A conundrum for a sed wizard

rather than try to convince sed to match your corrupt line via regexp, just use awk to split the file into 2 pieces - the lines before the corrupt line (into file A) and the lines after the corrupt line (into file B).

Then cat the 2 files together, with your fixed line in the middle:

    cat A > fixed.xml
    echo "my fix" >> fixed.xml
    cat B >> fixed.xml

Works fine for me:-

    $ cat infile.xml
    1
    <load address=’11c1�����ze=’08′ />
    3
    $ cat seddy.sed
    2s@^.*$@<load address=’11c1385b’ size=’08’ />@
    p
    $ sed -n -f seddy.sed infile.xml
    1
    <load address=’11c1385b’ size=’08’ />
    3
    $

So I'm guessing it's locale related (LC_CTYPE is en_GB.UTF-8 for me), or the string that forms his corrupt line is different from what was pasted. The output of:-

    sed -n '35185222p' infile.xml | hexdump -b

would help in that regard.

This just looks painful. Not the problem, the 35 million line XML document. Which probably will not validate after eating that line.

None the less. I'm sure the constants are a bit off here.

head -35185222 infile.xml > OhGodThisIsHorrific.xml

tail -n+35185224 infile.xml >> OhGodThisIsHorrific.xml

I'm guessing a 35 million line file, is like, what, enterprise Java version of hello world? (just kidding, sorta)

(Edited to fix the line spacing. Also the double greater than is not a typo)

This seems like a great question for StackOverflow.

Are you using unicode handling?