R with GUI

[MarkW]

After quite a bit of work, I have re-built R with the Rcmdr GUI, and will be re-releasing it within the next week (I shelved work on RStudio for the time being - I ran into a lot of brick walls and am still smarting from the experience). :unsure:  Rcmdr uses tcl/tk, so the interface isn't exactly sparkling (hey, this is statistics  ):

but it may help those who run from the command line!

I will be pre-installing as many of the Rcmdr packages as I can, but if anyone wants any other packages (regular or otherwise) pre-installed in the base pnd, please let me know. I know some of the regular packages require access to a toolchain for pre-compilation, so now is your chance to get stuff into the next release - more or less hassle free. Whilst I will do my best with the additional packages - 'fraid there are no guarantees and it is entirely possible I may miss some shared libs the first time around (not even I'm mad enough to test all of the packages 100%).

If, going forward people find they need more 'stuff', just let me know and I will include any additional packages I can in future maintenance releases.

 

[Askarus]

Why don't you port RKWard?

 

[MarkW]

Why don't you port RKWard?

Two reasons. First RCmdr basically allows you to extend the functionality of R infinitely by writing your own plugins - tcl/tk isn't exactly rocket science.

Second, I do not fancy wrestling with half of the K libs just for one pnd.

Other than that, it would be a grand idea.

 

[Askarus]

I see.

Then I have to use the one from PanDebian.

I need to do psychological statistics.,

As RKWard looks like SPSS it's the easiest frontend for me.

 

[MarkW]

As RKWard looks like SPSS it's the easiest frontend for me.

It seems you had your solution before you posed the question. If it fits your needs, go for it.

 

[Askarus]

I just mean it'd be nice not to use PanDebian.

Also prefer LibreOffice as PND.

If you refer to SPSS it's only the program we had some lesson at university.

Now I will completely switch to R and RKWard.

My teacher supports me as well.

 

[peelie]

great news!

can you add these for nlp processing?

http://cran.r-project.org/web/views/NaturalLanguageProcessing.html

thanks!

 

[MarkW]

great news!

can you add these for nlp processing?

http://cran.r-project.org/web/views/NaturalLanguageProcessing.html

thanks!

I'll see what  can do. Why do I get the feeling this pnd is going to grow - rather big! Its already @>50meg - bare bones (and compressed)!

 

[peelie]

great news!

can you add these for nlp processing?

http://cran.r-project.org/web/views/NaturalLanguageProcessing.html

thanks!

I'll see what  can do. Why do I get the feeling this pnd is going to grow - rather big! Its already @>50meg - bare bones (and compressed)!

hehe at a minimum these packages from list above would be great to start:

openNLP and openNLPmodels, gsubfn, languageR, zipfR, tm.

cheers

 

[MarkW]

great news!

can you add these for nlp processing?

http://cran.r-project.org/web/views/NaturalLanguageProcessing.html

thanks!

I'll see what  can do. Why do I get the feeling this pnd is going to grow - rather big! Its already @>50meg - bare bones (and compressed)!

hehe at a minimum these packages from list above would be great to start:

openNLP and openNLPmodels, gsubfn, languageR, zipfR, tm.

cheers

groan - what have I started??!!! :huh:

 

[MarkW]

I just mean it'd be nice not to use PanDebian.

Also prefer LibreOffice as PND.

If you refer to SPSS it's only the program we had some lesson at university.

Now I will completely switch to R and RKWard.

My teacher supports me as well.

Its just the amount of work. Yes, RKward is preety - and would probably be in greater demand the Rcmdr. However, Personally, the CLI is enough for my needs. People asked for a GUI, so I'm providing one - it may not suit everyone, but then again nothing ever will.

Even after figuring out the overall approach of building with Rcmdr & tcl/tk (and that took some head scratching), it still took me the best part of a day to put the the basic package together so it runs from a read-only, stand-alone environment. LDD doesn't fully report all linked libraries, so you have to use an iterative process of building, packing, testing just to get the thing running. I only had 8 libraries missing after the initial LDD test - so that was only 8 pnd rebuilds and tests (and reboots to guarantee a completely clean unmount of codeblocks) to identify the basic missing libraries. Given this, and the fact that the development team of RKward cannot be bothered to specify which libraries are needed (visit their page to see what I mean - its all pretty vague), I'm not going to spend my time poking around completely in the dark. I relish a challenge, but there are limits.

At least with tcl/tk, if I include everything of both packages (as I have done) - aka the blunderbuss approach, things are moving in the right direction at a fair pace - and size. If I take the same approach with K ... who knows? Everything would have to be lifted, moved, integrated, packed, tested etc etc. Groan. I'm dedicated to the cause, but not to that extent. If the developers of RKward were more diligent in their documentation, I would be motivated - but they're not, so it doesn't even appear on my radar. End of discussion I think.

 

[MarkW]

great news!

can you add these for nlp processing?

http://cran.r-project.org/web/views/NaturalLanguageProcessing.html

thanks!

I'll see what  can do. Why do I get the feeling this pnd is going to grow - rather big! Its already @>50meg - bare bones (and compressed)!

hehe at a minimum these packages from list above would be great to start:

openNLP and openNLPmodels, gsubfn, languageR, zipfR, tm.

cheers

I take it you are not interested in the Spanish version of openNLPmodels?

