> I'm very sorry to bother the list with this problem but I've been > searching in the web the couples hours to find an answer and still > haven't find any.
It's a common sed problem and in the FAQ http://sed.sourceforge.net/sed1line.txt (scan for the "emulates grep" bit) Fortunately, the vim list has a lot of friendly regexp wonks like myself :) > The problem is that I have a txt file of 3.5GB containing all the info > of Human chromosome 6. I want to save into one another file all lines > that have the pattern rs10946398 (occurring only ones). I know that vi > cannot handle files so big. I used ed in Fedora5 but this too cannot > stream it. I hope that grep or sed can do this but cannot figure how to. > I tried the following for sed but doesn't work: > > sed '/rs10946398/p' chr6.txt > > Can someone help? you should be able to use grep 'rs10946498' chr6.txt > out.txt or either of the sed variants: sed -n '/rs10946398/p' chr6.txt > out.txt sed '/rs10946398/!d' chr6.txt > out.txt If you misspelled and meant "occurring only once", you can then post-process the above with grep 'rs10946498' chr6.txt | grep -v 'rs10946498.*rs10946498' > out.txt Sed might allow it in one pass with something like sed -e '/rs10946398/!d' -e '/rs10946398.*rs10946398/d' chr6.txt > out.txt The first sed variant (-n) doesn't print anything ('-n') unless it matches the pattern. The 2nd sed variant deletes lines that don't match the given pattern ("!d") akin to a ":v" command in vim. I suspect that neither vi/vim nor ed streams the file and that both suffer similar problems in that they try to map the whole 3.5GB file into memory which is a cruel thing to do to an OS. :) If you just want lines with the pattern in it, I'd use the grep command. If you want lines with the pattern in it once and only once, I'd go with the last sed command. My $0.02 -tim
