RELEASE NOTES 

		spatstat 1.8-6

	 	23 February 2005

	
	We thank Agnes Gault, Mandy Hering and Mark Stevenson
	for contributions.


OVERVIEW

	Many bugs in the handling of empty point patterns have been fixed.
	Various minor improvements are introduced.


NEW FUNCTIONS

	unique.ppp
	duplicated.ppp

		new methods for `unique' and `duplicated'
		for point patterns. 

		These make it possible to identify and remove 
		any duplicated points in a point pattern.

	multiplicity.ppp

		Counts the number of duplicates of each point
		in a point pattern.

	dilate.owin
		performs morphological dilation of a window
		(compare erode.owin)
	

IMPROVEMENTS/ENHANCEMENTS

	empty point patterns
		In spatstat, a point pattern object is allowed to
		be empty (i.e. to contain no points). Such patterns
		are occasionally generated in simulations. 
		In this release we have improved the handling of
		empty patterns throughout spatstat, and fixed 
		several bugs [1]-[5] which occurred with empty patterns.

	diagnose.ppm
	plot.diagppm
	plot.im
	persp.im
	contour.im
		These functions now pay (more) attention to 
		their "..." arguments which modify the plot behaviour.

	qqplot.ppm
		Now prints some diagnostic information about the
		simulated point patterns.

	summary.ppp
		Prints a message if the point pattern 
		contains duplicated points.

BUGS

	[1] ppm
		If a pairwise interaction model was fitted to
		an empty point pattern,	ppm() crashed with the message
			"Error in apply(POT, c(2, 3), sum): 
			dim(X) must have a positive length".
		This was a bug in pairwise.family.
		Fixed.

	[2] qqplot.ppm
		Sometimes crashed with the error message
			"Error in apply(POT, c(2, 3), sum): 
			dim(X) must have a positive length".
		This was an instance of bug [1]. It occurred randomly,
		when one of the simulated point patterns was empty.
		Eliminated.

	[3] qqplot.ppm
		In some cases the simulated patterns were ALL empty.

		This is a quirk of the simulation method.
		The default behaviour of qqplot.ppm is to simulate patterns 
		on an expanded window (specified through the argument
		`control') in order to avoid edge effects.
		The trend is extrapolated over this expanded
		window. If the trend is strongly inhomogeneous, the 
		extrapolated trend may have very large (or even infinite)
		values. This causes the simulation algorithm to 
		produce empty patterns.

		This problem was most likely to occur when the 
		original data were in a non-rectangular window.
		In that case the default expansion window (computed by
		default.expand) was much larger than the original window.

		We have changed the default expansion window for
		non-rectangular windows, so that the problem is
		now very unlikely to occur.

		The problem may still occur in extreme cases.
		qqplot.ppm will now issue a warning when it occurs.

		The only way to suppress this problem entirely is to
		prohibit the expansion of the window, by setting
		the `control' argument to qqplot.ppm, e.g.
			qqplot.ppm(....., 
			  control=list(nrep=1e6, expand=1))
		where `expand=1' means there will be no expansion.

	[4] print.ppp, summary.ppp, summary.quad
		Generated an obscure error message when the point pattern
		is empty. This was a bug in summary.quad.
		Fixed.

	[5] ppm
		If a uniform Poisson model was fitted to 
		an empty point pattern, the fitted model had
			$maxlogpl=NaN.
		Fixed.

	[6] plot.im
		Refused to plot an image containing logical values.
		Fixed.

	[7] anova.ppm
		Erroneously complained that `some models were not fitted
		with use.gam=TRUE' when one of the models was the uniform
		Poisson process. Fixed.

	[8] density.psp
		Yielded an image containing only NA's,
		if one of the line segments had zero length.
		Fixed.

	[9] envelope
		Did not correctly handle the argument 'r'.
		Fixed.