ggplot to loon transformation
the ggplot and its states
Begin with a ggplot defined using the interactive
grammar extension:
ggp <- ggplot(data,
mapping = aes(x = A, y = B)) +
ggtitle("Some title") +
geom_point(color = "grey", size = 5) +
linking(linkingGroup = "my plots")
plot(ggp)
Information on the ggplot corresponding to
loon states is had using ggplot_build()
# get the ggplot data corresponding to loon "states"
ggp_states <- ggplot_build(ggp)$data[[1]]
ggp_states
#> x y PANEL group shape colour size fill alpha stroke
#> 1 19 68 1 -1 19 grey 5 NA NA 0.5
#> 2 19 63 1 -1 19 grey 5 NA NA 0.5
#> 3 25 63 1 -1 19 grey 5 NA NA 0.5
#> 4 62 13 1 -1 19 grey 5 NA NA 0.5
#> 5 34 55 1 -1 19 grey 5 NA NA 0.5
#> 6 58 78 1 -1 19 grey 5 NA NA 0.5
#> 7 62 14 1 -1 19 grey 5 NA NA 0.5
#> 8 40 14 1 -1 19 grey 5 NA NA 0.5
#> 9 24 NA 1 -1 19 grey 5 NA NA 0.5
#> 10 60 28 1 -1 19 grey 5 NA NA 0.5
#> 11 70 NA 1 -1 19 grey 5 NA NA 0.5
#> 12 40 55 1 -1 19 grey 5 NA NA 0.5
#> 13 40 57 1 -1 19 grey 5 NA NA 0.5
#> 14 34 40 1 -1 19 grey 5 NA NA 0.5
#> 15 26 78 1 -1 19 grey 5 NA NA 0.5Note that some of the x and y values are
NA, representing missing values.
the loon plot and its states
Now construct the loon plot from this
ggplot:
The loon plot looks slightly different:
- title in a different position,
- the relative sizes of the margins and drawing areas,
- axis tic marks and labels,
- font sizes and styles, and
- points are a slightly darker grey in the
loonplot.
Plots in loon have more constrained layouts than those
in ggplot2, since focus is primarily on interactive as
opposed to publication quality graphics.
Some differences (e.g., colour) are also because loon is
based on Tcl for drawing primitives.
Plot states in loon are accessible using []
and the name of the state. Values are assigned in the same way. Names of
interactively changeable states in loon are
names(lp)
#> [1] "glyph" "itemLabel" "showItemLabels" "linkingGroup"
#> [5] "linkingKey" "zoomX" "zoomY" "panX"
#> [9] "panY" "deltaX" "deltaY" "xlabel"
#> [13] "ylabel" "title" "showLabels" "showScales"
#> [17] "swapAxes" "showGuides" "background" "foreground"
#> [21] "guidesBackground" "guidelines" "minimumMargins" "labelMargins"
#> [25] "scalesMargins" "x" "y" "xTemp"
#> [29] "yTemp" "color" "selected" "active"
#> [33] "size" "tag" "useLoonInspector" "selectBy"
#> [37] "selectionLogic"
# and accessed with [] as in
lp["title"]
#> [1] "Some title"(See also ?l_info_states.)
Some, like x, y, and color,
are n-dimensional states whose values correspond to those on
individual observations. These may be changed using the
[] <- notation as well.
Because loon plots exclude observations missing values
in any its n-dimensional states, the loon plot will have
fewer observations than the corresponding ggplot, even
though the same number of points may be plotted. The actual number of
observations in the loon plot is
which is less than the number of observations recorded in the
ggplot structure (namely nrow(ggp_states) =
15).
changing from ggplot to loon
Differences between ggplot structure and
loon plot structure (and how these affect the
visualizations) is worth exploring in a little more detail. A better
understanding will be helpful when transforming back and forth between
the two.
shape to glyph
In R the point symbol plotted is associated with a
numerical code pch. In loon the point symbol
is a glyph, identified by a string.
ggp_states$shape
#> [1] 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19
lp["glyph"]
#> [1] "circle" "circle" "circle" "circle" "circle" "circle" "circle" "circle"
#> [9] "circle" "circle" "circle" "circle" "circle"As with other plot “states”, there is not always a glyph
in loon that matches a pch in R.
Values of pch with no counterpart in loon will
map to the default glyph in loon.
point size
Size is another plot state that differs between ggplot
and loon.
ggp_states$size
#> [1] 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5
lp["size"]
#> [1] 26.78116 26.78116 26.78116 26.78116 26.78116 26.78116 26.78116 26.78116
#> [9] 26.78116 26.78116 26.78116 26.78116 26.78116The point size in loon identifies the
approximate area of the point symbol, so when
transforming from ggplot to loon, the
size parameter of the ggplot is transformed to
an integer value in loon that tries to match the area of
the point in ggplot.
colours
Comparing the plots, the points are slightly different
"grey" in one plot than in the other.
This is because the two plots have different values stored as their colour state:
ggp_states$colour
#> [1] "grey" "grey" "grey" "grey" "grey" "grey" "grey" "grey" "grey" "grey"
#> [11] "grey" "grey" "grey" "grey" "grey"
lp["color"]
#> [1] "#808080808080" "#808080808080" "#808080808080" "#808080808080"
#> [5] "#808080808080" "#808080808080" "#808080808080" "#808080808080"
#> [9] "#808080808080" "#808080808080" "#808080808080" "#808080808080"
#> [13] "#808080808080"First, there is a slight difference between what R
regards as "grey" and what Tcl (used by
loon) regards as "grey". Several other
named colours in R also differ slightly
from those in Tcl and hence in loon.
R colours whose name is the same as that
in Tcl but whose colour is different are
only the following:
(See ?tkcolors in loon for more
information.)
Second, loon plot colour values are strings of
twelve hexadecimal digits, as in tk (e.g.,
see ?tkcolors); in contrast, the ggplot are
either strings corresponding to the named R colours (see
?grDevices::colors in R) or to
six hexadecimal digits (two for each of the red, green,
and blue components).
Twelve hex digit colours are turned into six hex digits using the
loon function hex12to6(). To convert the
R named colours to six hex digits, the following function
can be used.
tohex <- function(x) {
sapply(x, function(xi) {
crgb <- as.vector(col2rgb(xi))
rgb(crgb[1], crgb[2], crgb[3], maxColorValue = 255)
})}So, comparing the two "grey" colours, the
ggplot has the R hex colour
tohex("grey") = #BEBEBE and the loon plot has
Tcl hex colour (converted to hex 6)
hex12tohex6(lp["color"][1]) = #808080, which are clearly
different RGB values.
effect of missing data
The loon plot does not include the missing data
nrow(ggp_states) == lp["n"]
#> [1] FALSE
# Compare
ggp_states$y
#> [1] 68 63 63 13 55 78 14 14 NA 28 NA 55 57 40 78
lp["y"]
#> [1] 68 63 63 13 55 78 14 14 28 55 57 40 78Note that, absent the missing NAs, the
order of the points is identical.
on linking keys
Default linking keys in loon are from "0"
to "n-1" where "n" is the number of rows in
the original data set.
Here nrow(data) = 15. Note that this "n" is
not the number of points actually plotted because loon
drops the missing data, unlike ggplot.
The linking keys reflect this missingness:
which is missing "8" and "10". Note also
that lp["n"] = 13 is the number of points plotted in the
loon plot.
Some care needs to be taken when dealing with linking, especially when there is missing data.
for more information
See the vignette Linking
and the loon vignette Logical
Queries for more information.
The left plot,