| |
| |
| |
|
Page: 1 2 3 4
Comments:
<0> Hi
<1> hey folks
<0> is it possible to replace a string in a file without losing the tabulation ?
<1> i'm trying to apply a block filter, but I only need to apply to the first block it matches, ignoring the rest that could match
<2> Lutin: yes
<1> Lutin: you can limit the regex using \t
<1> is that possble ?
<0> csmanx, ok
<2> csmanx: Yes.
<2> csmanx: A less elegant method is to simply print the rest of the input in a loop: :a n ba
<2> sed -e '/foo/,/bar/{ s/xxx/yyy/g; /bar/{ :a' -e 'n;ba' -e '}' -e '}'
<1> i don't get it after the second /bar/ =/
<2> It's a loop. : is a label, n prints and reads, b jumps to the label.
<1> right
<2> So, when it finds the end of the first block, it simply reads and prints all of the remaining lines.
<1> too many '
<1> prec: you mean, when it finishes processing the first block
<2> No, there's just enough '
<2> ;)
<2> csmanx: Right.
<1> is -e acting recursively ?
<2> No, sed takes all the -e options and concatenates them into a single script which it then executes.
<2> printf '/foo/,/bar/{\n\ts/xxx/yyy/g\n\t/bar/{\n\t\t:a\n\t\tn\n\t\tba\n\t}\n}\n' >csmanx.sed
<2> sed -f csmanx.sed
<1> =)
<1> prec: i'm intrigued about the -e decoration
<1> what is the second one for in the :a loop?
<2> ?
<2> All the -e options are concatenated into a single script.
<2> The : and b commands are terminated by the end of the line, hence you cannot (portably) use: sed -e ':a;n;ba'
<1> prec: right , but they seem not be executing anything ' -e ' ?
<2> What's your script look like?
<1> prec: not refering to my script, but to the example here
<1> =)
<2> It works for me. Are you using an sh-compatible shell?
<1> so :a starts the loop , but '-e' does what?
<2> -e doesn't do anythinig
<2> -e is not part of the script
<2> -e is how you specify the script.
<2> In my example, there are 8 command line arguments to sed.
<1> i thought it meant "execute this"
<2> They consist of four -e script pairs.
<1> i see
<2> All four option arguments are concatenated into a single script _before_ sed starts its execution phase.
<2> sed -e 's/1/2/g' -e 's/3/4/g'
<2> sed -e 's/1/2/g;s/3/4/g'
<1> that one i know
<2> sed 's/1/2/g;s/3/4/g'
<1> right
<2> Those are all the same.
<1> but going back to the question, sed doesn't have an explicit n blocks to process switch, right?
<2> Nope.
<2> By "block", you mean some address range, right?
<1> yeah
<2> OK.
<1> if i have <!-- <thing_to_enable ... /> -->
<1> like five times
<1> and I just want to remove the comments from the first block because I only need 1
<1> btw, removing --> just one time doesn't work =(
<2> ?
<1> it removes all of them
<1> from all matching blocks
<2> What's your script look like?
<1> sed -e '/<element_to_enable/,/-->/{s/-->//g;/-->/{:a' -e 'n;ba' -e '}' -e '}'
<2> OK, the problem is that the /-->/ is never matched because it occurs on the same line as the start address.
<1> it is matched and removed from all blocks, i don't know about the loop part
<2> OK, it is matched by the s command, yes.
<2> But it is never matched by the second address in your address range.
<1> right
<1> what would be the start address there?
<1> the one that is gone?
<1> right
<1> =)
<2> seq 19 22 |sed -e '/2/,/0/!d'
<2> vs.: seq 19 22 |sed -e '/0/,/1/!d'
<2> In other words, the block does not end until the _next_ line which matches the end address.
<2> brb
<1> the next line that matches, not the next match in the same line
<2> csmanx: Exactly.
<2> However, we can fix your script.
