Plots with ROOTs

The live notebook names this table titanic_by_class. Rows represent three passenger classes and Crew, a separate group, not a fourth class. Counts include everyone in titanic_by_class.csv from Zenodo (v1.0.1).

The live notebook's first useful comparison plots survivors and deaths side-by-side (listing 1). Start with the source, then read the column mapping and appearance options from left to right.

plot data using 3:xtic(1) title "Survived", \
     ''   using 4         title "Died"

Compare the stacked counts in listing 4 with the sex comparison in listing 9. Both plots are printed beside their code. Blue represents survivors in the first, women in the second. Linetype numbers select styles, not meanings; the series and legend define what each colour represents.

Listing 1 draws both count columns; using 3 alone draws only survivors. Listing 6 adds deliberate colours, guides and axis styling.

The table itself gives useful checks: 201 + 123 = 324 in first class and 211 + 679 = 890 for Crew. Across all four groups there are 2,207 people, 711 of whom survived. If a later edit breaks these sums, check the table before debugging the plot.

Gnuplot reads the first numeric field as y if only one is given. For numeric x and y, write both fields: using 1:2. Parentheses distinguish a calculation from a column number: using ($3/$2*100) means a percentage, while using 3 means column 3.

Labels versus coordinates: xtic(1) labels equally spaced categories; it does not set x positions. For ages 1, 4 and 10, use the age column as x to preserve gaps of 3 and 6 years. Row numbers would make the gaps equal.

For each class, survival rate is survived / aboard. In first class that is 201 / 324 = 62.0%. The live notebook computes this from columns 3 and 2, then plots bars and percentage labels (listing 2).

set yrange [0:100]
set format y "%.0f%%"
plot data using 0:($3/$2*100):xtic(1) \
       with boxes notitle, \
     '' using 0:($3/$2*100): \
       (sprintf("%.1f%%", $3/$2*100)) \
       with labels notitle

The bar height is the rate, not the survivor count. Third class has 181 survivors but 709 passengers aboard; the denominator is what makes the bar comparable with the smaller classes.

The plot command above draws the same values twice: bars first, labels second. offset 0,1 moves labels above the bars; it is measured in character heights. set offsets 0.5,0.5,0,0 leaves horizontal room at both ends.

Use 711.0 / total when the result must be fractional: integer-only division can discard the fractional part. Variables are case-sensitive; ordinary variables are named total, while $3 refers to a data column.

The comparison in listing 9 uses each group's own aboard count.

First-class women: 139/144 = 96.5%. First-class men: 62/180 = 34.4%.

Use grouped bars when the values should be compared side by side. In the Titanic figure, each class has one bar for survivors and one for deaths.

set style data histograms
set style histogram clustered gap 1
set yrange [0:*]
plot data using 3:xtic(1) title "Survived", \
     ''   using 4         title "Died"

set style histogram rowstacked
plot data using 3:xtic(1) title "Survived", \
     ''   using 4         title "Died"

rowstacked puts each class's series in one bar. Its total height is everyone aboard: 324 in first class, 284 in second, 709 in third, 890 crew. This answers a different question from the grouped chart.

Divide each outcome by the class total before stacking. The surviving and dead portions now add to 100% in each class (listing 5).

set style histogram rowstacked
set yrange [0:100]
set format y "%.0f%%"
plot data using ($3/$2*100):xtic(1) title "Survived", \
     ''   using ($4/$2*100)         title "Died"

Every bar reaches 100%. Compare the blue portions: 62.0%, 41.5%, 25.5% and 23.7% survived. The grey portions are the complementary death shares.

Crew has the most deaths and is also the largest group. Rates compare chances within groups. A pie is also a share plot, but is harder to compare precisely; use it for one composition only.

Counts and rates describe different things. Third class had more deaths (528) than first and second combined (289), while the survival rate compares each class using its own total.

<<plot-style>> inserts the notebook's shared settings (Noweb). Edit plot-style once to change the defaults for every plot.

set border 3 lw 1.9
set tics out nomirror scale 0.9
set grid ytics lc rgb "#d5d8dc" lw 0.5
set key top left
set linetype 1 lc rgb "#14507d"
set linetype 2 lc rgb "#aab0b8"
set style fill solid 0.9 border -2

Each line connects men's and women's rates within a group (listing 7). Its length is the gap in percentage points.

plot sex using ($6/$5*100):0: \
  (($3/$2-$6/$5)*100):(0):ytic(1) \
  with vectors nohead notitle, \
  '' using ($3/$2*100):0 with points title "Women", \
  '' using ($6/$5*100):0 with points title "Men"

For 2D borders, add flags: bottom 1, left 2, top 4, right 8. set border 3 selects bottom + left (1 + 2). Other examples: set border 1 (bottom), set border 12 (top + right), set border 15 (all four); unset border removes all borders.

