R help - May 2005 - working with CGIwithR

<headline>Short question</headline>: 
Do people have advice on debugging R programs running after CGIwithR 
inputting of data from forms?  Is there a way of setting up fast 
local versions if your local machine has to be a windoze (2k) machine 
(R 2.1.0) and your server is a Debian, ssh shell only set up running 
R 1.8.0?  Are there simple guides to ways of not having to invoke R 
each time you submit data from a form?  No complaints here about R or 
CGIwithR: both brilliant.  I just need some help partly because my 
hardware resources are slow.

TIA, Chris

<H>Long question for those with time and wanting more detail</H>:

I'm not a statistician though I've spent a lot of the last thirty 
years of my life working with statisticians and doing my own stats 
part time.  I've worked from roll my own in dBaseII, through BMDP, 
loads of SPSS, SAS & SAS/IML, S+ and now, over the last two to three 
years, from S+ to R and loving it.  However, at times I really hit 
the limits of my amateur skills. .... and this is one of them!

I've put together some scripts to deal with Jacobson et al.'s ideas 
of "reliable change and clinically significant" change.  They work 
fine on R 2.10 on my Win2K box and can output to ascii & a windows 
graphics plot or to HTML and jpg using R2HTML.  Now I'm trying to 
transfer them to my www server which runs Debian stable and R 1.8.0 
and I'm using CGIwithR to get the input from people.  

All probably fine but the of course I struggled with some aspects of 
getting the data entry flexible and now something is simply not 
working as I pass the data to the main bits of the program that I've 
worked out for local use.  I'm sure the problem will turn out to be 
some trivial incorrect class or the like that's afflicting one of the 
parameters or bits of data.  I won't (yet) waste time here with 
detail but I'd love advice on ways to get more debugging information 
and on speeding things up as my server is slow, the only one I can 
currently afford, and each invocation of the script takes minutes to 
turn around and even with:
  zz <-file("Rmess.all",open="wt")
  sink(zz,type=c("output","message"))
I'm not trapping any error messages, the program just appears to skp 
entering a function I've written and continue on with nary an apology 
or gentle complaint about my stupidity!  Most unlike R!

Thanks to anyone with suggestions!

Chris



-- 
Chris Evans <chris at psyctc.org>
Consultant Psychiatrist in Psychotherapy, Rampton Hospital; 
Research Programmes Director, Nottinghamshire NHS Trust, 
Hon. SL Institute of Psychiatry
*** My views are my own and not representative of those institutions 
***

On 8 May 2005 at 19:45, Chris Evans wrote:
> Do people have advice on debugging R programs running after CGIwithR
> inputting of data from forms?  
... rest of my original post snipped, I'm replying to my own post for 
the archives! ...

I had two kind responses, one shrewd one from Tom Short noting that 
you can have problems with error messages in angle brackets if you're 
looking at HTML sink output.  Noted but that wasn't the problem.  He 
and David Firth, who tells me there's a new version, 0.70, of 
CGIwithR uploaded to CRAN which helps 2.1.0 compatibility, both tell 
me I'll have to use ssh and hack things on the server from there.  
True indeed and I think one message is that you have to be prepared 
to work out how to get your data into the program directly if 
bypassing form input.  I've debugged things now and the program runs.

Now some more questions separately for healthy archiving!

Thanks all,

Chris
-- 
Chris Evans <chris at psyctc.org>
Consultant Psychiatrist in Psychotherapy, Rampton Hospital; 
Research Programmes Director, Nottinghamshire NHS Trust, 
Hon. SL Institute of Psychiatry
*** My views are my own and not representative of those institutions 
***