<2> /blah/,/-->/{/-->/{s///;:a n ba } }
<2> sed -e '/blah/,/-->/{/-->/{s///;:a' -e'n;ba' -e \} -e \}
<1> hmmm
<2> Perhaps I should test that ...
<1> if --> matches do the matching again and if it matches replace with nothing and print the rest ?
<1> lemme try
<1> it worked
<1> but i got the stuff printed twice
<2> Yeah, it works here.
<2> What stuff?
<1> the blocks
<1> echo '<!-- <element_to_enable /> --> <!-- <element_to_enable /> -->' | sed -e '/<element/,/-->/{/-->/{s///;:a' -e'n;ba' -e \} -e \}
<2> They don't print twice for me.
<1> huh?
<1> geez...
<1> heh, my devel box is broken then
<1> GNU sed version 4.0.7
<2> I've got 4.1.4 for some reason.
<1> now you freaked me out with the \} and the not-really-unmatched ' =P
<2> But it is pretty standard sed.
<1> hmmm
<2> Hmm? I've got two pairs of '' in my last example.
<1> yeah, i'm just freaking out with { syntax
<2> I get this output: "<!-- <element_to_enable /> <!-- <element_to_enable /> -->"
<1> right
<1> i get the same on my mandriva laptop
<1> but not on the rhel 3
<3> how would I remove the abc from this string? "<P***word>abc</P***word>"
<2> s/abc//
<4> heh
<3> heh
<2> <input xml2pyx |sed '/^(P***word/,/^)P***word/d' |pyx2xml >output
<2> <input xml2pyx |sed '/^(P***word/,/^)P***word/{/^-/d;}' |pyx2xml >output ## this leaves the element, but removes any CDATA content.
<3> I don't have xml2pyx and can't install it. I'd also like to be able to change abc to things other than blank.
<2> cekc: You will need to use an XML parser (easy) or write an XML parser in sed (very difficult).
<2> I have a version of "xml2pyx" written in sed.
<2> However, I don't have pyx2xml yet, so ... ;)
<3> why must I use an XML parser? Is it the < and > characters?
<2> Because you are parsing XML, yes?
<2> By definition, you need to use an XML parser to do that.
<3> I am sanitizing p***words from an XML file, and the p***words will always be between the strings <P***word> and </P***word>
<3> I'd be just as happy removing BBB from AAABBBCCC where BBB is variable and AAA and CCC are known
<2> [^<] might come in handy then.
<3> I have to take off now, might be back on later to continue figuring it out. Thank you very much for the help
<5> hi there
<5> anyone up?
<5> :)
<1> hello again =)
<6> i'm trying to match lines which contain either TE or TA: echo TAST | sed -ne '/(te|ta)/I p'
<6> but the (te|ta) does not seem to work
<6> any hint?
<7> veal: echo TAST | sed -ne '/te\|ta/I p'
<6> thank you!
<7> veal http://www.student.northpark.edu/pemente/sed/sed1line.txt AND sed.sf.net
<1> cl***ic mistake
<1> my just-two-line matcher doesn't work =(
<1> a simple s/foo\nbar/boo/ won't work
<1> =(
<7> csmanx: edit between regexs: sed '1,/foo/!{ /bar/,/foo/!s/^/added text /; }' filename
<1> gnubien: what is the 1 for ?
<7> csmanx: try it without the 1 to see the diff
<1> ok
<8> wouldn't it be easier to use sed '/foo/{N;s/foo\nbar/boo/}' file
<8> ?
<7> helloman: nice
<1> helloman: exactly
<1> helloman: i just go that
<1> by myself ^_^
<1> but now, i just want to replace the foo
<8> when there is bar on the next line?
<1> hmmm the boo should work
Return to
#sed
or
Go to some related
logs:
mkpart =-1s
#openzaurus
nachplus
MAXVOLRAT
exception in __init__ python
ubuntu no sda1
wrong current_timestamp mysql
#css
suse kdrc default
slashcot
|
|