[Rd] possible improvement to ?with examples

Martin Maechler maechler at stat.math.ethz.ch
Tue Feb 21 17:42:00 CET 2017


>>>>> Ben Bolker <bbolker at gmail.com>
>>>>>     on Thu, 16 Feb 2017 15:37:13 -0500 writes:

    > A querent on StackOverflow asked about the with() function
    > http://stackoverflow.com/questions/42283479/why-when-to-use-with-function#42283479

    > and asked about the example in ?with

    > library(MASS)
    > with(anorexia, {
    >  anorex.1 <- glm(Postwt ~ Prewt + Treat + offset(Prewt),
    >                 family = gaussian)
    >  summary(anorex.1)
    > })

    > which saves little or no typing relative to

    > anorex.1 <- glm(Postwt ~ Prewt + Treat + offset(Prewt),
    > family = gaussian, data=anorexia)

    > (I would argue that the latter is better practice anyway).

    > Could we have something more sensible like

    > with(mtcars,mpg[cyl==8 & disp>350])

    > ?  (It could be contrasted directly with

    > mtcars$mpg[mtcars$cyl==8 & mtcars$disp>350]

    > )

I now have done something like the above, and have added a
\note{ .. }  to warn about "over - use" of with().

Also added a link to Thomas Lumley's paper
  Thomas Lumley (2003)  \emph{Standard nonstandard evaluation rules}.
  \url{http://developer.r-project.org/nonstandard-eval.pdf}

    > I'm happy to submit a bug report/patch if that seems appropriate.

Thank you, Ben!
Martin



More information about the R-devel mailing list