2018 Jan 08	4.1-12	Merge ChangeLog files

2017 Mar 12	4.1-11	Include directly-needed headers, update o/p for R
3.4.0.

2015 Jun 29	4.1-10	Tweak imports.

2015 Feb 11	4.1-9	Update Korean translations.
			Remove some unused assignments.

2014 Mar 28	4.1-8	Update Polish translations.

2014 Mar 24	4.1-7	Update French and German translations.
	                Fix array-overrun in gini.c (detected by valgrind
	                checks for adabag).

2014 Mar 07	4.1-6	model.frame() method could fail when the recorded
			call was rpart::rpart().

2014 Jan 25	4.1-5	Avoid abbreviation in tests/treble.R
			More comprehensive Description: field.

2013 Dec 10	4.1-4	Change pre-defined structure sizes to mitigate 
	                false positives from Undefined Behaviour Sanitizer.
	
2013 Sep 01	4.1-3	Document TMT change to predict() output.

2013 Aug 15	4.1-2	Replace calls to as.name().
	                Remove unused and un-exported rpartpl().
	                Correct plot.rpart.Rd as to where things are stored.
	                Increase version dependence to >= 2.15.0,
	                remove conditional paste0 from this package.

2013 Mar 20	4.1-1	Add ko translations, update reference output for 3.0.0.

2012 Nov 29	4.1-0	Remove rpconvert() (converted rpart2 trees).
	
			Clean up a lot of the R code.
	
			C code reformatted with GNU indent and
			-i4 -nut -ncdb -d1 -br -ce -il0 -npcs -brs
			then use whitespace-cleanup in Emacs.

			Call R_CheckUserInterrupts at each phase of
			cross-validation.

			Ensure that surrogates really are better than the
			default split (adj > 1e-10), and that at least 2
			cases of non-zero weight go each way.

			Removed unused file src/s_xpred.c.

2012 Nov 18	4.0-3	Add 'importance' to rpart object and to summary():
	                from TMT.

	                'minlength' arg for tree() method.

	                Improved handling of weights with zero fit,
	                including bugfix for ordered factors.

	                free_tree was commented out in rpart.c, and
	                could crash as the structure was not zeroed.  As a
	                precaution, pointers which are freed are NULLed.

	                s_xpred.c did not free the first instance of a tree.

	                make_cp_list used calloc but did not free.

	                Spell-check help pages and vignettes.

			Several examples needed par(xpd = TRUE) at
			default plotting sizes.

2012 Nov 11	4.0-2	Add car90 dataset from TMT.

	                'agree' in choose_surg.c needs to be double
	                for fractional weights.

	                Not use paste0() so R 2.14.x works.

2012 Oct 26	4.0-1	Merge in updates from TMT: see inst/NEWS.Rd 
	                and below

	                Use .Call for all C code.

	                Use an environment in the package for 'parms' from
	                plot.rpart().

	                Force byte-compilation for consistency.

2012 Oct 03	3.1-55	Force use of registered symbols in R >= 3.0.0
	                Update Polish translations.
	                Work on message formats.
	                text(fancy = TRUE) gains a 'bg' argument.

2012 Jun 27	3.1-54	Add Polish tranlations.

2012 Jun 01	3.1-53	rpart, rpart.matrix: allow backticks in formulae.
	                tests/backtick.R: regession test

2012 Mar 04	3.1-52	src/xval.c: ensure unused code is not compiled in.

2012 Jan 11	3.1-51	Change description of 'margin' in ?plot.rpart
			as suggested by Bill Venables.

2011 Apr 09	3.1-50	Change licence to GPL-2 | GPL-3
	                Remove set-but-unused variable in src/xval.c

2011 Mar 06	3.1-49	Update testall.Rout.save for R 2.13.0

2010 Dec 08	3.1-48	Avoid partial match to args, unnecessary as.vector.
			Update reference output for survival change.
			Correction to plot.rpart(compress = TRUE) from
			Stephen Milborrow.