Also, rjava is causing issues with one of your packages (as a dependency). What sort of java environment have you got on your OP?

So far, openNLPmodels.en, gsubfn, tm and languageR all seem to have gone in without problems. openNLP fails because of the rjava dependency. Haven't tried zipfR, but since it doesn't have any dependencies, I'm not anticipating problems with it.

I may well back up a couple of stages, and recompile R slightly differently. To date, I haven't explicitly included java support (as I didn't realise java was used outside the help system)! TBH, I'm not entirely sure how R&java will/can work together on the OP - but I'm willing to give it a bash and see!

2nd edit (!)

I've had a look at WizardStan's java Runtime pnd. This now includes the Oracle jdk - which means I can recompile R with full java support. Obviously with this approach I'm tied to that pnd for Java support, (and the java pnd would need to be mounted for things to work) - but this approach could solve all sorts of java path problems, effectively removing the moving target that would otherwise inevitably occur if I made assumptions about other people's installations.....worth a punt?

Let me know your thoughts.

 

[Askarus]

I just mean it'd be nice not to use PanDebian.

Also prefer LibreOffice as PND.

If you refer to SPSS it's only the program we had some lesson at university.

Now I will completely switch to R and RKWard.

My teacher supports me as well.

Its just the amount of work. Yes, RKward is preety - and would probably be in greater demand the Rcmdr. However, Personally, the CLI is enough for my needs. People asked for a GUI, so I'm providing one - it may not suit everyone, but then again nothing ever will.

Even after figuring out the overall approach of building with Rcmdr & tcl/tk (and that took some head scratching), it still took me the best part of a day to put the the basic package together so it runs from a read-only, stand-alone environment. LDD doesn't fully report all linked libraries, so you have to use an iterative process of building, packing, testing just to get the thing running. I only had 8 libraries missing after the initial LDD test - so that was only 8 pnd rebuilds and tests (and reboots to guarantee a completely clean unmount of codeblocks) to identify the basic missing libraries. Given this, and the fact that the development team of RKward cannot be bothered to specify which libraries are needed (visit their page to see what I mean - its all pretty vague), I'm not going to spend my time poking around completely in the dark. I relish a challenge, but there are limits.

At least with tcl/tk, if I include everything of both packages (as I have done) - aka the blunderbuss approach, things are moving in the right direction at a fair pace - and size. If I take the same approach with K ... who knows? Everything would have to be lifted, moved, integrated, packed, tested etc etc. Groan. I'm dedicated to the cause, but not to that extent. If the developers of RKward were more diligent in their documentation, I would be motivated - but they're not, so it doesn't even appear on my radar. End of discussion I think.

Thanks for the clear information .

 

[peelie]

hi not interested in spanish nlp, an interested in english and french.

i have java pnd installed so your approach sounds good.

cheers

 

[MarkW]

hi not interested in spanish nlp, an interested in english and french.

i have java pnd installed so your approach sounds good.

cheers

'tis moving forward - slowly. I mount the java.pnd from a shell script, which will eventually be done through the pnd run.sh. I'll put a little logic and a bit of zenity stuff in there as there are a couple of gotcha's that could catch people out (java could vanish from underneath R if someone launches a .jar application and then exits it, since this causes the pnd to unmount - irrespective of who mounted it). However, as I had hoped, this approach does indeed make the full java environment visible to R. R picks up all it needs directly from setting just two environment variables which are set in a modified version of bin/R. All that is working well. Still not quite got rJava in, although I think I'm through with the Pandora specific problems and am now just encountering the same problems everyone else seems to have  :blink:

There will be a requirement to use a specific version of the java pnd, and it is likely that I will also say that the java.pnd will also have to be in the same pandora directory as R. Other than these few things, I think this will work out OK.

Also, re you msg. I do not see a french OpenNLPmodels.fr package....

 

[peelie]

my bad  about there being a french package

 

[peelie]

another request - shiny package

thanks!

 

[MarkW]

another request - shiny package

thanks!

FInally nailed rJava - now celebrating with a cool beer! 

I see, you give me challenges - I eventually rise to them, and them you give me ........... more!!!! :rolleyes:

 

[peelie]

another request - shiny package

thanks!

FInally nailed rJava - now celebrating with a cool beer! 

I see, you give me challenges - I eventually rise to them, and them you give me ........... more!!!! :rolleyes:

good news! enjoy yr beer

 

[MarkW]

another request - shiny package

thanks!

All of your packages (and their dependencies) are now in. No errors were reported. Be very sure to test them carefully when I release the pnd; I haven't a clue what they do beyond a very superficial understanding   and therefore cannot test them. None of the packages reported linking to  external shared libs that I noticed (which is my main concern) - but its always possible I have missed something, or that something didn't get reported properly during the install output.

I've got a bunch of snapshots of the build in progress, so I can always roll back to any one of them and rebuild if necessary.

If you have any more requests, let me know. I'm looking to release the pnd by the end of this coming working week. In the meantime, I'll clear up the rest of the Rcmdr packages, sort out the run.sh script to include sensible ish mounting of the java pnd and prepare a thread post to accompany the pnd release.