I don't remember where I saw this first - probably in the Debian Installer code. Anyway, given an n-tuple it can be somewhat unwieldy to unpack it:
VAR="Aho Sethi Ullman" ONE="$(echo ${VAR} | cut -d' ' -f1)" TWO="$(echo ${VAR} | cut -d' ' -f2)" THREE="$(echo ${VAR} | cut -d' ' -f3)"
The trick being presented is to generate the boilerplate code and then eval it:
VAR="Aho Sethi Ullman" eval $(echo ${VAR} | awk '{ print "ONE=" $1 " TWO=" $2 " THREE=" $3 }')
Here's a more concrete example, with apologies to Python:
for PAIR in \ "one Aho" \ "two Sethi" \ "three Ullman" do eval $(echo ${PAIR} | awk '{ print "k=" $1 " v=" $2 }') # .. do something with $k and $v done