How can I use sed to generate an awk file? - awk

Question

How do I write sed commands to generate an awk file.
Here is my problem:
For example, I have a text file, A.txt which contains a word on each line.
app#
#ple
#ol#
The # refers when the word starts/ ends/ starts and ends. For example, app# shows that the word starts with 'app'. #ple shows that the word ends with 'ple'. #ol# shows that the word has 'ol' in the middle of the word.
I have to generate an awk file from sed commands which reads in another file, B.txt (which contains a word on each line) and increments the variable start, end, middle.
How do I write sed commands whereby for each line in the text file, A.txt, it will generate an awk code ie.
{ {if ($1 ~/^app/)
{start++;}
}
For example, if I input the other file, B.txt with these words into the awk script,
application
people
bold
cold
The output would be; start = 1, end = 1, middle = 2.

Answer 1

I'd use ed over sed for this, actually.
A quick script that creates A.awk from A.txt and runs it on B.txt:
#!/bin/sh
ed -s A.txt <<'EOF'
1,$ s!^#\(.*\)#$!$0 ~ /.+\1.+/ { middle++ }!
1,$ s!^#\(.*\)!$0 ~ /\1$/ { end++ }!
1,$ s!^\(.*\)#!$0 ~ /^\1/ { start++ }!
0 a
#!/usr/bin/awk -f
BEGIN { start = end = middle = 0 }
.
$ a
END { printf "start = %d, end = %d, middle = %d\n", start, end, middle }
.
w A.awk
EOF
# awk -f A.awk B.txt would work too, but this demonstrates a self-contained awk script
chmod +x A.awk
./A.awk B.txt
Running it:
$ ./translate.sh
start = 1, end = 1, middle = 2
$ cat A.awk
#!/usr/bin/awk -f
BEGIN { start = end = middle = 0 }
$0 ~ /^app/ { start++ }
$0 ~ /ple$/ { end++ }
$0 ~ /.+ol.+/ { middle++ }
END { printf "start = %d, end = %d, middle = %d\n", start, end, middle }
Note: This assumes that the middle patterns shouldn't match at the start or end of a line.
But here's a attempt using sed to create A.awk, putting all the sed commands in a file, as trying to this as a one-liner using -e and getting all the escaping right is not something I feel up to at the moment:
Contents of makeA.sed:
s!^#\(.*\)#$!$0 ~ /.+\1.+/ { middle++ }!
s!^#\(.*\)!$0 ~ /\1$/ { end++ }!
s!^\(.*\)#!$0 ~ /^\1/ { start++ }!
1 i\
#!/usr/bin/awk -f\
BEGIN { start = end = middle = 0 }
$ a\
END { printf "start = %d, end = %d, middle = %d\\n", start, end, middle }
Running it:
$ sed -f makeA.sed A.txt > A.awk
$ awk -f A.awk B.txt
start = 1, end = 1, middle = 2

Answer 2

Off the top of my head, and not tested:
/\(.*\)#$/s//{if ($1 ~ /^\1/) start++; next}/
/#\(.*\)$/s//{if ($1 ~ /\1$/) end++; next}/
/\(.*\)/s//{if ($1 ~ /\1/) middle++; next}/
The construct \(.*\) matches any text and saves it in a back-reference, then \1 recalls the back-reference. The empty pattern following the s command refers back to the pattern that matched the line. The next prevents the third pattern from matching after one of the other two has already matched.