FEATURES ******** o rbind.zoo() currently ignores the column names (if any), should behave like rbind.data.frame() in this case. o design: re-think shared design of parameter and panel functions for plotting. o bugs: rollmax, rollmedian give Error for k=1. o zoo methods for dbReadTable and dbWriteTable to facilitate using zoo objects with DBI o register S3 methods and use namespaces so that this works: f <- local(function(x) zoo::index(x), baseenv()) o provide a [<-.zoo method that is fully consistent with [.zoo so that the following works: a <- zoo(matrix(1:10,5,2), 2001:2005) a[I(2003), 2] <- NA (Note: as in window<-.zoo) o preserve attributes in [.zoo and possibly certain other zoo methods o allow plot.zoo to accept a list of zoo objects -- this would be particularly useful in the case of line plots of multiple series with different time bases since NAs are introduced by merging and they can cause the drawn lines to be undesirably broken. (Current workaround is to use na.approx.) o more graphical facilities, e.g., a pairs function, make plot.zoo(x, y) work with multivariate series (plotting every combination of plot(x[,i], y[,j])) o incorporate runmean in package caMassClass o allow log to be specified per plot in plot.zoo o performance enhancement when using arithmetic on zoo objects. Check if they have same index first. o speedup for indexing strictly regular series. Test with is.regular first. set.seed(1) x <- ts(rnorm(5000)) xz <- as.zoo(x) system.time({ ################################ for(i in 1:5000) if (is.regular(x)) x[(i - tsp(x)[1])*frequency(x) + 1] <- 0 else window(xz, start = i, end = i) <- 0 }, gc = TRUE) system.time( for (i in 1:5000) xz[I(i)] <- 0, gc = TRUE ) system.time( for (i in 1:5000) window(xz, start=i, end=i) <- 0, gc = TRUE o zoo() - support zoo data.frame objects (and zoo list objects) - fully support zoo factor objects [limited support already available] - idea: new class "Zoo" that is not a matrix plus attribute but a list with slots "coredata", "index" and "frequency". o merge() - names not processed correctly when retclass = "list" o misc functions - interface to additional statistical routines such as the ones referenced in: http://CRAN.R-project.org/doc/contrib/Ricci-refcard-ts.pdf [Z: with as.ts.zoo() most of the functions work directly by coercing the first argument to "ts"] o QA: regression test suite o make.par.list. names, other than correspondence names, are not allowed on x. Perhaps this should be allowed if m = 1 or perhaps it should be allowed if x is not a list and correspondence names only allowed for lists. o rollapply.default o as of lattice 0.12-14 lattice:::updateList could be replaced with modifyList o read.zoo. The following situations have all appeared on r-help: - the index is of the form yyyy-mm-dd hh:mm:ss at the beginning of each line but sep is whitespace. To handle this read.zoo would have to figure out special cases in which there is whitespace in the format and use two columsn rather than one. I think it would be good enough if it only handled a single whitespace. - header does not have a component corresponding to the index. It would be nice if read.zoo could automatically figure out that situation and handle it. - the date format is yyyy.mm . In that case it would be convenient if it were possible to specify: read.zoo(..., FUN = as.yearmon, format = "%Y.%m") and, in general, if format were passed to FUN if both are specified. o identify.zoo o allow index.column in read.zoo to be a vector in which case it pastes them together so that 9999-99-99 99:99:99 99 99 99 can be read directly as dates and times. o add class= arg to as.zoo.ts o cut/seq/floor/ceiling/round/trunc on yearmon/yearqtr objects INTERFACES ********** o provide fCalendar interface routine to the Rmetrics projects DOCUMENTATION & COMMUNICATION ***************************** o R news article based on the vignette o some data to include with zoo o more examples for the help files o as of R 2.6.0 the example in aggregate.zoo.Rd that uses duplicates(..., fromLast=) could be taken out of the \dontrun{}