[R] with vs. attach
peter dalgaard
pdalgd at gmail.com
Mon May 9 14:12:51 CEST 2016
On 09 May 2016, at 02:46 , Bert Gunter <bgunter.4567 at gmail.com> wrote:
> ... To be clear, Hadley or anyone else should also feel free to set me
> straight, preferably publicly, but privately if you prefer.
Not really to "set anyone straight", but there are some subtleties with mode call objects versus expression objects and formulas to be aware of.
E.g.,
> a <- 2
> do.call("print", list(a*pi))
[1] 6.283185
> do.call("print", list(quote(a*pi)))
[1] 6.283185
> do.call("print", list(expression(a*pi)))
expression(a * pi)
> do.call("print", list(~a*pi))
~a * pi
Thing is, if you insert a call object into a parse tree, nothing is there to preserve its nature as an unevaluated expression. Similarly, in
> call("print", quote(a*pi))
print(a * pi)
the result is identical to quote(print(a * pi)), so when evaluated, quoting is not seen by print().
As far as I understand, this is also the reason that for math in ggplot, you may need as.expression(bquote(....)).
In general, I think that a number of things in R had been more cleanly implemented using formulas/expression objects than using substitution and lazy evaluation, notably subset and offset arguments in lm/glm. It would have been so much cleaner to have
lm(math ~ age, data = foo, subset = ~ sex=="1")
than the current situation where lm internally chops its own head off and substitutes with model.frame, then evaluates the call to model.frame() which in turn does eval(substitute(subset), data, env). Of course, at the time, ~ was intended specifically for Wilkinson Rogers type formulas; "abusing" it for other kinds of expressions is something of an afterthought.
-pd
>
> Cheers,
> Bert
> Bert Gunter
>
> "The trouble with having an open mind is that people keep coming along
> and sticking things into it."
> -- Opus (aka Berkeley Breathed in his "Bloom County" comic strip )
>
>
> On Sun, May 8, 2016 at 5:28 PM, Bert Gunter <bgunter.4567 at gmail.com> wrote:
>> Jeff:
>>
>> That's easy to do already with substitute(), since you can pass around
>> an unevaluated expression (a parse tree) however you like. As I read
>> it, (admittedly quickly) what it's main feature is that it allows you
>> more control over the environment in which the expression is finally
>> evaluated -- as well as permitting nested expression evaluation fairly
>> easily.
>>
>> But maybe we're saying the same thing ... IMHO I think Hadley has
>> gone overboard here, worrying about rarely important issues, as you
>> seem to be intimating also.
>>
>> Feel free to set me straight... or ignore.
>>
>> Cheers,
>> Bert
>> Bert Gunter
>>
>> "The trouble with having an open mind is that people keep coming along
>> and sticking things into it."
>> -- Opus (aka Berkeley Breathed in his "Bloom County" comic strip )
>>
>>
>> On Sun, May 8, 2016 at 4:02 PM, Jeff Newmiller <jdnewmil at dcn.davis.ca.us> wrote:
>>> The lazyeval package addresses the problem of how to delay evaluation even when the function you want to do the evaluation in is buried two or more function calls below where the original call was made. If you are not building nested function calls with delayed evaluation then you probably don't need that package.
>>> --
>>> Sent from my phone. Please excuse my brevity.
>>>
>>> On May 8, 2016 3:30:16 PM PDT, Spencer Graves <spencer.graves at effectivedefense.org> wrote:
>>>> Hi, Hadley et al.:
>>>>
>>>>
>>>> Hadley's link requires his development version of "lazyeval",
>>>> which can be obtained as follows:
>>>>
>>>>
>>>> library(devtools)
>>>> install_github("hadley/lazyeval")
>>>>
>>>>
>>>> Hadley's link describes real problems with elegant solutions.
>>>>
>>>>
>>>> However, David's solution solved my immediate problem, and it's
>>>> not immediately obvious to me how his "expr_text" function (or other
>>>> functions in "lazyevel") to produce a better solution.
>>>>
>>>>
>>>> Thanks again to David, Peter and Hadley for their replies.
>>>>
>>>>
>>>> Spencer Graves
>>>>
>>>>
>>>> On 5/6/2016 5:08 PM, Hadley Wickham wrote:
>>>>> You may want to read http://rpubs.com/hadley/157957, which captures
>>>> my
>>>>> latest thinking (and tooling) around this problem. Feedback is much
>>>>> appreciated.
>>>>>
>>>>> Hadley
>>>>>
>>>>> On Fri, May 6, 2016 at 2:14 PM, David Winsemius
>>>> <dwinsemius at comcast.net> wrote:
>>>>>>> On May 6, 2016, at 5:47 AM, Spencer Graves
>>>> <spencer.graves at effectivedefense.org> wrote:
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> On 5/6/2016 6:46 AM, peter dalgaard wrote:
>>>>>>>> On 06 May 2016, at 02:43 , David Winsemius
>>>> <dwinsemius at comcast.net> wrote:
>>>>>>>>
>>>>>>>>>> On May 5, 2016, at 5:12 PM, Spencer Graves
>>>> <spencer.graves at effectivedefense.org> wrote:
>>>>>>>>>>
>>>>>>>>>> I want a function to evaluate one argument
>>>>>>>>>> in the environment of a data.frame supplied
>>>>>>>>>> as another argument. "attach" works for
>>>>>>>>>> this, but "with" does not. Is there a way
>>>>>>>>>> to make "with" work? I'd rather not attach
>>>>>>>>>> the data.frame.
