NEWS.md
"timechange" funbyApproach utilising timechange::time_floor() as the main function for transforming timestamps. This new approach generally is way faster than the existing ones for both families of TALFs and all time zones, hence it is the new package’s default.byY_____() and byYm____() TALFs of the "base" funbyApproach.print() method does not print its values’ key anymore, as it used to be data.table’s default before v1.15.0.aggregate() method, which could lead to incorrect results under the following conditions:
DTSg object is not UTC or equivalent.ignoreDST argument is TRUE..n column now contains 0 instead of NA in case all values of a certain temporal aggregation level are missing.na.rm argument.data.table’s GForce optimisation when the fun argument of the aggregate() method is provided with a character vector specifying summary functions. Please note that the column order of the resulting DTSg object is now different due to this fix.set() method was too anxious about removing all value columns.byFasttime*() TALFs. Please use the "fasttime" or even better the new "timechange" funbyApproach from now on.magrittr from the suggested packages list (vignettes now use R’s native pipe operator)."x,y,z", or the start and end column separated by a colon, for example, "x:z".interpolateLinear() function causing partial last observation carried forward behaviour when its roll argument was specified smaller than the size of the gap to be interpolated.interpolateLinear() function causing partial interpolation in certain cases when its roll argument was specified smaller than the size of the gap to be interpolated.funbyApproach argument to the new(), aggregate(), colapply() and subset() methods: allows for specifying the flavour of the applied temporal aggregation level functions (either "base" utilising as.POSIXct() or "fasttime" utilising fasttime::fastPOSIXct() or "RcppCCTZ" utilising RcppCCTZ::parseDatetime() as the main function for transforming timestamps). Custom approaches for user defined TALFs are also possible. Please note that the byFasttime* versions of the TALFs are now deprecated. Use this argument from now on instead.funbyApproach also to the list of helper data (funbyHelpers argument) passed on to TALFs.funbyApproach field reflecting the individual funbyApproach of a DTSg object (can also be actively set in order to change the utilised approach).names() method exclusive only to the R6 interface acting as an alias for the cols() method.mode and typeof arguments to the cols() method: allows for getting column names with a certain mode() and/or typeof(). This can be especially handy when making use of the units package.DTSgFast, DTSgFunbyApproach and DTSgNA.status options providing default values for the fast, funbyApproach and na.status arguments of the new() method.DTSgDeprecatedWarnings option: allows for disabling warnings from deprecated features.by______() and byFasttime______() TALFs."all" with the class argument of the cols() method in order to get all column names. Use the default value NULL instead for this. "all" is treated as a filter for classes of type all from now on.setCols() method showing how to set measurement units with the help of the units package.colapply() method showing how to calculate running correlations with the help of the runner package.".numerary" available to the class argument of the cols() method: allows for querying the names of integer and numeric columns in one go.raggregate(), rbind(), set() and setnames() methods exclusive only to the R6 interface acting as aliases for the rowaggregate(), rowbind(), setCols() and setColNames() methods.print() method now truncates the number of printed rows of the values more aggressively.pkgdown website.rowaggregate() method: allows for applying summary functions row-wise to a DTSg object.rowbind() method: allows for combining the rows of a DTSg object.setColNames() method: allows for renaming columns of a DTSg object.multiplier and funbyHelpers arguments to the aggregate(), colapply() and subset() methods: allows for adjusting the temporal aggregation level of TALFs and for passing on user defined helper data to TALFs.multiplier also to the list of helper data passed on to TALFs.helpers argument to the colapply() and rollapply() methods: controls if helper data is passed on to an applied function (makes occasionally needed anonymous function wrappers obsolete, e.g. x$colapply(fun = function(x, ...) {cumsum(x)}, funby = byYm____) can now be written as x$colapply(fun = cumsum, helpers = FALSE, funby = byYm____)).fasttime package now also support time zones equivalent to UTC (execute grep("^(Etc/)?(UCT|UTC)$|^(Etc/)?GMT(\\+|-)?0?$", OlsonNames(), ignore.case = TRUE, value = TRUE) for a full list of supported time zones).DTSg object with only one timestamp did not set the name of its .dateTime column as expected.subset() method: allows for filtering rows and/or selecting columns of a DTSg object.setCols() method: allows for setting the values of columns of, adding columns to and/or removing columns from a DTSg object.[ extract operator: acts as a shortcut for the getCol() method.colapply() method showing how to calculate moving averages with the help of the runner package instead of the rollapply() method.aggregate() method can benefit from data.table’s GForce optimisation now when its fun argument is provided with a character vector specifying summary functions.nas() method.funby argument of the colapply() method are not forced to return a POSIXct timestamp any longer. They are, however, forced to return an atomic mode (the same goes for the subset() method).getCol() method now is capable of also querying the .dateTime column.R6Method argument of the S3WrapperGenerator() function now also takes a public method of an R6ClassGenerator as a function and not only as an expression.merge() method despite an "explicit" na.status in some cases.getCol() method tried to query all numeric columns instead of only the first one by default.getCol() method: allows for querying the values of a single column of a DTSg object.funby and ignoreDST arguments to the colapply() method: allows for applying functions like cumsum() to a certain temporal level.na.status argument to the new() and alter() methods: allows for making missing values either "explicit" (default) or "implicit" or leaving them alone via "undecided".na.status field reflecting the status of missing values (can also be actively set).na.status also to the list of helper data passed on to temporal aggregation level functions.funby argument of the aggregate() method now also accepts a named list of functions: allows for calculating several summary statistics at once.periodicity field can now be actively set in order to change the periodicity of the time series.timezone field can now be actively set in order to convert the time zone of the series.DTSg object with only one timestamp accepted a missing value in its .dateTime column.rollback() function in upcoming R 4.1.0.memoryOverCPU argument to the rollapply() method: allows for preferring CPU over memory usage which makes the method more flexible in terms of resource consumption.DTSgClone in e.g. .RProfile are now respected and not overwritten by TRUE when the package is loaded.POSIXct .dateTime columns to POSIXct upon object creation in case they contained at least one missing value.tinytest instead of the testthat package.resultCols and suffix arguments to the colapply() and rollapply() methods: allows for adding return values of applied functions as new columns instead of replacing existing ones.class argument of the cols() method now accepts a character vector of class names.class argument of the cols() method from "all" to NULL, however, for backward compatibility "all" can still be used for the same result, but will eventually be treated as a filter for classes of type all.pattern and ... arguments to the cols() method: allows for searching column names.checkmate package instead of the assertive.base, assertive.numbers, assertive.sets and assertive.types packages.dygraphs package in at least some releases of R ≤ 3.5.3.swallow argument to the new() method: allows for a more resource efficient object creation.drop argument to the values() method: allows for a resource efficient destruction of a DTSg object while preserving its values.class argument to the values() method: can be used to return the values of a DTSg object as a data.frame instead of a data.table.alter() method and linked with it the new() and other methods are now way more resource efficient in some cases.timestamps field providing the total number of timestamps.print() method now omits empty metadata fields.covr package.xts from the suggested packages list (already comes along with the dygraphs package).