ggplot2 geom_bar() vs geom_col() in R: Bar Charts Made Easy
The geom_bar() function in ggplot2 counts rows by category and draws bars; geom_col() uses values you provide directly. Use geom_bar for frequency tables, geom_col for pre-aggregated data.
ggplot(df, aes(x = category)) + geom_bar() # count rows ggplot(df, aes(x = category, y = value)) + geom_col() # use y values ggplot(df, aes(x = category, fill = group)) + geom_bar() # stacked ggplot(df, aes(x = category, fill = group)) + geom_bar(position = "dodge") # grouped ggplot(df, aes(x = category, fill = group)) + geom_bar(position = "fill") # 100% stacked ggplot(df, aes(x, y)) + geom_col() + coord_flip() # horizontal ggplot(df, aes(x = reorder(category, -count), y = count)) + geom_col() # sorted
Need explanation? Read on for examples and pitfalls.
What geom_bar() and geom_col() do in one sentence
geom_bar() counts rows in each x category and draws a bar of that height; geom_col() uses the y values you provide directly without counting. They produce identical-looking output but expect different inputs: geom_bar wants raw rows; geom_col wants pre-aggregated (category, value) pairs.
If you find yourself writing geom_bar(stat = "identity"), you actually want geom_col(). They are equivalent but geom_col() is the modern, less-confusing form.
Syntax
Both functions take an aes() mapping. geom_bar() requires only x; geom_col() requires both x and y.
The full signatures:
geom_bar(mapping = NULL, data = NULL, stat = "count", position = "stack",
..., width = NULL, na.rm = FALSE, orientation = NA, show.legend = NA,
inherit.aes = TRUE)
geom_col(mapping = NULL, data = NULL, position = "stack", ...)geom_col() whenever you have pre-aggregated data. geom_bar(stat = "identity") produces the same result but is harder to read. geom_col is the cleaner alternative introduced in ggplot2 2.2+.Seven common patterns
1. Count rows with geom_bar
geom_bar() calls stat = "count" internally, tallying rows per class. No y aesthetic needed.
2. Use precomputed values with geom_col
geom_col() plots y as bar height. Use this when you have a frequency table or any table with one row per category.
3. Stacked bars with fill aesthetic
fill = drv automatically stacks bars colored by drivetrain. Total bar height = total count per class.
4. Side-by-side (dodged) bars
position = "dodge" places each fill group's bar SIDE-BY-SIDE instead of stacked. Useful for comparing magnitudes across groups within each x category.
5. 100% stacked bars
position = "fill" makes every bar reach 1.0 (or 100%). Inside each bar, fill segments show PROPORTIONS of each group within that x category.
6. Horizontal bars
coord_flip() swaps x and y. Useful when category labels are long and would overlap on a vertical axis.
7. Sorted bars
fct_infreq() from forcats reorders the factor levels by descending frequency. Combined with coord_flip(), this produces a clean ranked horizontal bar chart.
geom_bar() and geom_col() are nearly identical visually but their INPUT differs. geom_bar expects raw row data and counts internally. geom_col expects pre-aggregated values. Reach for whichever matches your data: row-level → geom_bar; aggregated → geom_col.geom_bar() / geom_col() vs base R barplot()
Base R barplot() works on tables/vectors; ggplot2's bar geoms work on data frames with grouping baked in.
| Task | ggplot2 | Base R |
|---|---|---|
| Count rows per category | aes(x=cat) + geom_bar() |
barplot(table(df$cat)) |
| Pre-aggregated values | aes(x, y) + geom_col() |
barplot(values, names = labels) |
| Stacked | aes(fill=grp) + geom_bar() |
barplot(table(df$grp, df$cat)) |
| Grouped (dodged) | position = "dodge" |
beside = TRUE |
| Horizontal | + coord_flip() |
horiz = TRUE |
| Sort by value | aes(x = reorder(...)) |
sort + barplot |
When to use which:
- Use ggplot2 for grouped, faceted, or publication bar charts.
- Use base
barplot()for one-line frequency plots.
Common pitfalls
Pitfall 1: using geom_bar with pre-aggregated data. If your data is already (category, value), geom_bar will try to COUNT rows, ignoring your y values. Use geom_col() or geom_bar(stat = "identity", aes(y = value)).
Pitfall 2: fill vs color confusion. Bar charts use fill for the inside color and color for the outline. aes(color = group) only changes outlines; aes(fill = group) changes the bar interiors.
Pitfall 3: x axis ordering. ggplot orders categorical x alphabetically by default. To reorder by frequency: aes(x = forcats::fct_infreq(class)). By a value column: aes(x = reorder(category, -value)).
Try it yourself
Try it: Build a horizontal bar chart from mpg showing the COUNT of cars per manufacturer, sorted from highest to lowest count. Save to ex_plot.
Click to reveal solution
Explanation: fct_infreq() orders factor levels by descending frequency. After coord_flip(), the highest-frequency manufacturer appears at the TOP of the y axis (visually at the top of the chart).
Related ggplot2 functions
After mastering geom_bar() and geom_col(), look at:
geom_histogram(): bar-like display for continuous x (binned)geom_freqpoly(): line version of a histogramgeom_text(): annotate bars with their valuesposition_dodge(),position_stack(),position_fill(): explicit position controlsforcats::fct_reorder(),fct_infreq(): factor-level ordering helpersscale_y_continuous(labels = scales::percent): format y as percentages
For pictogram-style displays, the waffle package extends ggplot2 with geom_waffle().
FAQ
What is the difference between geom_bar and geom_col in ggplot2?
geom_bar() counts the number of rows in each x category. geom_col() uses values from a y column directly. If your data has one row per observation, use geom_bar. If you have pre-aggregated (category, value) pairs, use geom_col.
How do I make a stacked bar chart in ggplot2?
Map fill = group inside aes(): geom_bar(aes(fill = group)). Default position is "stack". For side-by-side bars, add position = "dodge". For 100%-stacked, use position = "fill".
How do I sort bars in ggplot2 by count or value?
By count (frequency): aes(x = forcats::fct_infreq(category)) then geom_bar(). By value: aes(x = reorder(category, -value)) then geom_col(). Add coord_flip() for horizontal sorted bars (often more readable for long labels).
How do I add value labels to bars in ggplot2?
Add geom_text() after geom_col(): + geom_text(aes(label = value), vjust = -0.3). The vjust controls vertical position relative to the bar top.
How do I make horizontal bars in ggplot2?
Add coord_flip() after the bar geom: geom_bar() + coord_flip(). Newer ggplot2 versions also support aes(y = category) directly which auto-flips. Use whichever feels cleaner.