2010 Nov 03	3.1-47	Update rpart-Ex.Rout.save for 2.12.x

2010 Jan 03	3.1-46	Update rpart-Ex.Rout.save

2009 Jul 28	3.1-45	Add rpart-Ex.Rout.save file

2009 May 18	3.1-44	Add German translations
	
2009 Mar 09	3.1-43	Spelling in man/snip.rpart.Rd.
			Spacing issue in tests/testall.Rout.save
			Remove environments from fit$functions if basic
	                plotcp() allows 'ylim' to be passed in

2008 Oct 21	3.1-42	Make use of 1L etc, update plot.rpart to use dev.new

2008 Apr 10	3.1-41	Add Russian translations

2008 Mar 28	3.1-40	cosmetics on .Rd files, Date: field in DESCRIPTION

2008 Feb 18	3.1-39	summary.rpart was missing a drop=FALSE.

2007 Oct 03	3.1-38	Remove obsolete \non_function{} notation.

2007 Jul 26	3.1-37	Correct spelling errors in man pages
			DESCRIPTION: GPL-2 only
			point to www.r-project.org for GPL-2.

2007 Jun 12	3.1-36	Qualify nchar() where needed
			Update tests/testall.Rout.save for 2.6.x
			Add reference to usersplits.R in ?rpart.

2007 Feb 23	3.1-35	Correct 'label' in text.rpart
	 		C-level formatg is replaced by sprintf.

2006 Dec 24	3.1-34	Spelling corrections

2006 Nov 29	3.1-33	Use control=NULL in deparsed calls

2006 Sep 26	3.1-32	Missing 'drop=FALSE' in rpart, add tests/surv_test.R

2006 Sep 04	3.1-31	Add depends on standard packages.

2006 Jul 05	3.1-30  Update output for R 2.4.0's naprint
			Expand the LICENCE, and install it
			Update cu.summary.rda

2006 Apr 13	3.1-29  Update tests output for changes in all.equal

2005 Dec 30	3.1-28	Use registered symbols in .C/.Call

2005 Dec 09	3.1-27	Add French and en@quot translations.

2005 Nov 15	3.1-26	Add back entry-point registration.

2005 Nov 09	3.1-25	Drop obselete test for existence of .checkMFClasses

2005 Oct 17	3.1-24  Add missing drop=FALSE in na.rpart.
			Clarify predict.rpart.Rd and rpart.object.Rd.
			Add na.action arg to predict.rpart (instead of
			using the na.action used during fitting).

2005 Apr 15	3.1-23	Use xpd=NA in example(rpart)

2005 Feb 01	3.1.22	Improve error messages for possible translation.

2004 Nov 17     3.1-21  Change logic for setting params on a device 
	                in plot.rpart.
			text.rpart.Rd: Mention use of xpd=TRUE.

2004 Aug 25     3.1-20  Stop attempts to plot a degenerate tree

2004 Aug 03     3.1-18  Conversion for R 2.0.0 & LazyData

2004 Jun 22     3.1-17  Fix possible use of uninitialized `split' in bsplit.c
			Add drop=FALSE for probs prediction for a single case.

2004 Jun 06     3.1-16  Replace long* by int* in rpartexp2.c

2003 Dec 08     3.1-15  Update NAMESPACE for R 1.9.0
	                Capitalization issues in DESCRIPTION file

2003 Nov 18     3.1-14  Test newdata types in predict.rpart
	                Correct documentation for `y' in rpart.Rd

2003 Jul 20     3.1-13  Remove unused vars

2003 Mar 15	3.1-12	Update NAMESPACE file
			Use post not post.rpart

2003 Mar 03	3.1-11	Reinstate rpart.matrix etc, as ipred used it (even
			though they were documented as for use in rpart).

2003 Mar 01	3.1-10	Use namespace, REprintf.

2002 Dec 10	3.1-9	Apparent typo in rpartcallback.s spotted by 
			Torsten Hothorn.
			formatg uses e+/-0n not 00n under Windows

2002 Jun 20	3.1-8	Remove use of registration fiasco

2002 Jun 05	3.1-7	T -> TRUE in tests

2002 Mar 26	3.1-6	based on rpart3 'release'.
			Bug fix from TMT for empty classes in training set.
			Bug fix for prediction from root-only tree.
			Register .C/.Call entry points.
			Add PACKAGE= to .C/.Call calls.
			Replace is.Surv by its definition
			Don't need FUN1 in text.rpart any more

2002 Jan 14	3.1-5	path.rpart needs descendants(), node.match().

2002 Jan 04	3.1-4	Allow ylim to be passed to plotcp.
			Add NAOK=TRUE to formatg.
			Workaround for multiple symbols for MacOS X.

