help: R on Mac's is "buggy", department reconsidering

Rob - my experience is that the crashes have to do with graphics,  
but I am going to be more systematic about this henceforth.

Stupid question: Simon, when you say to report the crash, how  
exactly do we do this? Is it written somewhere how to do this? If  
not, then where is the log and do we need to send you that via email  
or is automated somehow?

On Jan 31, 2008 11:00 AM, Rob Goedman <goedman at mac.com> wrote:

Matthew,

Thanks for your input. As some others noted, we might need some more
input to really be able to work on this problem.

I do know of a sure way to crash R.app (up to R-2.7-devel).

Below very simple lattice graphics example will crash R.app after
switching to the display to full screen (green button) and back
once or sometimes several times.

I have been caught by this problem while teaching/using R.

Could that be a/the possible reason for the crashes? By the way,
if you manually enlarge/shrink the quartz() window the problem
(crash) does not occur. Simon and I have been aware of this
issue but have not been able to find the cause. As far as I
recall it has been there for about 2 years (in fact since the
last lattice upgrade) and for me was the main reason to switch
to using an external editor. Once I figured out it was the little
green button, I felt more confident showing the nice internal
editor features to new users.

The updated graphics in R-2.7-devel does solve this issue,
but of course R-2.7 is still out some time and is not
of release quality yet.

Rob






On Jan 31, 2008, at 8:40 AM, Matthew Keller wrote:

Hello all,

I dread writing this because it may come across as being  
ungrateful to
the hard work of those (esp. Simon Urbanek) who have made an  
otherwise
fantastic R interface for the mac. Nevertheless...

I do most of my work in R using Aquamacs and had, until recently,
little interaction with the R Gui. People in my department  
(psychology
at CU Boulder) are mainly mac users, and a few of us (including the
chair and the statistics professors) are making a push to make our
department an "R" department (such that it is the package taught to
undergrads and grad students). I have been teaching a four week  
intro
to the R language this semester to students, postdocs, and faculty,
and the undergrad stats professors have taken the leap and are
teaching R in the undergrad stats classes this semester. So far so
good... but...

There is beginning to be a backlash, not from what I had expected  
("R
is too user unfriendly"... this doesn't seem to come up!), but  
rather
that R is "too buggy" and crashes too often. Since switching to  
the R
Gui to teach this class, I myself have had R crash several times
(previously a rare occurrence when I used Aquamacs). I cannot find a
rhyme or reason to when it happens. The students/colleagues in my
class who are mac users also complain about it (none of the windows
users have). And this morning, an email went out from one of the
undergrad stats teachers who said much the same thing (see his email
below).

Being one of the people who really pushed for this switch to R, I  
hate
to see it go down in flames because of this. I have considered  
trying
to teach using R via Aquamacs, but this presents its own set of
difficulties.

Does anyone out there who has experience with the R Gui for Macs  
have
any suggestions re making it more stable? Are there steps we can  
take
or advice we can give to students that would lessen the number of
crashes we are experiencing? Might the crashes be due to the newness
of the Leopard OS, etc?

I apologize for breaking the listserve rules - I have not given a  
good
explanation of the conditions under which these crashes occur. I  
could
try to get the students to document exactly when they happen and  
send
a summary back out to the list if this would be helpful.

I really want this (our department's switch to R) to go through,  
but I
am afraid the "bugginess of R" may be jeopardizing this. Any help
would be most appreciated. And as I said before, thank you to the
people who have selflessly spent their time making the R Gui - I
wouldn't even know where to start! Best,

Matt



<message from an undergrad stats prof here at CU> :

In the labs for my course, the R GUI has been buggy - i.e. will
crash at
random times. I thought this had to do with using R Commander, but
after
looking into it further, it has nothing to do with R Commander -
when I
ran through some of the tutorial/homework for my course using the R
GUI
(rather than R Commander) the crashes were still there.  I also
thought
it might have to do with computing labs (perhaps installing ALL R
packages created the problem), but a quick survey in my course
suggested
that people's personal apple computers are experiencing the same
thing.
Greg and Matt Keller also mentioned to me that some of their  
students
were having issues with the R GUI on Mac OS X. As everyone knows,
there
are A LOT of Mac users in the undergrad student body.

As far as I can tell (after discussing with Matt Keller), it has
nothing
to do with the R "engine" - it's all about the R GUI interface (for
example, I've never had a problem submitting R jobs in pseudo-batch
mode).  I get slightly less buggy-ness on my Mac (intel dual-core,
Leopard), but there are still some issues.

Anyway, no real action items here other than to note that it appears
to
be fairly common. The students are tolerating it, and I just keep
telling them that it's "free" and "free" always comes at a price. If
anyone has stumbled a more stable configuration, please let me know!



--
Matthew C Keller
Asst. Professor of Psychology
University of Colorado at Boulder
www.matthewckeller.com

-- 
Matthew C Keller
Asst. Professor of Psychology
University of Colorado at Boulder
www.matthewckeller.com