Shell Programming and Scripting

You are right too, it may be better to cater for such a case even if it is not part of the specification.

It could be mitigated in a way similar to what is built into your suggestion:

Code:

awk '$1=="select"{if(p)print ";"} {p=$NF!~/;$/}1' RS= ORS='\n\n' file

Thank you so much Rudic and Scrutinizer. Both your solutions worked for me. What Rudic said is right if the previous line terminated with 1 then it would replace that with semi colon. I hope that I don't have such a condition in the rest of my files where I want to do this search and replace.
My goal is that I want to separate each SQL statement in a file by a semi colon. In my SQL I don't have appropriate semicolons separation for all the SQL statements. My end goal is to run this SQL file via Teradata Bteq utility successfully.

This sed script prints the semicolon before the select if the preceding line is empty (there are two consecutive newlines)

Code:

sed '
/^select/{
x
/^$/i\
;
x
}
h
' file