plot.xts axis only shows two labels

[eddelbuettel]

Description

The x-axis annotation is somewhat overly sparse on irregular data

Expected behavior

I would expect the usual four to five ticks to be shown

Minimal, reproducible example

The image comes from running the example which requires the package. The same effect appears
just with the code not commented out here

## library(RcppUTS)
## example(EMAnext)
##
## or directly:
suppressMessages(library(xts))
times <- ISOdatetime(2010, 1, 2, 8, 30, 0) + c(0, 1.0, 1.2, 2.3, 2.9, 5.0)
values <- seq(0, 10, by=2)
plot(xts(values, order.by=times), type="b",
         main="Series and last/next/linear EMAs")
#lines(xts(EMAlast(times,values, 1),  values, order.by=times),
#          type="b", col="lightblue")
#    lines(xts(EMAnext(times,values, 1),  values, order.by=times),
#          type="b", col="darkblue")
#   lines(xts(EMAlinear(times,values, 1),  values, order.by=times),
#          type="b", col="mediumblue")
    addLegend("topleft", legend.names=c("series", "EMAlast", "EMAnext", "EMAlinear"),
              lty=rep(1,4), lwd=rep(1,4),
              col=c("black", "lightblue", "darkblue", "mediumblue"))

will re-create the same plot shown on the repository README.md and here:

Session Info

Current CRAN versions of everything.

[eddelbuettel]

BTW I tried to research this in old issue tickets and did not see it posted. My bad if I missed one.

/cc @rossb34

[joshuaulrich]

Thanks for the report. This has bothered me too, but never enough to take a look. Seems like the solution could be as simple as setting major.ticks = "auto" and grid.ticks.on = "auto" as the plot.xts() defaults. It works for your example.

[eddelbuettel]

That helps indeed. The documentation on those options and these values is a bit ... sparse.

[rossb34]

I agree with @joshuaulrich that setting the default arguments to "auto" for the grid and axis ticks is a good solution. The default is "months" which is only reasonable for several months of data on a daily or lower frequency.

plot.xts uses xts:::axTicksByTime2 which is a port from what is used in quantmod::chart_Series, also mentioned in #74.

[rossb34]

@eddelbuettel thanks for the note on documentation. I'll add language to improve the documentation.

[eddelbuettel]

I am also looking at what rtsplot::rtsplot() does -- and it is pretty compelling. Just extended some intra-daily data from one to several days and it put weekdays on the axis.

[joshuaulrich]

rtsplot::rtsplot() uses graphics::axis.POSIXct(), which plot.xts() cannot use because of its different geometry. That said, we may be able to use a similar approach.

One possible problem: xts objects are often irregularly spaced, which means there could easily be no observations where you would want to place a tick mark based on the logic in graphics::axis.POSIXct(). This is why axTicksByTime() uses endpoints() to find tick mark locations.

[eddelbuettel]

Casual usage of both over the last couple of days with generally irregular data has rtsplot() as a winner.

I am sure there are counter-examples. I have not yet hit any. I'll follow up if I do.