[Rd] `mgp[1:3]' are of differing sign (PR#14130)

[Peter Ehlers]

Peter Dalgaard wrote:
> Peter Ehlers wrote:
>>
>> cornell.p.gonschior at iem.fh-friedberg.de wrote:
>>> Full_Name: Cornell Gonschior
>>> Version: 2.10.0
>>> OS: Linux
>>> Submission from: (NULL) (212.201.28.40)
>>>
>>>
>>> Hi,
>>>
>>> in the introduction to R, you can find the following sentence in the 
>>> par()
>>> chapter:
>>> "Use tck=0.01 and mgp=c(1,-1.5,0) for internal tick marks."
>>> I thought that's nice, because I wanted to have tick marks and tick 
>>> labels
>>> inside and the axis title outside.
>>>
>>> But:
>>>> plot(z, las=1, tck=0.01, mgp=c(1,-1.5,0))
>>> Warnmeldungen:
>>> 1: In plot.window(...) : `mgp[1:3]' are of differing sign
>>> 2: In plot.xy(xy, type, ...) : `mgp[1:3]' are of differing sign
>>> 3: In axis(side = side, at = at, labels = labels, ...) :
>>>   `mgp[1:3]' are of differing sign
>>> 4: In axis(side = side, at = at, labels = labels, ...) :
>>>   `mgp[1:3]' are of differing sign
>>> 5: In box(...) : `mgp[1:3]' are of differing sign
>>> 6: In title(...) : `mgp[1:3]' are of differing sign
>>>
>>>> par(las=1, tck=0.01, mgp=c(1,-1,0))
>>> Warnmeldung:
>>> In par(las = 1, tck = 0.01, mgp = c(1, -1, 0)) :
>>>   `mgp[1:3]' are of differing sign
>>>
>>> Was there a recent change, couldn't find anything useful searching 
>>> the web.
>>>
>>> Regards,
>>> Cornell
>>
>> Well, it's only a warning, making you aware of a possibly
>> unintended par setting. Warnings are good things but if you
>> don't want to see them, they can be suppressed.
>>
>> Certainly not a bug.
> 
> Hmm, then again, I tend to agree with Cornell that there are a bit too 
> many cases where mgp[1:3]' would sensibly have differing sign, compared 
> to cases where it is a mistake. In addition to internal tick marks and 
> labels, there are also cases where the whole axis is shifted into the 
> plot area. I'd more likely use axis(pos=...) for that, but still.
> 
And, I suppose, if R-Intro mentions it, it would be less confusing
if there were mention of the warning. But use of R conditions one to
seeing warnings as a good thing. Can't count the number of times
I've seen "longer ... is not a multiple of shorter ..." and having
it catch a silly user error.

-- 
Peter Ehlers
University of Calgary
403.202.3921