[Rd] I wish xlim=c(0, NA) would work. How about I send you a patch?

Henrik Bengtsson hb at biostat.ucsf.edu
Mon Apr 16 22:11:39 CEST 2012

```Reversing the direction of an axis currently needs an explicit 'xlim'
such that diff(xlim) < 0 [I think].  Thus, the assumption to do a
"forward" plot when "leaving out" one element in 'xlim' by setting it
to a missing value is not too bad.

However, what about using +Inf and -Inf as instead?  Disclaimer: I
haven't bumped into one, but it could be that there plot functions
that include -Inf/+Inf.  Example:

## Case #1: Infer 2nd argument in 'xlim' from data
# (a) Forward
plot(x, xlim=c(0,+Inf))
# (b) Reverse
plot(x, xlim=c(0,-Inf))

## Case #2: Infer 1st argument in 'xlim' from data
# (a) Forward
plot(x, xlim=c(-Inf,0))
# (b) Reverse
plot(x, xlim=c(+Inf,0))

## Case #3: Infer both arguments in 'xlim' from data
# (a) Forward
plot(x, xlim=c(-Inf,+Inf))
# (b) Reverse
plot(x, xlim=c(+Inf,-Inf))

Note how the latter also supports a use case currently not supported.

/Henrik

On Mon, Apr 16, 2012 at 12:15 PM, Duncan Murdoch
<murdoch.duncan at gmail.com> wrote:
> On 12-04-16 2:26 PM, William Dunlap wrote:
>>
>> plot(1:10, xlim=c(10,1)) reverses the x axis.
>> If we allow plot(1:10, xlim=c(5,NA)), which
>> direction should it go?    Would this require new
>> parameters, {x,y}{min,max} or new paremeters
>> {x,y}{axisDirection}?
>
>
> I'd rather not add another parameter.  So if I were to implement this, I'd
> probably choose Paul's original suggestion.  If someone wants c(5, NA) to
> mean c(5, min(data)) rather than c(5, max(data)) they'd need to code it
> explicitly.
>
> Duncan Murdoch
>
>
>>
>> Bill Dunlap
>> Spotfire, TIBCO Software
>> wdunlap tibco.com
>>
>>
>>> -----Original Message-----
>>> From: r-devel-bounces at r-project.org
>>> [mailto:r-devel-bounces at r-project.org] On Behalf
>>> Of Duncan Murdoch
>>> Sent: Monday, April 16, 2012 11:14 AM
>>> To: Paul Johnson
>>> Cc: R Devel List
>>> Subject: Re: [Rd] I wish xlim=c(0, NA) would work. How about I send you a
>>> patch?
>>>
>>> On 12-04-16 1:52 PM, Paul Johnson wrote:
>>>>
>>>> I'm looking for an R mentor.  I want to propose a change in management
>>>> of plot options xlim and ylim.
>>>
>>>
>>> Your suggestion sounds reasonable, but because plot limits are such a
>>> commonly used parameter, it would have to be done quite carefully.  The
>>> questions I'd ask before implementing it are:
>>>
>>>   - Are there other locations besides plot.default where xlim and ylim
>>> are specified?  I'd like to have them updated consistently.
>>>
>>>   - Are there any conflicting uses of NA for a limit in published
>>> packages?
>>>
>>>   - Which package authors would need to be told about this change, so
>>> they could make a compatible change?
>>>
>>>   - Where does it need to be documented?
>>>
>>> Duncan Murdoch
>>>
>>>>
>>>> Did you ever want to change one coordinate in xlim or ylim? It happens
>>>> to me all the time.
>>>>
>>>> x<- rnorm(100, m=5, s=1)
>>>> y<- rnorm(100, m=6, s=1)
>>>> plot(x,y)
>>>>
>>>> ## Oh, I want the "y axis" to show above x=0.
>>>>
>>>> plot(x,y, xlim=c(0, ))
>>>>
>>>> ##Output: Error in c(0, ) : argument 2 is empty
>>>>
>>>>   plot(x,y, xlim=c(0,NA ))
>>>> ## Output: Error in plot.window(...) : need finite 'xlim' values
>>>>
>>>>
>>>> I wish that plot would let me supply just the min (or max) and then
>>>> calculate the other value where needed.
>>>> It is a little bit tedious for the user to do that for herself.  The
>>>> user must be knowledgeable enough to know where the maximum (MAX) is
>>>> supposed to be, and then supply xlim=c(0, MAX). I can't see any reason
>>>> for insisting users have that deeper understanding of how R calculates
>>>> ranges for plots.
>>>>
>>>> Suppose the user is allowed to supply NA to signal R should fill in the
>>>> blanks.
>>>>
>>>> plot(x,y, xlim=c(0, NA))
>>>>
>>>>
>>>> In plot.default now, I find this code to manage xlim
>>>>
>>>>     xlim<- if (is.null(xlim))
>>>>          range(xy\$x[is.finite(xy\$x)])
>>>>
>>>> And I would change it to be something like
>>>>     ##get default range
>>>>     nxlim<- range(xy\$x[is.finite(xy\$x)])
>>>>
>>>>     ## if xlim is NULL, so same as current
>>>>      xlim<- if (is.null(xlim)) nxlim
>>>> ## Otherwise, replace NAs in xlim with values from nxlim
>>>>      else xlim[ is.na(xlim) ]<- nxlim[ is.na(xlim) ]
>>>>
>>>>
>>>> Who is the responsible party for plot.default.  How about it?
>>>>
>>>> Think of how much happier users would be!
>>>>
>>>> pj
>>>
>>>
>>> ______________________________________________
>>> R-devel at r-project.org mailing list
>>> https://stat.ethz.ch/mailman/listinfo/r-devel
>
>
> ______________________________________________
> R-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel

```