I want to depend on littler, please export something

[gaborcsardi]

I want to use littler from my package, but I don't want to Depend on it, because of the lengthy startup message. I think the message is useful in general, but maybe not so much if littler is just the background machinery of another package.

Then the problem is that I cannot use Imports, either, because then I need to import sg to avoid an R CMD check warning. But there is nothing to import.

I quite like the fact that littler does not export anything. But I think it might be worth exporting a little something so that I can Import the package cleanly. E.g. how about importing a single variable or function r which just points to the binary of littler? That's even (slightly) useful, I don't need to call system.file(), just use littler::r.

Alternative suggestions are also welcome.

[eddelbuettel]

What message?

edd@don:~$ r -e 'q()'
edd@don:~$ r -e 'cat("Hi Gabor\n"); q()'
Hi Gabor
edd@don:~$ 

[gaborcsardi]

I want to use it from an R package. Then if one loads my package, one gets

Loading expect
Loading required package: littler
The littler package provides 'r' as a binary.
On OS X, 'r' and 'R' are the same so 'lr' is an alternate name for littler.

because I need Depends...

[eddelbuettel]

Dunno. Can you not suppressPackageStartupMessage() when Importing? Seems like an R bug.

[gaborcsardi]

I depend on your package, so I don't think I can specify suppressPackageStartupMessage.....

[eddelbuettel]

Try suppressMessages(library(myPackageInQuestion)) which is what I do.

Else, I suppose, I could be silent if getOption("littlerPleaseBeSilent", FALSE) returns TRUE ie when the option is set. But all this is a wee bit orthogonal to the normal usage of the package ... (but new to us now that the package is in fact on CRAN).

[gaborcsardi]

I can't do suppressMessages(library(myPackageInQuestion)) because I don't call library() at all. I have:

...
URL: https://github.com/mangothecat/expect
BugReports: https://github.com/mangothecat/expect/issues
Depends:
    littler
Imports:
    R6
...

An option would be great!

I think this is a bug in R CMD check. An R package can be useful and required, even if one does not import anything from it. One can still use the files it supplies....

I suppose this is not the regular littler usage, but it is indeed one of the "benefits" of having it in a CRAN package. :)

[gaborcsardi]

I am not sure that an option would work, actually. littler is loaded before my package is loaded, so I might not have time to set the option. Maybe in .onLoad.

[gaborcsardi]

Btw it looks like to find the executable from R, I need sg like

file.path(system.file(package="littler"), "bin",
                   paste0("r", ifelse(.Platform$OS.type=="windows", ".exe", "")))

(from littler itself).

Maybe it makes sense to provide this as an r() function, because it is quite error prone to write it?

[eddelbuettel]

Maybe. I can look at what tools and utils do.

[eddelbuettel]

BTW there is no .Platform$OS.type=="windows". It is "OSs with an X only".

[eddelbuettel]

So this, I guess, it close:

R> littler:::test("cat(system.file(package='littler', 'bin', 'r'))")
[1] "/usr/local/lib/R/site-library/littler/bin/r"
R> 

[gaborcsardi]

Well, I took the code from liittler, and I assumed that it was right. :)
https://github.com/eddelbuettel/littler/blob/738371378d4e03f27adb3b01e38ecd5ea4873c2a/R/test.R#L2-3

[eddelbuettel]

Yes, I later read up on test.r -- not something I wrote initially, but simply accommodating a PR by somebody else.

[gaborcsardi]

Thanks! Btw. I just found out that if you need to import littler, you can also just import the whole package, and apparently R CMD check is fine with that (even if nothing will be imported).

I still think that an r() function makes sense, but it seems that you don't actually need it to be able to import the package. FYI.

[eddelbuettel]

Well...now it is in there. It just a one-liner, really, and a few lines of roxygen.