2001 Nov 11	3.1-3	Change to zero-split case in rpart.
			Fixes from TMT re pruning single-node trees.

2001 Sep 25	3.1-2	Fixes to predict.rpart, residuals.rpart.

2001 Aug 23	3.1-1	Further fixes from TMT, xpred.rpart was not
			intepreting fit$parms correctly.

2001 Aug 08	3.1-0	Further updates from TMT.
			More documentation updates and corrections.

2001 Jul 25	3.0-2	Correct documentation for predict.rpart.
			Remove left-over frame$yprob in residuals.rpart.
			Use >= vs < in the labels for continuous splits.

2001 Jul 03	3.0-1	Restore use of FUN1 in text.rpart, as NAs are
			handled differently in R.

2001 May 25	3.0-0	New sources from TMT with user-specified splits.
			Use format(nsmall=) and naresid/naprint from
			  R 1.3.0.
			Change na.rpart to make use of passing down the
			  terms attribute in R 1.3.0.
			Explicitly get/set .rpart.parms* in user workspace.
			
2001 Mar 31	2.0-3	Add priority: recommended
			Re-licence under GPL2

2000 Dec 05	2.0-2	Update for R 1.2.0: more careful use of malloc

2000 Aug 12	2.0-1	Update for 2000/02/25 release of rpart, which
			added case weights

2000 Feb 07	1.1-2	Header file changes for 0.99.0 (especially re error)
			Escape # in post.part.Rd and text.rpart.Rd

1999 Dec 15	1.1-1	New version which uses weights.
			Change all occurrences of longs.

1999 May 11     1.0-7   Fix bug in graycode.c, from TMT.
			All the examples now run correctly.

1999 Apr	1.0-6	Add index of (test) datasets.

1999 Feb 24	1.0-5	Correct rpart.branch.s to get(parms, inherits=TRUE).
			Add tolerance to tree.depth, needed for the Windows 
			  version.

1999 Jan 06	1.0-4	Remove model.frame.rp, which was no longer needed
			  now model.frame.default uses xlevels.
			Modified rpart.matrix to allow - in model formulae.
			Examples are now all executable (or commented out).

1998 Jul 24  	1.0-3	Added identify.rpart to work around R's limited
			  identify.
			text is now generic in R, so removed from zzz.R.
			levels is now generic in R, so removed from zzz.R.
			Manual pages re-converted with Sd2Rd version 0.3-1.
			predict now uses xlevels to force agreement of
			  levels of factors in newdata.

1998 Jun 22  	1.0-2	Manual pages converted with Details section
			snip.rpart implemented.

1998 Jun 16  	1.0-1	Original port

------------------- Former file PORTING -------------------
src/*.{h,c} long -> int

man/*.Rd  convert *.d by Sd2Rd

Don't implement code using naresid, which R does not have.
F -> FALSE, T -> TRUE

R/labels.rpart.s	replace call to prlabel by S code.
R/model.frame.rpart.s 	deparse calls
R/na.rpart.s		as x has no attributes in R, redesign
R/plot.rpart.s 		change frame=0 to .GlobalEnv
R/post.rpart.s		change `title' trickery
R/print.rpart.s		remove nsmall in call to format
			attr(x, 'ylevels') not 'ylevel'
R/rpart.branch.s	get with inherits not frame=0
R/rpart.matrix.s	adjust for different terms structure
R/rpart.s		single -> double
			sys.parent() -> sys.frame(sys.parent())
R/rpartco.s		remove frame=0 several times
R/snip.rpart.mouse.s	drop frame=0
R/summary.rpart.s	remove justify="left" in format
			remove comma in paste(... ,,collapse ...)
			attr(x, 'ylevels') not 'ylevel'
R/text.rpart.s		text.default prints "NA", so remove these
			remove density=0 in calls to polygon
R/xpred.rpart.s		single -> double
			sys.parent() -> sys.frame(sys.parent())
R/zzz.R			a few missing functions.


------------------- Former file ChangeLog.TMT -------------------
The changes documented since 3/2002, which is the date stated in Brian's 
version of the Description file.

