NEWS.md
The default breaks of mixtime scales now follow the calendar structure of the time being plotted, stepping by how its chronon nests within coarser granules (e.g. every 3 or 6 months within a year, every 7 days within a week) and aligning time points to them (each January, each Monday), rather than round numbers of the chronon (#37).
autoplot() of ACF/PACF/CCF results labels lags with plain numbers, as their unit is already in the axis title (#37).
breaks coarser than the data (e.g. yearmonth() breaks on a daily axis) are now placed at the start of each time point, for compatibility with mixtime refusing temporally indeterminate conversions of discrete time.geom_candlestick()/stat_ohlc(): summarises the values within each time granule by their first (open), largest (high), smallest (low), and last (close) value, drawn as a candlestick chart. Rising and falling candles are coloured with scale_colour_candlestick() and scale_fill_candlestick().
coord_calendar() gains cell_ratio, which fixes the aspect ratio of a calendar cell as coord_fixed() fixes that of the data units of the axes. cell_ratio = 1 gives the square cells of a printed calendar (#33).
The plot helpers (gg_season(), gg_subseries(), gg_lag(), gg_tsdisplay(), and autoplot() of ACF/PACF/CCF results) are now all built from mixtime. Legacy Date, POSIXct, and tsibble yearweek()/yearmonth()/yearquarter() indices are converted to mixtime for plotting, so existing code continues to work. Labels for cyclical displays of time (e.g. time-of-day) have been improved as a result.
gg_season() is now drawn with coord_loop(), keeping time continuous rather than offsetting it into a single season. The lines connecting consecutive seasons (e.g. December to January) are now drawn, split across the left and right margins of the plot.
The period of gg_season(), gg_subseries(), and gg_lag() can now be a mixtime duration (e.g. mixtime::weeks(2L)) or a string of a whole number of a single unit (e.g. "year" or "2 weeks"), and is no longer converted to a lubridate period.
gg_subseries() now facets by the smallest seasonal period of the data by default (e.g. a week for daily data, a day for hourly data).
gg_lag() now only colours by season for periods with a fixed number of observations, and gives an informative error if the requested period doesn’t (e.g. days in a month).
autoplot() of ACF/PACF/CCF results now labels lags as durations of time (e.g. “12 months”) rather than a count of observations.
Looping coord_radial() with coord_loop() no longer labels the end of the loop with its start (e.g. “Jan/Jan”).
gg_season() and related plots now give a clear error for zero-row data instead of failing with “object ‘found_format’ not found”.
gg_season() now includes lines connecting observations across seasonalities (e.g. from December of one year to January of the next).
ggtime no longer imports lubridate.
This release introduces the grammar of temporal graphics: a set of time-aware geoms, scales, and coordinate systems that extend ggplot2 to work with mixtime time vectors. These grammar elements sit alongside the plot helper functions from previous releases, and are what the helpers are now built from.
geom_time_line(): a time-aware extension of geom_line() that keeps a line’s slope an accurate reflection of the rate of change, even across timezone changes, gaps, and duplicated time points.scale_x_mixtime()/scale_y_mixtime(): the position scale behind every mixtime time axis, applied automatically whenever a mixtime vector is mapped to a plot. It maps time points of different granularities onto one shared axis, and takes calendrical durations for breaks (time_breaks) and calendar-aware format strings for labels (time_labels).scale_colour_mixtime()/scale_fill_mixtime()/scale_alpha_mixtime()/ scale_size_mixtime()/scale_linewidth_mixtime(): mixtime-aware equivalents for other aesthetics.coord_loop(): loops the time axis around a calendrical period, so a continuous time axis can be compared cyclically instead of being discretised into a seasonal factor.coord_calendar(): arranges time into a calendar-like grid of rows and columns, useful for visualising events over short intervals within a long time span, such as holidays.aes_nudge(): specifying alignment of discrete time poisitions on continuous time plot axis with sensible defaults.transform_warp(): a scales transform for warping time.This release completes the migration of graphics functions from {feasts} and {fabletools}, including a deprecation process for a gradual migration to {ggtime}. If you previously used time-series graphics from {feasts} or {fable}, you should now include library(ggtime) in your script to avoid deprecation notices. Existing code will continue to work with a deprecation warning.
The following plot helper functions from {fabletools} are included in this release:
autoplot(<fbl_ts>): Forecast plots to show forecast intervals with historical data.autolayer(<fbl_ts>): Forecast layers for fable objects.autoplot(<dcmp_ts>): Decomposition plots to show components of a dable object.fortify(<fbl_ts>): Fortify method for converting fable objects into basic data frames.Initial release including all plot helper functions from across the tidy time series analysis packages. The following plot helper functions are included in this release:
autoplot(<tbl_ts>): Time plots to show overall patterns in tsibble objects.autolayer(<tbl_ts>): Time plot layers for tsibble objects.gg_season(): Seasonal plots to show the shape of seasonal patterns.gg_subseries(): Seasonal sub-series plots to show seasonal changes over time.gg_lag(): Lag plots to show relationships between now and the past.gg_irf(): Impulse response function plots to be used with IRF() results.gg_arma(): Plot the characteristic ARMA roots.gg_tsdisplay(): An ensemble graphic useful in exploring time series data.gg_tsresiduals(): An ensemble graphic useful in diagnosing model residuals.