names<- method should protect against invalid names

[hadley]

i.e. the following code should generate errors:

library(tibble)

df <- tibble(x = 1, y = 2)

names(df) <- "x"
names(df) <- c("x", "x")
names(df) <- c("", "")

[jennybc]

I think this could create some self-contradiction because, if I see how things are going, there will be a way to authorize, e.g., creation of a tibble with duplicate names, if you really want to.

I suspect the only thing that is categorically not allowed is names(df) <- NULL. And maybe names<- could also enforce this principle (if adopted): treat a name of NA as "".

[hadley]

We could still allow (e.g.)

names(df, .names = "tidy") <- "x"

[krlmlr]

I see there's names<-.POSIXlt(), so I guess we might be able to implement names<-.tbl_df() . But would it affect rlang::set_names() ?

Would it work if we warn first, instead of throwing an error, to reduce friction with existing code?

[jennybc]

So is this a go for 1.5.0?

To restate @hadley's example with current syntax, it could look like:

names(df, .name_repair = "minimal") <- "x"

and presumably .name_repair would generally work as in tibble() and as_tibble()?

[krlmlr]

With 6165f02 I'm seeing:

devtools::load_all("~/git/R/tibble")
#> Loading tibble
df <- tibble(a = 1)
names(df, .name_repair = "minimal") <- NULL
#> Error in `names<-`(`*tmp*`, .name_repair = "minimal", value = NULL): 3 arguments passed to 'names<-' which requires 2

^{Created on 2018-10-19 by the reprex package (v0.2.1.9000)}

So a names<-() with an extra argument doesn't seem to be supported.

I'll add a simple "minimal" check for now, we'll see how many downstream packages are affected. Also, the user can always subvert with rlang::set_names() .

[krlmlr]

Seems like set_names() somehow forwards to names<-(), so we're safe here.

[github-actions]

This old thread has been automatically locked. If you think you have found something related to this, please open a new issue and link to this old issue if necessary.