Skip to content

Rcmd install mypkg

13 messages · Uwe Ligges, Dirk Eddelbuettel, Gabor Grothendieck +2 more

Original
Duncan Murdoch
#
On 9/30/2006 10:54 PM, Gabor Grothendieck wrote:
This is a documented change--see the CHANGES file.

Duncan Murdoch
Gabor Grothendieck
# 
This would seem to violate the user interface stereotype that
command lines are case insensitive on Windows.
On 10/1/06, Duncan Murdoch <murdoch at stats.uwo.ca> wrote:
Uwe Ligges
#
Gabor Grothendieck wrote:
This is an improvement. R tries to be as unique as possible on all 
platforms.

Uwe Ligges
Gabor Grothendieck
#
On 10/1/06, Uwe Ligges <ligges at statistik.uni-dortmund.de> wrote:
Not sure what the above means but if you mean it works the same
on all platforms that is an error in the design of the software in my view.
The objective should be that it fits into the environment its working in.

Any Windows user can tell the difference between "clunky" UNIX-derived
software which does not feel right on the Windows platform and
software that adapts its interface to work like all other Windows programs
in that environment.
Dirk Eddelbuettel
#
On 1 October 2006 at 08:53, Gabor Grothendieck wrote:
| On 10/1/06, Uwe Ligges <ligges at statistik.uni-dortmund.de> wrote:
| > Gabor Grothendieck wrote:
| > > This would seem to violate the user interface stereotype that
| > > command lines are case insensitive on Windows.
| >
| > This is an improvement. R tries to be as unique as possible on all
| > platforms.
| 
| Not sure what the above means but if you mean it works the same
| on all platforms that is an error in the design of the software in my view.
| The objective should be that it fits into the environment its working in.
| 
| Any Windows user can tell the difference between "clunky" UNIX-derived
| software which does not feel right on the Windows platform and
| software that adapts its interface to work like all other Windows programs
| in that environment.

Generally speaking, I think many of us would beg to differ and much rather
(continue to) see consistency of R across platforms.

Personally speaking, I would think that taking what I consider to be a clunky
OS and UI to be the gold standard as somewhat questionable. But that may just
be me.

Cheers, Dirk
Gabor Grothendieck
#
On 10/1/06, Dirk Eddelbuettel <edd at debian.org> wrote:
I don't want to get into a debate on the merits of an OS but in terms of
user interface principles one wants to fit into the environment that the
software is working in rather than have the user familiarize themselves
with multiple styles of interaction.

If you are running UNIX and Windows you can expect differences when
you move from one to the other but you do not expect differences when
running one program vs. another in the same OS.
Duncan Murdoch
#
On 10/1/2006 7:20 AM, Gabor Grothendieck wrote:
And command options use a / prefix in Windows, but a - or -- prefix in 
R, and the path separator in Windows is \ in most places, but / is 
accepted in R.  There are a large number of inconsistencies between 
Windows and other systems, and R mostly follows the GNU conventions; 
it's a GNU project, after all.

Duncan Murdoch
Gabor Grothendieck
#
On 10/1/06, Duncan Murdoch <murdoch at stats.uwo.ca> wrote:
That's not a bad point but is there really any reason to make it even
less windows-like?  After all, it did work in either case before and
windows filenames are case insensitive so its not like there is
a valid distinction between install and INSTALL if the word refers
to a filename.
Duncan Murdoch
#
On 10/1/2006 1:10 PM, Gabor Grothendieck wrote:
This is why Rgui tries to follow Windows conventions (not entirely 
successfully, and about 10 years behind the times, but anyway...). 
However, the command line utilities are used in the build process for 
packages and for R itself, and so it makes maintenance a lot easier for 
those to follow consistent cross-platform conventions.

Requiring Perl script names to be specified in a case-sensitive way 
doesn't affect the R build process, but it does mean that a script 
written on Windows is slightly more likely to work when someone runs it
on another platform.  There's an "install" command in at least some 
flavours of Unix that would be invoked instead of R's INSTALL script if 
you said "R CMD install" on one of those platforms, and that's not a 
good thing.

So what I'd suggest you do if you find typing INSTALL to be painful is 
to install an alias or command script that wraps it up in a way that 
feels more comfortable to you.

Duncan Murdoch
Gabor Grothendieck
#
On 10/1/06, Duncan Murdoch <murdoch at stats.uwo.ca> wrote:
It would be possible to do some processing in Rcmd.bat in batchfiles
but I am somewhat loathe to do that since it would result in a potential
difference between Rcmd.bat in batchfiles and Rcmd delivered with R
and right now its entirely consistent -- it just passes the command line
right through.
Hadley Wickham
# 
Will this be changing on OS X too?  My version of 2006-09-20 r39433
accepts either upper or lower case commands.

Hadley
Duncan Murdoch
#
On 10/1/2006 4:30 PM, hadley wickham wrote:
I think OS X uses the same scripts as other Unix platforms.  So if 
they're happy working on a case-insensitive file system, it's unlikely 
we'll force a change, whether it would be a good idea or not.

In fact, I'm fairly ambivalent about this one:  I'm a long-time believer 
in case-insensitivity.  I'd prefer if R were case-insensitive in all 
respects.  However, it's not, so consistency pulls me to case-sensitivity.

Duncan Murdoch
Gabor Grothendieck
# 
I think the really annoying thing is that it was changed for the worse.
Why not just keep it how it was before?
On 10/1/06, Duncan Murdoch <murdoch at stats.uwo.ca> wrote: