All exported functions now use snake_case names, so that no function name can
be mistaken for an S3 method (e.g. create.bm previously read as a create()
method for class bm). The old names have been removed:
| Old name | New name |
|---------------------|---------------------|
| create.bm() | create_bm() |
| add.points() | add_points() |
| add.segment() | add_segment() |
| add.labels() | add_labels() |
| delete.labels() | delete_labels() |
| delete.skeleton() | delete_skeleton() |
| concat.bm() | concat_bm() |
| first.passage() | first_passage() |
sim(), layers(), refine() and plot() are unchanged.
The points() method for BrownianMotion objects, a leftover from an
earlier base-graphics version of the plotting, has been removed. Use
plot(), or bm[, "t"] and bm[, "W_t"] to draw the skeleton yourself.
The package no longer imports graphics.
print() methods for Brownian motion objects (one and several
dimensions, and the transformed view bm$Z). Printing a path now gives an
overview of its skeleton, layers, jumps and labels, followed by a table of
its points showing how each arose and the layer and bounds over the
interval to the next point, in place of the bare environment. Long paths
show only their first and last points; see ?print.BrownianMotion for the
n, digits and width arguments. Matching format() methods return the
same output as a character vector.summary() methods, which report the time range and spacing of the
skeleton, the start, end and extremes of the path, any jumps and labels,
the layers of each type with the time they cover, any gaps between layers
and the bounds which contain the whole path over the layered time, and the
default settings. The summary is a list, so these values can also be used
programmatically (see ?summary.BrownianMotion).create_bm() accepts cov as a vector of length dim, giving a diagonal
covariance matrix (as described in the tutorial vignette).U %*% t(U) instead of
cov = t(U) %*% U.concat_bm() join) now draws from the correct distribution.
Previously the right end of the auxiliary path was taken after the jump
rather than before it, giving draws outside the layer bounds.add_segment() now moves the bounds of localised bridge layers by the
correct amount.concat_bm() of multi-dimensional paths now returns a complete object;
previously bm$Z, add_points() and add_segment() failed due to cov and
chol_inv variables not being copied.plot() of a multi-dimensional path, or of its $Z view, now works when
the path has no layers.what <- "W_t"; bm[1, what]; previously only a literal string worked, and
anything else (including indexing from inside a function) returned NULL.bm[i, var] or
bm[t = times, var], or for an unknown var, instead of returning NULL
or silently ignoring an index.bm[c(2, 3, 1), "W_t"])
now returns values in the requested order.delete_skeleton(type = "label") now deletes only user labels; previously
it also deleted the layers in the interval.refine() now checks its arguments. Previously a mult of zero or less made
it loop forever, and invalid s or t silently did nothing.create_bm() now rejects non-numeric and non-finite t, W_t and cov
with a clear error, rather than failing later.q) from
simulation or from named times passed to add_points(). Names on times
are still used as labels.layers(type = "localised") no longer fails with "missing value where
TRUE/FALSE needed".BrownianMotion object with exactly one layer now works both at
and off skeleton points.sim.condbessel_() is now sim_condbessel_(), and update.lyr_(),
which R treated as a stats::update() method, is now update_lyr_()).add.labels() and delete.labels() for managing user labels on existing
path times; delete.skeleton() gains the ability to delete labels.create.bm(dim = , cov = ), with access to the transformed path through $Z.layers() (Bessel, intersection and
localised layers).add.points(), add.segment() and concat.bm().delete.skeleton() for removing path observations and layers.[ indexing of Brownian motion objects by time, index, time range and label.sf suggested for
union layer plots).