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> With the following two functions "eval.w.attach"
>>>>>>>>>> works but "eval.w.with" fails:
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> dat <- data.frame(a=1:2)
>>>>>>>>>> eval.w.attach <- function(x, dat){
>>>>>>>>>> attach(dat)
>>>>>>>>>> X <- x
>>>>>>>>>> detach()
>>>>>>>>>> X
>>>>>>>>>> }
>>>>>>>>>>
>>>>>>>>>> eval.w.with <- function(x, dat){
>>>>>>>>>> with(dat, x)
>>>>>>>>>> }
>>>>>>>>>>
>>>>>>>>>> eval.w.attach(a/2, dat) # returns c(.5, 1)
>>>>>>>>> How about using eval( substitute( ...))?
>>>>>>>>>
>>>>>>>>> eval.w.sub <- function(expr, datt){
>>>>>>>>> eval( substitute(expr), env=datt)
>>>>>>>>> }
>>>>>>>>> eval.w.sub(a/2, dat)
>>>>>>>>> #[1] 0.5 1.0
>>>>>>>>>
>>>>>>>>>
>>>>>>>> Actually, I think a better overall strategy is to say that if you
>>>> want to pass an expression to a function, then pass an expression
>>>> object (or a call object or maybe a formula object).
>>>>>>>>
>>>>>>>> Once you figure out _how_ your eval.w.attach works (sort of),
>>>> you'll get the creeps:
>>>>>>>>
>>>>>>>> Lazy evaluation causes the argument x to be evaluated after the
>>>> attach(), hence the evaluation environment of an actual argument is
>>>> being temporarily modified from inside a function.
>>>>>>>>
>>>>>>>> Apart from upsetting computer science purists, there could be
>>>> hidden problems: One major issue is that values in "dat" could be
>>>> masked by values in the global environment, another issue is that an
>>>> error in evaluating the expression will leave dat attached. So at a
>>>> minimum, you need to recode using on.exit() magic.
>>>>>>>>
>>>>>>>> So my preferences go along these lines:
>>>>>>>>
>>>>>>>>> dat <- data.frame(a=1:2)
>>>>>>>>> eval.expression <- function(e, dat) eval(e, dat)
>>>>>>>>> eval.expression(quote(a/2), dat)
>>>>>>>> [1] 0.5 1.0
>>>>>>>>> eval.expression(expression(a/2), dat)
>>>>>>>> [1] 0.5 1.0
>>>>>>>>
>>>>>>>>> eval.formula <- function(f, dat) eval(f[[2]], dat)
>>>>>>>>> eval.formula(~a/2, dat)
>>>>>>>> [1] 0.5 1.0
>>>>>>> Hi, Peter:
>>>>>>>
>>>>>>>
>>>>>>> I don't like eval.expression or eval.formula, because they
>>>> don't automatically accept what I naively thought should work and
>>>> require more knowledge of the user. What about David's eval.w.sub:
>>>>>>>
>>>>>>>
>>>>>>> a <- pi
>>>>>>> dat <- data.frame(a=1:2)
>>>>>>> eval.w.sub <- function(a, Dat){
>>>>>>> eval( substitute(a), env=Dat)
>>>>>>> }
>>>>>>>> eval.w.sub(a/2, dat)
>>>>>>> [1] 0.5 1.0
>>>>>> I liked eval.expression and tested it with a bquote(...) argument to
>>>> see if that would succeed. It did, but it didn't return what you wanted
>>>> for `a/2`, so I tried seeing if a "double eval wuold deliver both yours
>>>> and my desired results:
>>>>>>
>>>>>> eval.w.sub <- function(a, Dat){
>>>>>> eval( eval(substitute(a),Dat), env=Dat)
>>>>>> }
>>>>>> x=2
>>>>>> eval.w.sub( a/2, dat)
>>>>>> [1] 0.5 1.0
>>>>>> eval.w.sub( bquote(2*a*.(x) ), dat)
>>>>>> [1] 4 8
>>>>>>
>>>>>> We are here retracing the path the Hadley took in some of his
>>>> ggplot2 design decsions. Unfortunately for me those NSE rules often
>>>> left me confused about what should and shouldn't be 'quoted' in the
>>>> as-character sense and what should be quote()-ed or "unquoted" in the
>>>> bquote() sense.
>>>>>> --
>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> This produces what's desired in a way that seems simpler to
>>>> me.
