www.phidot.org

I am trying to get MARK and RMark working with Linux. I can run models but it always says

********Following model failed to run : ....
and there is only NULL under the results.

However, if I look at the out files, they say " EXECUTION SUCCESSFUL ", and all the results are there. So it seems there is some miscommunication between MARK and RMark. I am using the most recent versions of R and RMark, and the recently (earlier this year) compiled MARK version for Linux (64 bit). Help would be appreciated!

thanks!
angie

angieluis wrote:I am trying to get MARK and RMark working with Linux. I can run models but it always says

********Following model failed to run : ....
and there is only NULL under the results.

However, if I look at the out files, they say " EXECUTION SUCCESSFUL ", and all the results are there. So it seems there is some miscommunication between MARK and RMark. I am using the most recent versions of R and RMark, and the recently (earlier this year) compiled MARK version for Linux (64 bit). Help would be appreciated!

thanks!
angie

Probably because the Linux version of MARK is no longer current (my guess is the RMark plays nice with the latest mark.exe, but not with earlier versions). Gary has added a number of new features, and this will require recompiling the code (which is not as simple as the usual config - make sequence). Unfortunately, this takes more time than I have at present. Sometime between now and the end of July is about as specific as I can get. Sorry. But, if the results are in the OUT files, then at least you have workable functionality.

One option is to run RMark through wine. Both R and MARK run well through wine. Hadn't tried RMark before, but I just loaded the library and ran the dipper example and everything seems to have gone smoothly. Once you've installed wine (http://www.winehq.org/), R for Windows (http://cran.r-project.org/), and MARK, just follow the Windows directions on the RMark software page (http://www.phidot.org/software/mark/rmark/).

Note that MARK will likely pop up an error message every time you run a model, but you can simply click the 'OK' button and it will happily proceed. In wine, it does this through the gui interface as well.

Cheers...

This is discussed in the notes for Linux on the RMark page although it is not linked to the error. For Windows it assumes MarkViewer=notepad.exe and for Linux it assumes MarkViewer=pico.exe. The viewer is used to show the output in the MARK out file. If you don't have pico then it will fail. Set MarkViewer= to whatever viewer/editor that you have on your Linux machine. Simon's use of Wine must encase it in Windows so it would use notepad.

--jeff

This is what is in the help file:
One other setting is the name of the editor/file viewer. The default is pico. If you want to use something other than pico or your machine does not have pico. Simply type MarkViewer="myviewer" in your R session before using RMark, where myviewer is the name of an editor or file viewer that is in your path. You can put the library call and the MarkViewer assignment into a function .First() and they will be run each time you start the R session in that directory.

Unfortunately, this takes more time than I have at present. Sometime between now and the end of July is about as specific as I can get. Sorry.

What is the likelihood of a Win64 build too? I've recently moved computer and am missing my favourite R package. I did try compiling from the available source but failed miserably! Tx Rob

robrob wrote:

Unfortunately, this takes more time than I have at present. Sometime between now and the end of July is about as specific as I can get. Sorry.

What is the likelihood of a Win64 build too? I've recently moved computer and am missing my favourite R package. I did try compiling from the available source but failed miserably! Tx Rob

Chances are pretty small, at least for now. Tests on 32-bit vs. 64-bit Windows builds show essentially no difference in speed. Further, 64-bit R is actually slower than 32-bit R. Everyone gets all excited about 64-bits. The only guaranteed gain you get with 64-bits is ability to access a larger memory space. You do not, in generally, gain anything in terms of speed. Don't believe the hype. For those applications that gain a lot by having more of the app i RAM (video editing, for example), there is a gain, but not, generally, for most statistical applications.

Thanks for the replies. Pico isn't installed on the machine, but nano is, and in the R command line before running the mark command, I say MarkViewer="nano", and I'm still having the same problem.

Would it help to use a previous version of RMark?

thanks!
angie

Angie,
I am using Karmic with Rv2.11.1 (5/31 build) with MARK 6.0 and the most updated RMark and I use MarkViewer="gedit" right after my library(RMark) call, and it works fine for me (I went back and tested the last 2 RMark releases and MARK 5.1 as well, still works on my end). I had at one time gotten the issue you are hitting (although, for the life of me, I cannot recreate it right now). This may not work, but try setting a specific working directory for our output files from MARK, e.g., use setwd() before you call mark() or your function to run your mark models and see if that helps (it should not matter, but its worth a shot).


Bret

You may also want to try using debug to see where it is failing. Type debug(run.mark.model) and then type mark(....) and step through and see where it fails. If it is failing inside a call to a function within run.mark.model then type debug(..) on that function to step into it as well.

When you track this down please post a message so as to help others down the line.

Sorry it wasn't the simple solution. I don't use Linux much so I'm not much of a help.


--jeff