plot.awk

#!/usr/bin/awk -f

# This program is a copy of guff, a plot device. https://github.com/silentbicycle/guff
# My copy here is written in awk instead of C, has no compelling benefit.
# Public domain. @thingskatedid

# Run as awk -v x=xyz ... or env variables for stuff?
# Assumptions: the data is evenly spaced along the x-axis

# TODO: moving average
# TODO: trend lines, or guess at complexities
# TODO: points vs. lines
# TODO: colourblind safe scheme
# TODO: center data around the 0 axis
# TODO: scanning for all float formats input, including -inf, NaN etc
# TODO: guess at whether to use lines or circles, based on delta within a window?

function hastitle() {
	for (i = 1; i <= NF; i++) {
		if ($i ~ /[^-0-9.]/) {
			return 1
		}
	}

	return 0
}

function amax(a,	i, max) {
	max = -1

	for (i in a) {
		if (max == -1 || a[i] > a[max]) {
			max = i
		}
	}

	return max
}

function normalise(	delta) {
	for (i = 1; i <= NF; i++) {
		max[i] = 0
		min[i] = 0

		for (j = 1; j <= NR; j++) {
			if (a[i, j] > max[i]) {
				max[i] = a[i, j]
			} else
			if (a[i, j] < min[i]) {
				min[i] = a[i, j]
			}
		}

		delta[i] = max[i] - min[i]

		for (j = 1; j <= NR; j++) {
			a[i, j] -= min[i]
			if (delta[i] > 0) {
				a[i, j] /= delta[i]
			}
		}
	}

	# TODO: rescale to center around 0

	# Here the data is squished slightly in descending order of deltas.
	# Each column is scaled independently anyway, so they're never to scale.
	# The idea here is to help show intutively which are smaller, but without
	# actually drawing them to size (since then very small deltas would not be
	# visible at all).

	k = 0
	prev = -1

	while (length(delta) > 0) {
		i = amax(delta)

		# Several columns can share the same delta
		# Formatting to %.3f here is just for sake of rounding
		if (prev != -1 && sprintf("%.3f", prev) != sprintf("%.3f", delta[i])) {
			k++
		}

		# +2 to squish things upwards a bit
		scale = (NF + 2 - k) / (NF + 2)

		# there's no need to scale by 1
		if (scale != 1) {
			for (j = 1; j <= NR; j++) {
				a[i, j] *= scale
			}
		}

		prev = delta[i]
		delete delta[i]
	}
}

# internal coordinates to svg coordinates
function point(x, y) {
	x = x * (chart_width - 2 * xmargin) + xmargin
	y = (height - 2 * ymargin) - y * (height - 2 * ymargin) + ymargin

	return sprintf("%u,%u", x, y)
}

function line(i) {
	printf "  <polyline stroke='%s%s' stroke-width='1.5' fill='none'", color[i], alpha
	printf " points='"

	for (j = 1; j <= NR; j++) {
		printf "%s ", point((j - 1) / NR, a[i, j])
	}

	printf "'/>\n"
}

function circles(i) {
	for (j = 1; j <= NR; j++) {
		p = point((j - 1) / NR, a[i, j])
		split(p, q, ",")
		printf "  <circle cx='%u' cy='%u' r='1.2' fill='%s%s' stroke='%s%s'/>\n",
			q[1], q[2], color[i], alpha, color[i], alpha
	}
}

function legend_text(i, title) {
	printf "  <g transform='translate(%u %u)'>\n", chart_width + gutter, i * line_height
	printf "    <circle cx='%d' cy='%d' r='3.5' fill='%s' stroke='%s'/>\n",
		-10, -line_height / 2 + 5, color[i], color[i]
	printf "    <text style='%s' xml:space='preserve'>%-*s[%.3g, %.3g]</text>\n",
		sprintf("fill: %s; font-size: %upx; font-family: mono", fg, font_size),
		(title_width > 0) ? title_width + 1 : 0, title, min[i], max[i]
	printf "  </g>\n"
}

function display() {
	print "<?xml version='1.0'?>"
	printf "<svg xmlns='%s' width='%u' height='%u' version='1.1'>\n",
		"http://www.w3.org/2000/svg",
		chart_width + gutter + legend_width, height

	title_width = 0
	for (i = 1; i <= NF; i++) {
		if (length(title[i]) > title_width) {
			title_width = length(title[i])
		}
	}

	for (i = 1; i <= NF; i++) {
#		line(i)
		circles(i)
	}

	if (length(title)) {
		for (i = 1; i <= NF; i++) {
			legend_text(i, title[i])
		}
	}

	print "</svg>"
}

NR == 1 {
	if (hastitle()) {
		for (i = 1; i <= NF; i++) {
			title[i] = $i
		}

		NR--
		next
	}
}

{
	for (i = 1; i <= NF; i++) {
		a[i, NR] = $i
	}
}

END {
	fg    = "#eeeeee"
	alpha = "ff"

	# Bang Wong's colour-safe palette, https://www.nature.com/articles/nmeth.1618
	# (using just the last five colours)
	color[3] = "#009E73"
	color[2] = "#F0E442"
	color[1] = "#0072B2"
	color[4] = "#CC79A7"
	color[5] = "#D55E00"

	color[6] = fg

	if (NF == 1) {
		color[1] = fg
	}

	if (NF > length(color)) {
		print "too many fields" >> "/dev/stderr"
		exit 1
	}

	chart_width=320
	legend_width=300
	height=120
	xmargin=0
	ymargin=5
	gutter=30
	font_size=15
	line_height=20

	# the data is scaled 0..1 for our internal coordinate space
	normalise()

	display()
}

Optional titles are detected from the data (in TSV format), when present:

; head -5 /tmp/data 
blorgh stuff things
0 1 0
1.40159 0.495601 1172922
-1.3012 0.0455791 761306
1.69096 -0.227783 800372

Now with (hopefully) colourblind-friendly colours. Colours are only used when there's more than one column to show.

Sorry for newb question: icat was totally new to me... is this using https://github.com/kovidgoyal/kitty or something else?

Yes, that's a shell alias to kitty icat