accessing .SD via data.table::.SD gives NULL

[spennihana]

I wasn't able to find an answer to this via searching SO/issues, so I'll just ask:

Is this legal way of accessing .SD:

someFun <- function(dt) {
  stopifnot(!is.null(dt))
  5
}

r <- data.table::data.table(iris)
r[ , someFun(data.table::.SD)    # Not OK
   , by=.(Sepal.Length)
 ]

# Error: !is.null(dt) is not TRUE

r[ , someFun(.SD)     # OK
   , by=.(Sepal.Length)
 ]

In someFun, the arg dt is NULL. Not a huge pain since I can do an explicit package import, but curious never-the-less.

edit: fixed up example

[MichaelChirico]

no, it's not a legal way of accessing .SD. the object is created within the [] call; it's not an exported object.

…

On Aug 16, 2017 7:20 PM, "spennihana" ***@***.***> wrote: I wasn't able to find an answer to this via searching SO/issues, so I'll just ask: Is this legal way of accessing .SD: someFun <- function(dt) { ... } r <- data.table::data.table(r) r[ , someFun(data.table::.SD, paramData) , by=.(col3) , .SDcols=c("col1", "col2") ] In `someFun`, the arg `dt` is NULL. Not a huge pain since I can do an explicit package import, but curious never-the-less. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#2306>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AHQQdWriPZlKLK35_cqKNwfmwmJ5Lvx2ks5sY3kxgaJpZM4O5mLh> .

[spennihana]

Thanks for your response; it's still bothersome that there's no way of telling data.table to get the groups without explicitly importing the package.

At first I was confused by what you meant by "not exported" (https://github.com/Rdatatable/data.table/blob/master/NAMESPACE#L27), but I see that you mean it's built on-the-fly.

I want to import and use data.table in this way, without explicitly setting it in my DESCRIPTION:depends. Any trick here?

[MichaelChirico]

why not just import it? using data.table without importing the C code is bound to end poorly.

…

On Aug 16, 2017 7:52 PM, "spennihana" ***@***.***> wrote: Thanks for your response; it's still bothersome that there's no way of telling data.table to get the groups without explicitly importing the package. At first I was confused by what you meant by "not exported" ( https://github.com/Rdatatable/data.table/blob/master/NAMESPACE#L27), but I see that you mean it's built on-the-fly. I want to import and use data.table in this way, without explicitly setting it in my DESCRIPTION:depends. Any trick here? — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#2306 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AHQQdStqPlsPngXB-Okg1bUduexug5DNks5sY4C7gaJpZM4O5mLh> .

[spennihana]

when developing a package id like to access via the package namespace. Among other things it avoids noisy startup messages that I don't want to appear.

[jangorecki]

@spennihana did you tried just .SD? You have to put to imports/depends every package that you use by ::, you don't have to import its functions or attach to search path. Unless you escape that function call with something like requireNamespace, then Suggests is enough.

[spennihana]

Tried bare .SD too, no luck.

It is desirable for all packages to be in Imports. I do not want them in Depends.

Yes, this R package is following a strict style of <pkg>::<fn>. But then we came across the current issue. I could look at requireNamespace to special-case data.table as a way of muting the startup messages; but it's unfortunate that this needs special handling...

[spennihana]

Actually, I got a hint on how to handle this from Matt via: #850

I will set these to NULL just before usage.

Thanks!

[f2ostie]

I'm having the same issue. Could somebody elaborate on the solution?

I imported data.table in the DESCRIPTION:

Imports:
  data.table (>= 1.10.0),
	RODBC (>= 1.3.0)

but

  r <- data.table::data.table(iris)
  r[ , print(.SD), by=.(Sepal.Length)]

just prints NULL.

Tried setting .SD to NULL before function to no avail.

 r <- data.table::data.table(iris)
.SD <- NULL
r[ , print(.SD), by=.(Sepal.Length)]

Can anybody fix my example? (Tried data.table::.SD but gave same result)

[HughParsonage]

Have you imported data.table in your package's NAMESPACE?

[f2ostie]

I have done that now and it indeed works, thanks!

[jangorecki]

@f2ostie see also the new vignette: https://rdatatable.gitlab.io/data.table/library/data.table/doc/datatable-importing.html