>>>>>>>
>>>>>>>
>>>>>>> By the way, I really appreciate Peter's insightful comments:
>>>>>>>
>>>>>>>
>>>>>>> eval.w.attachOops <- function(x, Dat){
>>>>>>> attach(Dat)
>>>>>>> X <- x
>>>>>>> detach()
>>>>>>> X
>>>>>>> }
>>>>>>>> eval.w.attachOops(a/2, dat)
>>>>>>> The following object is masked _by_ .GlobalEnv:
>>>>>>>
>>>>>>> a
>>>>>>>
>>>>>>> [1] 1.570796
>>>>>>>> eval.w.attachOops(b/2, dat)
>>>>>>> The following object is masked _by_ .GlobalEnv:
>>>>>>>
>>>>>>> a
>>>>>>>
>>>>>>> Error in eval.w.attachOops(b/2, dat) : object 'b' not found
>>>>>>>> search()
>>>>>>> [1] ".GlobalEnv" "Dat" "package:graphics"
>>>>>>> [4] "package:grDevices" "package:utils" "package:datasets"
>>>>>>> [7] "package:methods" "Autoloads" "package:base"
>>>>>>>> objects(2)
>>>>>>> [1] "a"
>>>>>>>
>>>>>>> *** NOTES:
>>>>>>>
>>>>>>>
>>>>>>> 1. This gives a likely wrong answer with a warning if "a"
>>>> exists in .GlobalEnv, and leaves "Dat" (NOT "dat") attached upon exit.
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> 2. A stray "detach()" [not shown here] detached
>>>> "package:stats". oops.
>>>>>>>
>>>>>>>
>>>>>>> *** Using "on.exit" fixes the problem with failure to detach but
>>>> not the likely wrong answer:
>>>>>>>
>>>>>>>
>>>>>>> detach()
>>>>>>> search()
>>>>>>> eval.w.attachStillWrong <- function(x, dat){
>>>>>>> attach(dat)
>>>>>>> on.exit(detach(dat))
>>>>>>> X <- x
>>>>>>> X
>>>>>>> }
>>>>>>> The following object is masked _by_ .GlobalEnv:
>>>>>>>
>>>>>>> a
>>>>>>>
>>>>>>> [1] 1.570796
>>>>>>>> eval.w.attachStillWrong(b/2, dat)
>>>>>>> The following object is masked _by_ .GlobalEnv:
>>>>>>>
>>>>>>> a
>>>>>>>
>>>>>>> Error in eval.w.attachStillWrong(b/2, dat) : object 'b' not found
>>>>>>>> search()
>>>>>>> [1] ".GlobalEnv" "package:grDevices" "package:utils"
>>>>>>> [4] "package:datasets" "package:methods" "Autoloads"
>>>>>>> [7] "package:base"
>>>>>>>
>>>>>>>
>>>>>>> Thanks again to Peter and David. Spencer
>>>>>>>
>>>>>>>> Peter D.
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>> --
>>>>>>>>> David.
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>> eval.w.with(a/2, dat) # Error ... 'a' not found
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> Thanks, Spencer Graves
>>>>>>>>>>
>>>>>>>>>> [[alternative HTML version deleted]]
>>>>>>>>>>
>>>>>>>>>> ______________________________________________
>>>>>>>>>> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more,
>>>> see
>>>>>>>>>> https://stat.ethz.ch/mailman/listinfo/r-help
>>>>>>>>>> PLEASE do read the posting guide
>>>> http://www.R-project.org/posting-guide.html
>>>>>>>>>> and provide commented, minimal, self-contained, reproducible
>>>> code.
>>>>>>>>> David Winsemius
>>>>>>>>> Alameda, CA, USA
>>>>>>>>>
>>>>>>>>> ______________________________________________
>>>>>>>>> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
>>>>>>>>> https://stat.ethz.ch/mailman/listinfo/r-help
>>>>>>>>> PLEASE do read the posting guide
>>>> http://www.R-project.org/posting-guide.html
>>>>>>>>> and provide commented, minimal, self-contained, reproducible
>>>> code.
>>>>>> David Winsemius
>>>>>> Alameda, CA, USA
>>>>>>
>>>>>> ______________________________________________
>>>>>> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
>>>>>> https://stat.ethz.ch/mailman/listinfo/r-help
>>>>>> PLEASE do read the posting guide
>>>> http://www.R-project.org/posting-guide.html
>>>>>> and provide commented, minimal, self-contained, reproducible code.
>>>>>
>>>>>
>>>>
>>>> ______________________________________________
>>>> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
>>>> https://stat.ethz.ch/mailman/listinfo/r-help
>>>> PLEASE do read the posting guide
>>>> http://www.R-project.org/posting-guide.html
>>>> and provide commented, minimal, self-contained, reproducible code.
>>>
>>> [[alternative HTML version deleted]]
>>>
>>> ______________________________________________
>>> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
>>> https://stat.ethz.ch/mailman/listinfo/r-help
>>> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
>>> and provide commented, minimal, self-contained, reproducible code.
--
Peter Dalgaard, Professor,
Center for Statistics, Copenhagen Business School
Solbjerg Plads 3, 2000 Frederiksberg, Denmark
Phone: (+45)38153501
Office: A 4.23
Email: pd.mes at cbs.dk Priv: PDalgd at gmail.com
More information about the R-help
mailing list