete axis correctly, presumably because ggplot 3.4.0 isn't flowing weight to fct_infreq(). fct_reorder() handles this case correctly.

geom_bar(aes(y = fct_reorder(<factor>, <weight>, sum), weight = <weight>))

It seems perhaps also worth noting the reordering FAQ entry for geom_bar() doesn't consider this case. So, depending what happens with this issue, it may be worth revising the FAQ.