Interoperability, at Last

Language is a very powerful way to describe behavior. Therefore even when I create pictures, instead of dragging around my mouse, I use declarative tools like GraphViz, gnuplot, and UMLGraph. These allow me to describe what I want to draw, instead of how I want the end-result to look like. The truth however is that the end-results are not always perfect. Today I realized that the state of the art has advanced to the point where I can create the drawing declaratively, and then visually polish the final drawing.

Here is an example. For an IEEE Software special-issue on development tools that I'm co-authoring we wrote a table listing the introduction date and level of abstraction of various tools. Here is an excerpt (I'm using an early draft version of the data).

# Abstraction:Year:Tool
4:1977:Make
5:1979:Lex/Yacc
6:1971:Smalltalk IDE
4:1972:SCCS
2:1973:Grep

#!/usr/bin/awk -F:
BEGIN {
	print "reset"
	print "set terminal svg"
	print "set output 'timeline.svg";
	print "unset ytics";
	print "set ylabel 'Level of Abstraction'"
	print "set xlabel 'Year'"
}
!/^#/ {
	# Random pertrubation by +/- 0.5
	$2 += rand() - 0.5
	$1 += rand() - 0.5
	print "set label \"" $3 "\" at " $2 ", " $1
}
END {
	print "plot [1970:2010] [0:9] -1 notitle"
	print "set output 'nul'"
}

However, I then loaded the resulting SVG file into Inkscape. It loaded perfectly, and I was thus able to easily adjust the labels' positions using the mouse. Here is the result after the fine-tuning.

Comments   Toot! Tweet Share