Get matching group from previous line - regex

Question

I'm working on a sed script to run through a file and make substitutions. The existing file will have a sequence of floating point numbers, and the sequence ends when a letter is found. Most of the substitutions are straightforward and look like this:
s/(-?[0-9]*\.?[0-9]*) (-?[0-9]*\.?[0-9]*) l/lineto(\1,\2);/g
To just replace the raw command with a function call.
Some commands have no 1:1 equivalent to a function call, because they depend on coordinates found on the previous line.
So I need to turn this:
1.068 7.399 m
-11.794 13.153 -11.843 12.234 v
Into this:
move(1.068,7.399);
curveto(1.068,7.399,-11.794,13.153,-11.843,12.234);
The last set of coordinates from the previous line needs to be used as the first set of coordinates for this line. The coordinates in the previous line don't always end in the same token, so that this:
-7.451 17.792 -10.366 16.42 -11.198 14.444 c
-11.794 13.153 -11.843 12.234 v
Needs to become this:
curveto(-7.451,17.792,-10.366,16.42,-11.198,14.444);
curveto(-11.198,14.444,-11.794,13.153,-11.843,12.234);
Here's my attempt (which is not working, broken into lines for readability, this is a one liner):
s/
.*(-?[0-9]*\.?[0-9]*) (-?[0-9]*\.?[0-9]*) [a-zA-Z]$^(-?[0-9]*\.?[0-9]*) (-?[0-9]*\.?[0-9]*) (-?[0-9]*\.?[0-9]*) (-?[0-9]*\.?[0-9]*) y/
curveto(\1,\2,\3,\4,\5,\6);/
g
What's the correct way to do this?

Answer 1

For your problem, you could try something along these lines:
$NF == "m" { print "move(" $1 "," $2 ");" }
$NF == "v" { print "curveto(" one "," two "," $1 "," $2 "," $3 "," $4 ");" }
$NF == "c" { print "curveto(" $1 "," $2 "," $3 "," $4 ");" }
{ one = $(NF-2); two = $(NF - 1) }
$NF is the last field of each line and is used to select which transformation to apply. The two fields preceding the command are assigned to the variables one and two (x and y might be a better choice).