VPdtw() now always returns the lowest-cost alignment. Previously a tie between
two steps could make it take a costlier path, so alignments and costs can change.
VPdtw() no longer misaligns penalty when reference contains NA.
Previously, each point after an NA got the penalty meant for an earlier point,
so results change if penalty varies.
dilation() no longer over-dilates the final span - 1 points of a signal.
Alignments whose penalty was built with dilation() can differ from those in
2.2.1.
VPdtw() now uses a single penalty value, such as the default penalty = 0,
along the whole alignment when query is a matrix. Previously, points after the
first got arbitrary penalties, giving wrong alignments.
dilation() no longer crashes R when span is much wider than the signal, and
now errors on a span below 1, a non-numeric y, or a y with no non-missing
values.
VPdtw() now errors on a maxshift below 1 and on a penalty shorter than
reference. A negative maxshift previously crashed R.
VPdtw() now warns when penalty contains negative values, which would reward shifting.
plot.VPdtw() no longer clips its legends in the three-panel layout when the plot
window is small.
plot.VPdtw and print.VPdtw.plot.VPdtw with
type = "Chromatograms".plot.VPdtw function so they no longer change user's par
settings.NEWS.md file to track changes to the package.