I have one file like:
ID_SOUR_CALENDAR BIGINT NOT NULL DEFAULT 0 COMPRESS 0 ,
UNIQUE PRIMARY INDEX ( CALENDAR_DATE );
ID ,
ID_SOUR ,
PRIMARY INDEX ( CALENDAR_DATE );
I want to replace the ',' by ')' in the line just before the line containing PRIMARY.
The result should be:
ID_SOUR_CALENDAR BIGINT NOT NULL DEFAULT 0 COMPRESS 0 )
UNIQUE PRIMARY INDEX ( CALENDAR_DATE );
ID ,
ID_SOUR )
PRIMARY INDEX ( CALENDAR_DATE );
Using GNU sed:
sed 'N;s/,\(\s*\n.*PRIMARY\)/)\1/;P;D' file
ID_SOUR_CALENDAR BIGINT NOT NULL DEFAULT 0 COMPRESS 0 )
UNIQUE PRIMARY INDEX ( CALENDAR_DATE );
ID ,
ID_SOUR )
PRIMARY INDEX ( CALENDAR_DATE );
N Read/append the next line of input into the pattern space.P Print up to the first embedded newline of the current pattern space.D Delete up to the first embedded newline in the pattern space. Start next cycle, but skip reading from the input if there is still data in the pattern space.Try reversing the file, doing the string replacement on the line after PRIMARY, then re-reversing the file:
tac file | sed '/PRIMARY/ {n; s/,$/)/}' | tac
PRIMARY
sed would do this on its own for the positive offset (AFTER a certain line/regex). You tried to judge the solution as for some other use case, not the one at hand.
try this,
perl -0777 -pe 's/,([^\n,]*\n[^\n]*PRIMARY)/)$1/g'
This will replace the last , with ) in lines before any line including PRIMARY.
Explanation
perl -0777 slurp whole file (read as one line)s/search_pattern/replacement/g Substitute search pattern with replacement globally
. meta char won't cross line boundaries, unless the /s modifier is in effect. Hence, you can change the regex to s/,(?=[^,]*\n.*PRIMARY)/)/g
If you use ed rather than sed you can use a regex address with negative offset:
g/PRIMARY/-1 s/,$/)/
Ex.
$ printf 'g/PRIMARY/-1 s/,$/)/\n,p\nq\n' | ed -s file
ID_SOUR_CALENDAR BIGINT NOT NULL DEFAULT 0 COMPRESS 0 )
UNIQUE PRIMARY INDEX ( CALENDAR_DATE );
ID ,
ID_SOUR )
PRIMARY INDEX ( CALENDAR_DATE );
or (for in-place editing)
printf 'g/PRIMARY/-1 s/,$/)/\nwq\n' | ed -s file
The replacement will match ALL instances of /PRIMARY/ - if you only want to replace the first, remove the g modifier.
With POSIX-ly sed you can do the following:
sed -e '
/PRIMARY/!{x;1!p;d;}
x;s/,/)/;$G
' input-file.txt
Output:
ID_SOUR_CALENDAR BIGINT NOT NULL DEFAULT 0 COMPRESS 0 )
UNIQUE PRIMARY INDEX ( CALENDAR_DATE );
ID ,
ID_SOUR )
PRIMARY INDEX ( CALENDAR_DATE );
Working:
