no visible binding for global variable 'x'

[mikemeredith]

When code includes a call to graphics::curve, R CMD check produces a NOTE re "no visible binding for global variable 'x' ". This is a known issue because of the non-standard way that curve uses 'x', but it does mean that the package will not be accepted by CRAN.

The work-around suggested by Duncan Murdoch on nabble is to insert function(x) before/as the first argument to curve. For example,

curve(plogis(beta0 + beta1*x), ...

becomes

curve(function(x) plogis(beta0 + beta1*x), ...

which does the same thing but doesn't set R CMD check barking.

[mikemeredith]

Unfortunately Duncan's work-around doesn't always work, and in any case it makes a mess of the y axis label, which curve generates from its first argument.

So I've implemented the kludge suggested by Henrik Bengtsson in the same R-help discussion, adding

x <- NULL ; rm(x) # Dummy to trick R CMD check

at the very start of the function which calls curve (at the start to avoid zapping any other object called x in the code).

[mikemeredith]

Maybe an even better kludge is

if(FALSE) x <- NULL

which does precisely nothing (and won't zap any object called x) but does keep R CMD check happy.