Shell Programming and Scripting

I need to reduce a file's size below 50MB by deleting chucks of text. The following sed does this.

Code:

sed '/^begpattern/,/endpattern/d' myfile

However, it's possible that the file size can get below 50MB by just deleting the first instance of the pattern. How do I code that into sed?
Or can awk do this?

Yes awk can do this:

Code:

awk '!s&&/^begpattern/ {s=++d} !d ; /endpattern/ {d=0}' infile

I've never seen this construct before. Can you explain what's going on?

OK here is the explination:

s variable is the started flag, and the statement !s&&/^begpattern/ will only match a begpattern if s if false (unassigned).

s=++d increments d (our delete flag) from unassigned to 1, and also assigns s to this incremented value ie 1.

!d only print a line if d is false/zero.

/endpattern/ {d=0} reset d (delete flag) back to false/zero when endpattern is seen.

I hope someone can comes up with a sed solution for you, I spent ages mucking around with sed trying to get it done, but my sed skills aren't up to it

Code:

 xSED=`sed -ne '/begpattern/,/endpattern/{=;!p;{/endpattern/q;}}' infile|sed -n '1p;$p'|sed 'N;s/\n/,/'` ; sed "$xSED d" infile

regards
ygemici

Yikes, need to run sed 3 times! This is the closest I got but 1st endpattern is still printed:

Code:

sed -n -e '1,/^begpattern/{/^begpattern/!p;};' -e '/endpattern/,$p'

My sed skills are now up to solving this, I knew it should be possible to do it in 1 pass:

Code:

sed -n -e ':b' -e '/^begpattern/bs;p;n;bb' -e ':s' -e '/^endpattern/br;n;bs' -e ':r' -e 'n;p;br'