Changelog
timeSeries 4032.109
CRAN release: 2024-01-14
deprecated function
returnSeries
is now defunct, usereturns
instead.a number of generic functions from base R now get only S3 methods for ‘timeSeries’ objects. Previously they were turned into S4 generics with S4 methods.
streamlined timeSeries methods for a number of functions. Left only S3 methods or only S4 methods were suitable.
consolidated the NAMESPACE.
timeSeries 4032.108
CRAN release: 2023-12-19
- fixed ‘Lost braces; missing escapes or markup?’ NOTE from CRAN.
timeSeries 4031.107
CRAN release: 2023-08-26
refactored the ‘timeSeries’ methods for
head
andtail
.fixed a bug in the ‘timeSeries’ method for
stats::na.contiguous
, which caused the wrong stretch to be returned in the case of tied longest stretches one of whom starts at the beginning of the series. Similar bug was present instats::na.contiguous.default
, see my bug report to R-devel from 2023-06-02 and the discussion there (https://stat.ethz.ch/pipermail/r-devel/2023-June/082642.html)removed deprecated functions
spreadSeries
,midquoteSeries
, anddurationSeries
. Usespreads
,midquotes
, anddurations
, respectively.removed deprecated function
colStdevs
, usecolSds()
instead.removed deprecated function
.description
, usedescription()
instead.removed deprecated ‘timeSeries’ method for function
cut()
, usewindow()
instead. The method was not compatible with the generic functioncut()
. Now applyingcut(x)
on a ‘timeSeries’ objectx
will work on the underlying time series data.-
replaced the S4 methods for
zoo::coredata
andzoo::'coredata<-'
. The ones forzoo::'coredata<-'
were not working at all, sincezoo::'coredata<-'
is an S3 generic and the methods dispatch on two arguments. It is also a mistery why the methods for the unexported S4 generics in ‘timeSeries’ were associated with the corresponding ‘zoo’ generics.If
zoo
is not attached, the calls need to be prefixed withzoo::
or, alternatively, since the new methods are exported, they can be called directly ascoredata.timeSeries()
and `coredata.'timeSeries<-'() <- value
. added a default method for
time<-
to improve its interaction with ‘zoo’.added ‘zoo’ to ‘Suggests:’.
removed the deprecated
dummySeries
, usedummyMonthlySeries
instead.added argument
FUN
to thetimeSeries
method forna.omit
to allow it to compute replacement values using functions, such asmean
,median
, or user defined.formally deprecated
removeNA
,interpNA
, andsubstituteNA
. These had been informally deprecated in the documentation for a long time.the help page for
orderStatistics
erroneously claimed that the input should be an univariatetimeSeries
object, while it is explicitly written to cover the multivariate case.moved package ‘methods’ back to ‘Depends’ to avoid subtle problems when ‘methods’ is loaded but not attached. For example, it seems that ‘Math’ methods for ‘structure’ are not seen for
cummin
and othercumXXX
functions, when called on time series objects (the other math functions work ok).cumsum
,cumprod
,cummin
, andcummax
now work on the columns of the ‘timeSeries’ object and keep its class and other attributes. This is a breaking change since previously the return value was numeric vector, the result of applying the base R functions to the data part of the object. This was not particularly useful, especilly for multivariate time series. With this change all functions from the S4Math
group return ‘timeSeries’ when their argument is ‘timeSeries’ object.stopped exporting some internal functions that were accidentally used by other packages (after those packages were updated on CRAN).
Numerous improvements to the documentation and further changes in the code.
timeSeries 4030.106
CRAN release: 2023-05-25
- removed UTF8 characters from NAMESPACE (fixes CRAN warning to that effect).
timeSeries 4021.105
CRAN release: 2022-10-15
updated and significantly improved the documentation.
class
timeSeries
now has a dedicated summary method. Previously it was falling back to the method for matrices.colCumsums
,colCummaxs
,colCummins
, andcolCumprods
no longer throw error fortimeSeries
objects when called withna.rm = TRUE
. Fixes bug #2121 reported by Shane Haas.corrected USDCHF dataset. The year information was wrong (the data started from year 8295). The bug had been introduced in version 2100.84 when the dataset file was converted from a
usdchf.csv
toUSDCHF.rda
.USDCHF@documentation
contains a short note about this change. Also changed the FinCenter to Zurich (neither the documentation nor the csv file contain FinCenter information).the original source file
msft.dat.csv
of theMSFT
data is included now asinst/extdata/msft.csv
(note the different name). The file had been removed in v2100.84. Note that there is a file `msft.dat.csv
intest/
but it is a modified and abbreviated version of the original file.dummySeries
has been renamed to the more expressivedummyMonthlySeries
. The old name is still available but is deprecated.The functions
returnSeries
andgetReturns
are no longer exported and will be removed in the near future. They are synonyms for the functionreturns
and their use was discouraged for many years. Just usereturns
.function
cut
is now formally deprecated. Usewindow
instead.deprecated function
seriesData
is now defunct. Useas.matrix()
instead.deprecated function
seriesPositions
is now defunct. Usetime()
instead.deprecated function
newPositions<-
is now defunct. Usetime<-
instead.deprecated function
colAvgs
is now defunct. UsecolMeans()
instead.deprecated function
colStdevs
is now defunct. UsecolSds()
instead.
Technical changes
stopped exporting (almost) all functions whose names start with a ‘.’. Historically, the package was exporting all functions, including those start with a ‘.’. This should be of no concern for users since these functions were not documented but the developers of some Rmetrics packages where using such functions.
the additional arguments of the S3
timeSeries
method fordiff()
are now in its signature, which previously wasdiff(x, ...)
. An intermediate function,.diff.timeSeries
, was eliminated in the process.the bodies of the methods of
series<-()
andcoredata<-
for signature"matrix"
of value were identical. Now the body is a separate, unexported function, which is used as the definition of both of these methods.eliminated
.merge.timeSeries
and other redundancy in the implementation of thec("timeSeries", "timeSeries")
method.eliminated
.rev.timeSeries
in the definition of therev
method.eliminated
.scale.timeSeries
in the definition of thescale
timeSeries method.same as above for
.sort.timeSeries
.eliminated
.start.timeSeries
and redundancy in the implementation of thetimeSeries
method.eliminated
.end.timeSeries
and redundancy in the implementation of thetimeSeries
method.the function
.applySeries
is now defunct. It was obsoleted long time ago and was exported for historical reasons only. UseapplySeries()
instead.