sed: print section of file from string to end of file

samyboy · 02-07-2008, 09:02 AM

Hi,

I've got a file with garbage (empty lines, weird chars) before the real content of my file.

In this case, I am deleting everything before the '<?xml' string.

My way is too much complicated because it uses three commands and a temp file ...

sed -n '/<?xml/,$p' file.xml > file.tmp
sed 's/^.*<\?xml/<?Xml/1' file.tmp > file.xml
rm -f file.tmp

Any knowledge to change it into one command ?

Thanks.
Sam

druuna · 02-07-2008, 09:59 AM

Hi,

Is this what you want/need:

sed -i -n '/<?xml/,$p' infile

This will edit the infile directly (the -i option) and print everything from the first <?xml to the last line (/<?xml/,$p). The -n will only print that what is wanted.

If you do want a backup of the original file you can extend the -i option. -ibak will generate a <file>.bak file and a -iorg will create a <file>.org file.

BTW: The -i option is only present from sed version 4.0 and higher.

Hope this helps.

samyboy · 02-08-2008, 08:19 AM

Quote:

Originally Posted by druuna

sed -i -n '/<?xml/,$p' infile

Hi,

thanks for your answer. However, the line you mentionned is the same as my first line.

I still need to merge it with this one below which "cleans" the garbage before the "<?xml" string (at the same line):

sed 's/^.*<\?xml/<?Xml/1' file.tmp

druuna · 02-08-2008, 08:40 AM

Hi,

Ok. Well in that case your initial set of commands aren't that complicated. I would personally re-write it to the following:

$ sed -n '/<?xml/,$p' infile | sed 's/^.*<\?xml/<?Xml/' > infile.tmp
$ mv infile.tmp infile

Too my knowledge there's no way to make a more compound sed (or other prog like awk) statement. But I could be wrong.....

Oh, you could make this a 'one-liner':
$ sed -n '/<?xml/,$p' infile | sed 's/^.*<\?xml/<?Xml/' > infile.tmp && mv infile.tmp infile

I used '&&' between the commands instead of a ';' to make sure that the move part is only done if the previous command succeeds.

samyboy · 02-26-2008, 07:23 AM

Nice thanks!