length(unique(x)) and n_distinct(x) return different answers for some lists

[jgabry]

The documentation for n_distinct says that it's an "equivalent of length(unique(x))"
but it gives a different answer than length(unique(x)) if x is a list with multiple (identical) elements.

library(dplyr)  # I checked versions 0.5.0 and 0.5.0.9000
x <- list(1,1)
length(unique(x)) # result: 1
n_distinct(x) # result: 2

Is this a bug or is there just a missing caveat in the doc?

[krlmlr]

This is because values in list columns are compared by reference, so n_distinct() treats them as different unless they really point to the same object:

a <- 1
n_distinct(list(a,a))
## [1] 1

Would you like to contribute documentation?

[jgabry]

Thanks for the explanation. That makes sense. I'd be happy to submit a PR
that clarifies this in the doc and adds an example demonstrating. Can
hopefully do this today.

On Monday, November 7, 2016, Kirill Müller <notifications@github.com
javascript:_e(%7B%7D,'cvml','notifications@github.com');> wrote:

This is because values in list columns are compared by reference, so
n_distinct() treats them as different unless they really point to the same
object:

a <- 1
n_distinct(list(a,a))## [1] 1

Would you like to contribute documentation?

—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
#2222 (comment), or mute
the thread
https://github.com/notifications/unsubscribe-auth/AHb4Q3nYXgpKdCguZGs0RwuFYhnxJTnaks5q70GUgaJpZM4KlbbQ
.

[hadley]

I think this is more of a bug than something we need to document

[jgabry]

When I originally opened the issue it was because it seemed like a bug. Kirill's explanation of the behavior makes sense but it does seem preferable to have it behave like the base R equivalent than to have to explain the unexpected deviation. On Tue, Jan 31, 2017 at 6:45 PM Hadley Wickham <notifications@github.com> wrote: I think this is more of a bug than something we need to document — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#2222 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AHb4QwazEidocYUzxcwAcOPltp5JPKVFks5rX8exgaJpZM4KlbbQ> .

[hadley]

Now part of #2355