davidhodge931/ggblanket

Simplify ggplot2 visualisation

ggblanket

Overview

ggblanket is a package of ggplot2 wrapper functions.

The primary objective is to simplify ggplot2 visualisation.

Secondary objectives relate to:

• Design: produce well-designed visualisation
• Alignment: use conventions aligned with ggplot2
• Scope: cover much of what ggplot2 does.

Computational speed has been traded-off to achieve these objectives.

Installation

install.packages("ggblanket")

Example

library(ggblanket)
library(palmerpenguins)

set_blanket()

penguins |>
  gg_histogram(
    x = flipper_length_mm,
    col = species,
  )

Get started

Click here to start learning how ggblanket works.

Thank you

Thanks to all authors of ggplot2, tidyverse, and the wider R ecosystem.

This R package is dedicated to my Dad (Peter Hodge, 1953–2023).

ggblanket's Issues

Add aesthetics for xmin/xmax/ymin/ymax xend/yend

Breaks should accept a function

x_zero & y_zero defaults for blank

Where there is continuous colour, zero-ing looks bad

Question:

• to make people convert limits to data min/max where it looks bad
• to change it in particular geoms where it makes sense to??

Change x_zero and y_zero args to `x_limits_include` & `y_limits_include`

More flexible...
but not sure

col_method names could be abbreviated

"bin" = "b"
"quantile" = "q"
"gradient" = "g"
"category" = "c"

geom_dotplot does not work

hms class does not work (i.e. time)

No scale present for time

lakers %>%
  slice_sample(n = 100) %>% 
  tibble() %>% select(time, points) %>% 
  mutate(time = lubridate::hm(time)) %>% 
  gg_blank(time, points) +
  geom_point() +
  scale_x_time()

naming col + fill aesthetic

dye
colfill
col_fill

x/y_breaks default limits should be adjusted

Sometimes scales don't use the space well. Especially during facetting

Better...

Or just adjust x_breaks_n to make them higher??

new function: add `gg_hex` function

Update licence & authors

boxplot: not building scale with respect for outliers

Allow code to work with when either x/y is not provided

x/y_balance

Ideally it would be:

x_breaks_mid = 0
y_breaks_mid = 0

datetime class does not work

is col_method the right approach?

Or should you follow col_method = "gradientn", "manual", "binned"...

x/y_zero

Ideally it would be:

x_limits_include = c(0)
y_limits_include = c(0)

col_method revise, Maybe call it `col_chop`

3 options:

1). Specify an argument, which does it for people.

evenly ...intervals (i.e. col_chop_n = blah) #by width
quantiles ...probs (i.e. col_chop_probs = blah) #by number of elements
mean_sd ..sds (i.e. col_probs_sds = blah) #by statistics
exactly ...breaks (i.e. col_breaks = blah).
pretty

Pro's:

• can make labels default to nice stuff, and link fmt to col_labels
• people can make their own var in the data if it's not working for them

Con's lots of args

2. People add a chop function

penguins %>%
gg_blank(bill_length_mm, flipper_length_mm, body_mass_g, x_zero = F, y_zero = F,
col_labels = ~ prettyNum(.x, big.mark = "", scientific = FALSE),
pal = pal_viridis_reorder(5),
col_chop = ~ santoku::chop_evenly(.x, intervals = 5)
) +
geom_point()

3. Make people do it themselves

Make argument col_legend_place

col_legend_place = "r", "t", "b", "n", "m"

Why? Because it is a combo of position/direction/ & title position...

x_breaks make the graph for facetting look bad

Maybe should use x_limits = c(NA, NA) for facetting

x & y scales bugs out when x_breaks or y_breaks specified

library(ggplot2)
library(palmerpenguins)

ggplot(penguins) +
  geom_point(aes(x = body_mass_g, y = flipper_length_mm)) +
  scale_x_continuous(breaks = scales::breaks_width(550))

x_breaks <- scales::fullseq(range(penguins$body_mass_g, na.rm = T), size = 550)

gg_point(penguins, body_mass_g, flipper_length_mm,
         x_breaks = c( 2200, 2750, 3300, 3850, 4400, 4950, 5500, 6050, 6600))

gg_point(penguins, body_mass_g, flipper_length_mm,
         y_breaks = c(160, 190, 240))

x_breaks <- scales::fullseq(range(economics$date, na.rm = T), size = "2 years")

Package names

library(ggrescale)

What is the package doing?
1). quick aes: uses quick syntax for aesthetics
2). facet as if an aes: simplifies facetting by treating facet as if it is an aesthetic
3). combines col/fill aes simplifies colour via pal argument, and combining col/fill
4). rescales x/y limits: creates adjustable x and y scales, where limits = min/max breaks
5). reordering: updates order of horizontal geom y and col, including legends, so everything is in order
6). default format: convert labels and titles to sentence case and comma

library(ggremix)

alpha & size: align with ggplot2

histogram methods for scale don't work

This occurs because the graph is built using ggplot2 default x/y scales

In histogram, the limits affect the data because in histogram the area between the x_limits affects the size of the y values computed.

