A word cloud displays text data where word size represents frequency or importance. Words are arranged to fill available space, creating a visual summary of text content that highlights prominent terms and patterns. This visualization is ideal for quickly identifying the most common themes or keywords in a body of text.
#' anyplot.ai
#' wordcloud-basic: Basic Word Cloud
#' Library: ggplot2 3.5.1 | R 4.4.1
#' Quality: 91/100 | Created: 2026-08-04
library(ggplot2)
library(tibble)
library(ragg)
set.seed(42)
# --- Theme tokens -------------------------------------------------------------
THEME <- Sys.getenv("ANYPLOT_THEME", "light")
PAGE_BG <- if (THEME == "light") "#FAF8F1" else "#1A1A17"
INK <- if (THEME == "light") "#1A1A17" else "#F0EFE8"
IMPRINT_PALETTE <- c("#009E73", "#C475FD", "#4467A3", "#BD8233",
"#AE3030", "#2ABCCD", "#954477", "#99B314")
BRAND <- IMPRINT_PALETTE[1]
# --- Data: term frequencies from a renewable-energy report scan ---------------
terms <- c(
"solar", "wind", "energy", "renewable", "grid", "battery", "storage",
"efficiency", "carbon", "emissions", "turbine", "hydropower", "biomass",
"geothermal", "investment", "innovation", "climate", "capacity",
"generation", "transition", "policy", "technology", "electricity",
"consumption", "demand", "supply", "market", "subsidy", "reliability",
"recycling"
)
rank_order <- seq_along(terms)
base_freq <- 150 / rank_order^0.65
frequency <- pmax(round(base_freq + rnorm(length(terms), mean = 0, sd = 4)), 5)
word_freq <- tibble(word = terms, frequency = frequency)
word_freq <- word_freq[order(-word_freq$frequency), ]
# Font size scales with sqrt(frequency) so rendered AREA (not height) tracks
# term frequency, matching how the eye compares word-cloud prominence.
size_min <- 3.2
size_max <- 18
freq_sqrt <- sqrt(word_freq$frequency)
word_freq$text_size <- size_min + (freq_sqrt - min(freq_sqrt)) /
diff(range(freq_sqrt)) * (size_max - size_min)
# --- Layout: box-bounded spiral placement with rectangle collision ------------
# geom_text's `size` aesthetic is in millimetres, so the layout below works
# directly in millimetres too (1 coordinate unit = 1 mm) and the panel is
# rendered at that exact physical scale via coord_fixed(ratio = 1) below —
# this keeps the hand-measured bounding boxes true to the rendered glyphs.
# Bounding-box multipliers are calibrated from grid::grobWidth/grobHeight on
# the bold sans face actually used for geom_text (see dev notes: width per
# char ranges ~0.55-0.68x fontsize, height ~0.735x fontsize; padded up here
# for a safety margin against overlap).
half_w_per_char <- 0.35
half_h_factor <- 0.42
box_half_w <- 86 # mm — half-width of the placement box (172mm total)
box_half_h <- 44 # mm — half-height of the placement box (88mm total)
x_stretch <- box_half_w / box_half_h
n_words <- nrow(word_freq)
placed_x <- numeric(n_words)
placed_y <- numeric(n_words)
placed_hw <- numeric(n_words)
placed_hh <- numeric(n_words)
max_radius <- sqrt(box_half_w^2 + box_half_h^2)
# Fallback grid, closest-to-center first, for words the spiral walk can't
# place within its step budget (a spiral path only samples one point per
# radius, so it can miss small pockets of free space that a grid catches).
grid_x <- seq(-box_half_w, box_half_w, by = 1)
grid_y <- seq(-box_half_h, box_half_h, by = 1)
fallback_grid <- expand.grid(x = grid_x, y = grid_y)
fallback_grid <- fallback_grid[order(sqrt((fallback_grid$x / x_stretch)^2 + fallback_grid$y^2)), ]
for (i in seq_len(n_words)) {
fs <- word_freq$text_size[i]
half_w <- nchar(word_freq$word[i]) * fs * half_w_per_char
half_h <- fs * half_h_factor
placed <- FALSE
if (i == 1) {
cx <- 0
cy <- 0
placed <- TRUE
} else {
theta <- 0
radius <- 0
step <- 0
repeat {
cx <- radius * cos(theta) * x_stretch
cy <- radius * sin(theta)
in_box <- (abs(cx) + half_w <= box_half_w) && (abs(cy) + half_h <= box_half_h)
overlap <- !in_box
if (in_box) {
for (j in seq_len(i - 1)) {
dx <- abs(cx - placed_x[j])
dy <- abs(cy - placed_y[j])
if (dx < (half_w + placed_hw[j]) * 1.06 && dy < (half_h + placed_hh[j]) * 1.15) {
overlap <- TRUE
break
}
}
}
if (!overlap) {
placed <- TRUE
break
}
theta <- theta + 0.12
radius <- min(radius + 0.18, max_radius)
step <- step + 1
if (step > 8000) break
}
}
if (!placed) {
for (k in seq_len(nrow(fallback_grid))) {
gx <- fallback_grid$x[k]
gy <- fallback_grid$y[k]
if (abs(gx) + half_w > box_half_w || abs(gy) + half_h > box_half_h) next
overlap <- FALSE
for (j in seq_len(i - 1)) {
dx <- abs(gx - placed_x[j])
dy <- abs(gy - placed_y[j])
if (dx < (half_w + placed_hw[j]) * 1.06 && dy < (half_h + placed_hh[j]) * 1.15) {
overlap <- TRUE
break
}
}
if (!overlap) {
cx <- gx
cy <- gy
placed <- TRUE
break
}
}
}
placed_x[i] <- cx
placed_y[i] <- cy
placed_hw[i] <- half_w
placed_hh[i] <- half_h
}
word_freq$x <- placed_x
word_freq$y <- placed_y
# --- Plot -----------------------------------------------------------------
plot_title <- "wordcloud-basic · r · ggplot2 · anyplot.ai"
title_ratio <- if (nchar(plot_title) > 67) 67 / nchar(plot_title) else 1.0
title_size <- round(12 * title_ratio)
p <- ggplot(word_freq, aes(x = x, y = y, label = word,
size = text_size, color = frequency)) +
geom_text(fontface = "bold", family = "sans") +
scale_size_identity() +
scale_color_gradient(low = BRAND, high = "#4467A3", guide = "none") +
scale_x_continuous(limits = c(-box_half_w, box_half_w), expand = c(0, 0)) +
scale_y_continuous(limits = c(-box_half_h, box_half_h), expand = c(0, 0)) +
coord_fixed(ratio = 1, clip = "off") +
labs(title = plot_title) +
theme_void(base_size = 8) +
theme(
plot.background = element_rect(fill = PAGE_BG, color = PAGE_BG),
plot.title = element_text(color = INK, size = title_size, hjust = 0.5,
margin = margin(b = 10)),
plot.margin = margin(t = 16, r = 24, b = 12, l = 24)
)
# --- Save -------------------------------------------------------------------
ggsave(
filename = sprintf("plot-%s.png", THEME),
plot = p,
device = ragg::agg_png,
width = 8,
height = 4.5,
units = "in",
dpi = 400
)
Part of Basic Word Cloud on anyplot.ai.