The values must already be percentages for %.0f%% to be truthful. Axis labels should say what a number measures, such as People aboard or Survived. The live notebook keeps plot titles in captions, so the graph stays uncluttered.

C-c C-v t tangles the named block to orgmode/plots/plot-style.gp. A plain gnuplot script can then use load "orgmode/plots/plot-style.gp". A personal ~/.gnuplot also sets defaults, but those settings are hidden from someone opening the script.

<<plot-style>>
set ylabel "People aboard"
plot data using 3:xtic(1) title "Survived", \
     ''   using 4         title "Died"

set linetype 1 lc rgb "#14507d"  # women
set linetype 2 lc rgb "#c47a2c"  # men
plot sex using ($3/$2*100):xtic(1) title "Women", \
     ''  using ($6/$5*100)         title "Men"

The colour belongs to the series; title gives the series its meaning in the legend. Readers should not have to guess what blue represents.

The live pie uses the counts in column 2. Each group becomes a wedge proportional to its share of 2,207 people. This is a constructed pie from circle wedges, a compact recipe that remains valid in gnuplot 6. The running angle angle_end carries each slice's end into the next row.

unset polar
set autoscale
stats data using 2 nooutput
total = STATS_sum
angle_for_count(n) = n * 360.0 / total
set angles degrees
set size ratio -1
set xrange [-1.05:1.05]
set yrange [-1.05:1.05]
set style fill solid 1 border linecolor rgb "white"
angle_end = 0
plot data using (0):(0):(1):(angle_end): \
  (angle_end = angle_end + angle_for_count($2)):($0+1) \
  with circles linecolor variable notitle

The Crew wedge is largest: 890/2207 = 40.3% of everyone aboard. This is a share of the dataset, not a survival rate. Starting at the right and moving anticlockwise, the unlabelled slices follow the table order: 1st, 2nd, 3rd, Crew. This keeps the first plotting step short. Run stats with polar mode off and before setting the fixed ranges.

Challenge 12 adds names and percentages as a separate label pass, using the middle angle of each slice. The geometry remains unchanged.

The notebook's heatmap places women at x=0, men at x=1, and one class per row. Cell colour encodes survival rate; the reversed y range puts first class at the top (listing 11).

The titanic_by_sex table stores women's aboard and survived counts in columns 2 and 3, men's in columns 5 and 6. Each rate uses the matching aboard count.

set palette defined (0 "#ffffff", 100 "#4f86b5")
set cbrange [0:100]
set xrange [-0.5:1.5]
set yrange [3.5:-0.5]
set xtics ("Women" 0, "Men" 1)
plot sex using (0):0:(0.5):(0.5):($3/$2*100) \
  with boxxyerror lc palette, \
     '' using (1):0:(0.5):(0.5):($6/$5*100) \
  with boxxyerror lc palette

Women in first class have the darkest cell: 139/144 = 96.5% survived. Men in second class have the lightest: 24/178 = 13.5%. The labels show exact values; the colour scale helps locate high and low values quickly.

The eight rates also appear in the grouped bars (listing 9) and dumbbell chart (listing 7). Bars encode rate by height, the dumbbell by position, and this heatmap by colour. Fixing cbrange [0:100] keeps colours comparable when another dataset is plotted.

The live spider plot uses five percentages for each class: overall survival, women's survival, men's survival, and the shares of women and men aboard (listing 12). Each row of titanic_by_sex makes a polygon; every plot component adds an axis.

set spiderplot
set for [p=1:5] paxis p range [0:100]
set grid spiderplot
plot sex using (($3+$6)/($2+$5)*100):key(1) \
       with spiderplot notitle, \
     sex using ($3/$2*100) with spiderplot notitle, \
     sex using ($6/$5*100) with spiderplot notitle, \
     sex using ($2/($2+$5)*100) with spiderplot notitle, \
     sex using ($5/($2+$5)*100) with spiderplot notitle

The course uses gnuplot 6.0 or newer. The shape depends on axis order; use bars when exact comparisons are more important than seeing a profile.

The two aboard shares sum to 100% in each class. First class reaches 96.5% on the women's survival axis but only 34.4% on the men's.

The fig-7e block in gnuplot_guide_orgmode.org draws annular sectors: survivors and deaths outside, class sizes inside. The positive inner radius leaves a real hole; no white covering circle is needed. Reset pos before each pass so the group boundaries align across both rings.

The two rings occupy separate radial intervals: 0.40-0.72 for class and 0.72-1.00 for outcomes. Class names sit halfway across the inner ring; rate labels sit outside radius 1. Labels use polar angle:radius positions.