For all geom's, the build of the ggplot is used to compute the x and y numeric scales. This is because the different positions and stats result in different x and y values.

For gg_histogram, you might have to do something different...

If x is not null, calculate the x_breaks and x_scale based on x_var_vctr, and feed into the ggplot2 build

If y is not null, calculate the y_breaks and y_scale based on the y_var_vctr and feed into the ggplot2

Maybe this should be done in blank??

col: for forcats, add_col_breaks arg in

From
col_scale_manual
breaks = col_breaks
values = pal

dates not working with fixed scales

Use dots to access other args in geom_

library(tidyverse)
library(palmerpenguins)

smoother <- function(data, x, y, method = "lm", ...) {
    ggplot(data) +
    geom_smooth(aes({{x}}, {{y}}), method = method, ...)
}

smoother(penguins, flipper_length_mm, body_mass_g)
smoother(penguins, flipper_length_mm, body_mass_g, method = "loess")

consider supporting `gg_contour` & `gg_contour_filled` functions

col discrete scale

library(tidyverse)

breaks <- c("Chinstrap", "Adelie", "Gentoo", "Another one")
values <- c("green", "blue", "red", "pink")
pal_df <- tibble::tibble(breaks, values)

ggplot(penguins) +
geom_point(aes(flipper_length_mm, body_mass_g, col = species)) +
scale_color_manual(values = pal_df$values, breaks = pal_df$breaks, drop = FALSE)

Modifications to gg_blank

Titles where no x/y aes
size = 0.5 or size = 1.6 for points
alpha = 0.75

Something cool

library(palmerpenguins)
library(tidyverse)

plot_data <- storms %>%
group_by(year, status) %>%
summarise(wind = mean(wind, na.rm = T))

plot_data %>%
ggplot(aes(year, wind, col = status)) +
geom_line() +
geom_line(aes(col = "All of NZ"),
data = plot_data %>% group_by(year) %>% summarise(wind = sum(wind)))

Change labelling defaults to prettyNum(x, big.mark = "", scientific = FALSE)

Avoid this happening all the time, which is annoying

new function: add `gg_bin2d` function

col_method quantile labels might be misleading

how to make a column chart with gg_blank

gg_blank(penguins, 
         x = body_mass_g, 
         y = flipper_length_mm, 
         dye = sex,
         y_zero = TRUE, 
         position = "fill", 
         x_limits = c(NA, NA),
         ) +
  geom_col(aes(x = body_mass_g, 
               y = flipper_length_mm, 
               col = sex, fill = sex), 
           position = "fill")

Where dye is not an aesthetic, the col/fill are not inherited

Make everything run off scale_*_manual

Position & stat are not inherited

This will need to be clarified in the plot

Also, shortcut functions will get around this...

colour method sorted :)

Default to continuous

Default to col_bins = NULL, which can colour continuously

Choose col_bins = "evenly" to turn chopping on
"evenly", "quantiles", "mean_sd"

Fine-tune with arg's:
col_intervals = 5
col_probs = seq(0, 1, 0.25)
col_sds = 1:3
col_pretty_n = 5
col_breaks = c(0, 300, 500, 1000)

Might need some work with col_breaks to make it work with gradient

Make it all left, and colour them using nice existing method

consider supporting `gg_density2d` & `gg_density2d_filled` functions

gg_histogram: default to 40 bins, alpha =0.75, + add x/ytitle

package name: ggblanket or blanket

library(blanket)
-pros: more readable
-cons: not linked to ggplot, unless you know about it, not a name for a leaf wrapper package available

library(ggblanket)

For area, bar, etc ... where bar-like polygons, default limits

if both x & y numeric, use...
x_limits = c(min(plot_data$year), max((plot_data$year))),

This is because, coord_flip has to be used to have it horizontal where both numeric

For points, lines etc, this would look less good

date_time *_breaks do not work

library(tidyverse)
library(lubridate)
library(ggblanket)

df <- lakers %>%
  slice_sample(n = 100) %>% 
  tibble() 

df %>% 
  gg_blank(time, points, x_breaks_n = 5) +
  geom_point()

x/y scales use cut_evenly instead??

x_zero is not working all the time

gg_blank(storms, x = year, y = wind, stat = "boxplot", position = "dodge2", x_zero = F) +
  geom_boxplot(aes(group = year), stat = "boxplot", position = "dodge2", alpha = 0.5)

Add breaks_width argument

library(tidyverse)
library(palmerpenguins)
library(ggblanket)

df <- penguins %>% 
  drop_na(body_mass_g)

breaks_n
breaks_width

scales::fullseq(range(df$body_mass_g), size = 750)
scales::fullseq(range(economics$date), size = "1 year")

Add group aes with options of 1 or a var. Default to NULL

Consider using expand_limits

Currently:

x_zero
y_zero

https://ggplot2.tidyverse.org/reference/expand_limits.html

y_limits_include= c(0, 10)
x_limits_include = 0

if y and col the same, don't reverse both

col: add limits code for when col_drop = TRUE