  In Sept 2012 the current R version and the Mayo version were merged.

14March02: When y was a factor, with no instances of one of the levels in
the "middle" of the levels list, the program would generate NA due to
division by 0.  Fairly simple fixes to gini.c and rpart.s.  The bug, and
a very nicely documented test case, was supplied by Matthew Wiener.

8Aug02: Small bug pointed out by Kai Yu in rundown.c - a missing pair of {},
which only apply if usesurrogate < 2 and there are lots of missings.  Appears
that the se of the xval error would be too small.

29Oct02: Error in xpred.rpart.s, for a user split with a response vector >1,
the "yback" vector was too short.  (Diff the lines creating eframe with those
of rpart.s, to see the obvious oversight!)  Leads to a core dump.

29Oct02: Fix error in branch.c, which was not watching out for missing
values in the surrogate variable.  Found due to close reading of the
C code by Kai Yu.  (I'm impressed!)

12Nov02: Add the "return.all" argument to xpred.rpart, to fit the needs of
Dan Schaid.  Add the test case xpred.s to test it.

28Nov02: Major changes to how indexing is done.  In the older version, at a 
lower branch on the tree one would find the following code again and again:
	for (i=0; i<n; i++) {
	    if (which[i] = current_node) {
		 ... all the work ...
	        }
If a node only had 20 of n=20,000 observations in it, the code still walked
through all 20000.  
   In the new code, the chain of routines is called with the pair
(n1, n2).  Consequences:

   a. The repeated code fragment is now the much more efficient
			for (i=n1; i<n2; i++) {
A trivial test (continuous y, n=20000, 20 xvals, cp=0 to force deep trees)
gave unix.time values of 107 vs 2526 seconds.  This was a bit of a shock -
I never assumed that quite so much of the time could be spent in
inefficient bookkeeping.  With shallower trees I would expect far less
gain, maybe 2 fold (at n=2000 it was 10 vs 32 seconds).
   b. In nodesplit.c, extra work has to be done to modify the rp.sort
array.  See comments in the last block of nodesplit for a full
explanation.
   c. Under the new regime, it turns out to be important to NOT sort
the columns of the x matrix.  As a consequence:
	1. There is no advantage to calling rpart with as.single(x) to
	  save memory.  The copy=F arg is a more efficient use of
	  resources.  (If the S manuals can be believed....)
        2. Internal variables have been changed to 'double', by redefining
	  the constant "FLOAT" in rpart.h.
	3. Since X is not disturbed, it is now safe to use the is_na
	macro to test it, rather than passing the result of is.na(X) as
        a separate argument to the C code.  This resulted in further
	algorithmic time savings in nodesplit.c and branch.c (depending on
	the number of surrogates that need to be used in splitting).
   d. Because tree building mixes up the rp.sorts array, we need to save
an original copy for cross-validation.  So the rewrite is not a total win --
we use somewhat more memory.  Subsetting out a group of subjects
turns out to be essentially the same code as was added to nodesplit.

29Nov02: If the primary split sends exactly the same number left and right
(same total weights actually), there should be no "go with the majority"
default.  Prior code always would choose left in this case, now chooses
neither.

29Nov02: The length of the response vector (numresp) added to the output
structure.  It was needed for xpred.rpart.

2Dec02: Based on concerns of a user (whose email I currently can't find) about
accuracy with asymmetric loss matrices, I spent the last 2 days creating
a completely worked example.  Their concerns, unfortunately, were completely
justified.
   a. With L=loss asymmetric, the altered priors were computed wrong -- they
were using L' instead of L!  Upshot -- the tree would not not necessarily
choose optimal splits for the given loss matrix.  Once chosen, splits were 
evaluated correctly.  The printed "improvement" values are of course the
wrong onese as well.  It's interesting that for my little test case, with
L quite asymmetric, the early splits in the tree are unchanged -- a good split
still looks good.  Simple fix to gini.c.
   b. Fixed a minor bug in the Gini criteria -- the printed "improvement"
values were too large by a constant = (#classes -1).  Since they are only used
to select the largest value, there was no practical impact.

17Jun05: Fix offset bug -- the line in rpart.s didn't actually get
the offset, but rather the variable number of the offset.
 
27March05: Finish up the test suite for priors.  Run test suite.  Code is
ready for distribution.