Skip to contents

Draw what a solve sequence covered

Source: R/plot_sequence.R
autoplot.sequence.Rd

autoplot() over sequence_coverage(): the units of a sequence drawn on the scale they were drawn from, so tiling, overlap and gaps are shapes to be seen rather than properties to be inferred from a table of runs.

Usage

# S3 method for class 'by_sample'
autoplot(
  object,
  type = c("coverage", "wall", "histogram"),
  year = NULL,
  var = "objective",
  by = NULL,
  bins = NULL,
  ...
)

# S3 method for class 'by_region'
autoplot(object, type = c("coverage"), geoscale = NULL, level = NULL, ...)

# S3 method for class 'recursive'
autoplot(object, type = c("coverage"), ...)

# S3 method for class 'guided'
autoplot(object, type = c("coverage"), ...)

Arguments

object

a by_sample, by_region, recursive or guided sequence.

type

which drawing to make; see Details.

year

the model year whose weekdays type = "wall" is arranged by. Without one the days are laid out in plain sequence.

var

for type = "histogram", what to draw the spread of: "objective", or the name of a solved variable.

by

for type = "histogram", the columns to group a variable on, one histogram each. Defaults to every key column, since any column left out would have to be aggregated away.

bins

for type = "histogram", the number of bins. Defaults to the square root of the number of samples.

...

for type = "histogram", passed to getData() to narrow what is drawn (year = , region = , ...); otherwise unused.

geoscale

a geoscales::GeoScale to draw on, or NULL for the sequence model's own.

level

the geoframe to draw. Defaults to the finest frame that contains every covered region.

Value

A ggplot object.

Details

One method per driver, because each samples a different scale.

  • by_sample draws the calendar. type = "coverage" is a tile grid over the calendar's two finest timeframes and works for any calendar; type = "wall" is the month-faceted wall calendar, which needs a day-resolution timeframe (YDAY or MDAY) and rolls each unit up to whole days. type = "histogram" leaves coverage behind and draws the spread of a result across the samples: bars weighted by each sample's year_fraction, a dashed rule at the weighted mean sample_summary() reports, and the free pass overlaid on the floored one when ncap_floor > 0. There is no histogram for by_region, whose units are shards of one system rather than replicates of it.

  • by_region draws the geoscale as a choropleth, falling back to a tile grid when the geoscale carries no geometry or sf is not installed.

  • recursive and guided draw the horizon as a Gantt chart, one row per step. For recursive the look-ahead milestones are filled apart from the decided ones: the overlap between steps is the look-ahead, and seeing it as such is the point.

Elements no unit covered keep the grey NA fill, and an element two units claimed is filled as "(shared)" rather than resolved silently. The subtitle counts both, plus the within-unit multiplicity a bootstrap draw carries.

See also

sequence_coverage(), solve_by_sample(), solve_by_region(), solve_recursive()

Examples

if (FALSE) { # \dontrun{
bs <- solve_by_sample(mod, samples = 12)
autoplot(bs)                        # which timeslices each sample drew
autoplot(bs, "wall", year = 2020)   # the same, as a wall calendar
autoplot(bs, "histogram")           # the objective's spread
autoplot(bs, "histogram", var = "vTechCap", by = "tech", year = 2050)
autoplot(solve_by_region(mod, group = list(c("R1", "R2"), "R3")))
autoplot(solve_recursive(mod, overlap = 1))
} # }