Despite consulting this thread, I struggle understanding the following output:
df <- tibble::tribble(
~color, ~a, ~b,
"blue", 1, 2,
"green", 6, 2,
"purple", 3, 3,
"red", 2, 3,
"yellow", 5, 1
)
Which gives:
> fct_reorder2(df$color, df$a, df$b, .fun = min, .desc = TRUE)
[1] blue green purple red yellow
Levels: purple green red blue yellow
I know you are supposed to use .fun differently with fct_reorder2. The min function here computes the mininum across all provided values, which are here the values in both df$a and df$b. I still would not expect the result I got. Can someone explain, please?
One of the linked answer's is downvoted for looking at the source but you're asking for the how so I think it makes sense to actually look at what the code for fct_reorder2 does.
# This is fine, just checking if it's a factor and assigning the value.
f <- check_factor(.f)
# Also fine, they're columns from a data.frame
stopifnot(length(f) == length(.x), length(.x) == length(.y))
# We're not using dots
ellipsis::check_dots_used()
With that out of the way we can use the subsequent code with original data:
summary <- tapply(seq_along(.x), f, function(i) .fun(.x[i], .y[i], ...))
# for us equivalent to
tapply(seq_along(df$a), df$color, function(i) {min(df$a[i], df$b[i])})
# blue green purple red yellow
# 1 2 3 2 1
This is in this case just the pairwise min of columns df$a and df$b
If you had multiple rows per color it would use the minimum value from any row or column of the factor level.
lvls_reorder(.f, order(summary, decreasing = .desc))
This just orders levels based on those values in descending order, so the color with the biggest pairwise minimum of columns a and b first. In the case of ties we can see it's sorted lexicographically leading to the output we see.
| color | a | b | pmin | dense rank | descending order (Lexicographically sorted for ties) |
|---|---|---|---|---|---|
| blue | 1 | 2 | 1 | 1 | 4 |
| green | 6 | 2 | 2 | 2 | 2 |
| purple | 3 | 3 | 3 | 3 | 1 |
| red | 2 | 3 | 2 | 2 | 3 |
| yellow | 5 | 1 | 1 | 1 | 5 |