Shell Programming and Scripting

I've got a report I need to make easier to read Using sh on HP-UX 11.12.

In short, I want to search for a regular expression and when found, examine the next line to see if it's blank. If so, then delete both lines. If not blank, move on to the next regexp. Repeat.

So far I've got:

Code:

sed -n -e '/[0-9]-[0-9][0-9]/p' infile >outfile

That finds my regular expression (which is a number by itself on one line and usually looks like 123456-99). I cannot seem to script the next part.

Any help is appreciated. Thank you.

how about this:

Code:

sed '/[0-9]-[0-9][0-9]/{N;/\n$/d}' infile > outfile

Hello Scottie1954 and welcome to the forums.

Perhaps this will do the trick for you:

Code:

sed '/PATTERN/{$q; N; /\n$/d;}'

That will print all lines until it finds pattern. When a line matching PATTERN is found, instead of printing it, the next line is read and appended. We then test to see if the text in the pattern buffer ends with a newline. Since sed never includes the newline from the text that's read from the file, the only way the newline could have gotten there is as a result of the preceding N command. That newline, if at the end of a line, indicates that the last line read was empty. If this condition is met, the d command discards the current pattern buffer (which at this point contains those two lines, one of them empty) and a new cycle begins with the next line in the file.

The $q command is there to immediately print and exit should the last line in the file match PATTERN. While there is no point in checking for a blank line when there are no further lines, the primary reason for this check is that the N command will not output the pattern buffer if there is nothing left to read from the file (if the last line of the file matches PATTERN, without this check it would be lost).

Regards,
Alister

---------- Post updated at 05:45 PM ---------- Previous update was at 05:43 PM ----------

I see that my archenemy, Chubler_XL beat me to it.

I'll leave mine intact for the explanation and $q difference.

Regards,
Alister

Thank you, Chubler_XL and Alister, for replying. I've tried both but hp-ux is having some difficulties.

For Chubler_XL's code, sed tells me:

Code:

sed: Function /[0-9]-[0-9][0-9]/{N;/\n$/d} cannot be parsed.

And for Alister's example:

Code:

sed: /[0-9]-[0-9][0-9]{$q; N; /\n$/d;} is not a recognized function.

I can understand Alister's explanation (that's been the clearest I've found about looking for the newline character), but I'm afraid I don't have enough sed experience to troubleshoot its different flavors among operating systems.

Anyway, if you've got a guess about what's making hp-ux cranky, I'd appreciate it. I'll keep experimenting myself.

~Scottie1954

@Alister - Nice explination and good pickup on pattern on last line of file

Scottie1955 - You could try using awk:

Code:

awk '/[0-9]-[0-9][0-9]/ { K=$0; next }
K { if(length) print K "\n" $0; K=""; next }
1
END { if(K) print K }' infile

Whoa! That worked. Thanks, Chubler_XL.

Chubler_XL's suggestion is missing a semicolon. To be posix-compliant, there should be a semi-colon after the d command:

Code:

sed '/[0-9]-[0-9][0-9]/{N;/\n$/d;}' infile > outfile

With regard to my example, it seems you may have mistyped it. In your error message, there is no slash delimiting the end of the pattern (what seems to have been omitted is in bold red):

Code:

sed '/PATTERN/{$q; N; /\n$/d;}'

Regards,
Alister