"[SOLVED] Rapidminer R extension [Fatal Error] Content is not allowed in prolog."

ad_astra

When trying to install the rapidminer R extension, I got a cryptic error like" Content is not allowed in prolog".  By googling around, I found this cryptic error might be caused by the XML reader barfing on a non-existent file, which in rapidminer's case is being looked for by user fragment documentation, whatever that is?  There are two warnings that pop up relative to this, and in some cases (I'm guessing it depends on your Java version?) causes the cryptic: [Fatal Error] Content is not allowed in prolog.

I was able to FIX this error by going into Rapidminer-->Tools-->Preferences-->R 

the first  field should be your library, /Library/Frameworks/R.framework/Versions/2.15/Resources/library/rJava/jri/libjri.jnilib

and I just copied in anything to the next two fields, so the XML parser wouldn't throw an uncaught exception.  So I just copied the exact same path from above (the one for the dll) into those two fields.  Of course it throws warnings because it isn't expecting the dll but it continues on with just a warning and the R extension works correctly!!!

I spent FOREVER messing with my PATH and R_HOME and JAVA_HOME thinking I was doing something wrong.  Maybe if my PATHs are set correctly then rapidminer would know where to find what its looking for, because I did actually manage to get the R extension working on my computer at some point in the past without using this funny work around.  Maybe there a command line option that can suppress that error and turn it into a warning?

There were three things that weren't explained well in the installation documentation for MacOSX users. 
1) when it asks for the dll, give it the dll listed above, I think if you pick one of other ones it also messes up, such as REngine or RLib
2) I didn't know where to change the DLL once it was set during the installation script, since it automatically rediscovers the one you set previously even if it is wrong. 
3) Whats up with the user fragment documentation warnings and in my case ERRORS, and should we be setting this manually if it doesn't get set automatically?

Best regards,
Robert

pmiddlet

One thing that's obvious here is that, for whatever reason, rapidminer is forced on startup to look for additional sources of user written code and help docs (this assumes that most people who write R code publish it, for example, internally in the R standard help doc format. However, this is generally not what I have seen in practice although ideally it should be.

Another user in this forum noticed that when they added a line to a file in ANY of the R settings (under Tools -> Preferences -> R) that all of a sudden, the prolog error (a java initiated error which is searching for an xml file to parse) disappears, although 'warnings' still persist.

Still another user found a workaround to get RM to work by successfully starting R in a Mac terminal, starting an interactive session with R, and then calling libPaths() in R. What precisely this accomplished under the hood for that user is a mystery to me.

The one thing here is certain. The code fragments and documentation, which are listed as optional, are attempting to be loaded by default. I think this perhaps a root cause of this problem. This is a software error for which one must develop a wierd workaround. This problem would be no different for the supported product.

The documentation for this software is simply not universal enough to include the Mac OS community - obviously of which is an important segment that shouldn't be disregarded. For this reason, I'm not inclined at this point to consider a full purchase of the product as I don't believe that there is demonstrated value in doing so based on these bugs, the shoddy documentation, and so forth. It's possibile that rapid-i isn't aware of these issues from the forum - else they are aware and simply don't know how to solve them. It is also possible, though I would like to believe otherwise, that they are purposefully withholding information as a simple strategy which frustrates users enough to purchase an enterprise license. This is not unheard of, but doesn't lend credibility to a company looking to profit from supported software.

I hope that one of the moderators is reading this and will move actionably to address an enormously widespread problem of complications involving inclusion of the R Extension. There is much chatter in this community and over the rest of the internet about this. So it would stand to reason that rapid-i would be working to address this problem, as other similar software integrates with R quite seamlessly. Just my $.02.