Inner ring: class size. Outer ring: outcome within each class. The outside labels give each class's survival rate.

titanic_by_age supplies counts and survival rates for the 120-by-24-pixel sparklines. Both include crew and omit unknown ages.

unset border; unset tics; unset key; unset grid
set margins 0,0,0,0
set yrange [0:100]
plot data using 0:3 with lines

Age-group counts and survival rates , across 25 three-year groups, from ages 0–2 to 72–74. Listing 15 shows how to select an SVG output file.

The table name is from Org. The name data on the left of the equals sign is chosen by the author; it is not a gnuplot keyword.

#+begin_src gnuplot :var data=titanic_by_class \
    :file orgmode/figures/by_class_outcomes_stacked.svg
<<plot-style>>
set style histogram rowstacked
plot data using 3:xtic(1) title "Survived", \
     ''   using 4         title "Died"
#+end_src

Export uses stored results: rerun changed plots first. A fresh gnuplot session avoids inheriting settings from an earlier plot.

These options complement the sparkline in listing 13. The terminal defines the SVG canvas; the document controls its printed size.

Counts peak at ages 21–23. Rates peak at 67.9% for ages 3–5; the three oldest groups have no recorded survivors and very small counts. Crew is included; only the two missing ages are excluded.

Run terminal/all.gp from terminal/ for the complete CSV examples. Listing 15 illustrates output setup using a $T data block.

set terminal svg size 900,400 font "sans,13"
set output "chart.svg"
set style data histograms
set yrange [0:*]
plot $T using 3:xtic(1) title "Survived", \
     '' using 4 title "Died"
unset output

In both cases the plot command reads the same columns. Only the way data and output reach gnuplot changes.

In the standalone script, change set yrange [0:*] to [0:250] and regenerate the class-count plot. Then change the file name in set output. The counts stay the same; the visible scale and saved filename change.

Further reading: Org Babel gnuplot tutorial (historical advice); Plotting tables with org-plot (#+PLOT: options).

Use Bash or Git Bash, gnuplot 6.0+, curl, a text editor and a browser. Install gnuplot separately when using Git Bash.

for p in bash gnuplot curl head cat mkdir; do
  command -v "$p" || printf 'MISSING: %s\n' "$p"
done
gnuplot --version

Fix missing programs first; gnuplot_guide_terminal.org has the full SVG check.

Run listing 17 from the course folder; then stay in terminal/.

mkdir -p terminal
cd terminal
mkdir -p data figures plots
export LC_NUMERIC=C

If LC_ALL is set, use unset LC_ALL first. Script paths are relative to the working directory: stay in terminal/, not plots/.

From terminal/, use the download loop in chapter 1 of gnuplot_guide_terminal.org to retrieve all five CSVs from Zenodo version 1.0.1 into data/. Both lessons use the published summaries directly; no AWK preparation is needed.

Counts cover 2,207 people and 711 survivors; age summaries exclude two unknown ages. Use set datafile separator "," and set datafile columnheaders so gnuplot reads CSV and treats the header as metadata. Do not also skip the first data record.

Save listing 18 as plots/by_class_aboard.gp inside terminal/. Use the .gp extension, not .gp.txt.

reset
set encoding utf8
set datafile separator ","
set datafile columnheaders
data = 'data/titanic_by_class.csv'
set terminal svg size 900,300 font 'sans,13'
set output 'figures/by_class_aboard.svg'
set style data histograms
set yrange [0:*]
plot data using 2:xtic(1) title "Aboard"
unset output

In Bash, run the saved file with listing 19:

gnuplot -d plots/by_class_aboard.gp

Open figures/by_class_aboard.svg in a browser. After editing: save the file, rerun the command, then reload the browser. SVG output does not open a plot window automatically.

Run listing 19 in Bash. Put plot, set and load in the script or at the gnuplot> prompt; do not copy the prompt itself. Bash expands $3 as a shell parameter; gnuplot reads it as column 3.

Save style.gp in terminal/ before using the styled plots in chapter 6. It changes appearance; it does not create the data or draw a figure.

all.gp is created in the rerun topic of the terminal lesson. Lines such as load 'plots/by_class_aboard.gp' run the saved plots. Prepare all sixteen scripts, style.gp and the three tables before gnuplot -d all.gp.

Cannot open a file? Check pwd, spelling and prerequisites. Tangling recreates solution files and overwrites local edits; students can write every file manually.

Companion: gnuplot_guide_terminal.org. Keep downloaded CSVs and saved scripts together to rebuild the figures offline. Optional book: Philipp K. Janert, Gnuplot in Action, Second Edition (Manning, 2016).