NEWS | R Documentation |
cosmetic changes in help pages
add angular distance/similarity; keep old behavior as cosine
Change coercion function for cosine similarity to distance to 1-x instead of 1-abs(x)
fix scaling for gower metric in case of cross-proximities to scale x and y together.y
fix tests in apply.R to comply to new default behavior of encoding character strings in data frames
smaller changes in C code to prevent false positives from automated CRAN code checking
Change registry class name to proxy_registry
to
prevent name clashes with registry package.
some more C-code PROTECT
ing.
print method for simil
objects added that prints
NA
instead of 0 for diagonal elements, if requested.
fix for binary distance (C-code): now has same behavior as the corresponding distance method in the stats package in case of missing values.
pr_simil2dist
now computes 1 - abs(x)
instead of 1 - x
.
add imports from recommended packages.
export as.matrix() generic.
Make binary similarities work with binary data in case of list data (by coercing list to a matrix).
Bug in Fager/McGown measure fixed.
Author/Maintainer removed.
PACKAGE="proxy" argument added to all .External-calls. "PACKAGE" field added to the registry, defaulting to "proxy", and used by .Call().
the method argument of dist()
now also accepts a
registry entry to avoid unnecessary lookups in repeated calls to
dist()
.
bugfix: names<-.dist
did not accept NULL
as assignment.
small vignette with an overview of existing methods added.
Podani, Chord, geodesic, Hellinger and Whittaker dissimilarities added.
Fager/McGown anf Faith similarities added.
pr_dist2simil
now computes 1 / (1 + x)
instead of 1 / (1 - x)
fix labeling bug for list data with named components
the C-level dist functions now conform with those in package stats with respect to NA handling (pairwise-complete observations).
as.matrix.dist
by default now uses 0 for diagonal elements
fix a bug in Kullback-Leibler-distance
fixed Gower-dissimilarity
fixed dist/simil subsetting
self-proximities now default to NA
data frames are auto-converted to matrices if all variables are
either numeric, logical, or complex (to be compatible to
stats::dist
)
pairwise distances can now be computed
Registry is now case-insensitive
Initial release