| |
| |
| |
|
Comments:
<0> redduck666: your script doesn't work if more than one whitespace characters are between the first and the second word
<1> [ ][ ]* looks like it will work.
<1> What it won't handle is the case where the first word has only one letter. ;)
<2> 23:16:06 ~ $ echo 'word1 word2 word3' | sed 's/[ \t]*\([^ \t]\)[^ \t]*\([^ \t]\)[ \t][ \t]*\([^ \t]\).*/\1\2\3/'
<2> w1w
<2> goedel: now it works :)
<3> oh, i was afk, sorry.. redduck666, thank you, it works
<0> redduck666: sorry to disappoint you; it doesn't work if the first word consists only from one character
<0> s/from/of/
<1> hehe.
<0> my script is also not perfect: http://pastebin.com/870710
<0> perhaps better: awk '{print substr($1,1,1) substr($1,length($1)) substr($2,length($2))}'
<0> but OT
<3> goedel, thanks alot
<0> rew: you're welcome
<0> good night; time too sleep
<4> I have a bunch of CSS: I want to print the selector when the block contains certain text. The css is laid out like 'selector {\n\tproperty:\tvalue;\n\tnextproperty:\tnextvalue;\n...\n}
<4> I'd like to print out 'selector {' if the junk between { and } contains 'float'
<4> I guess I can do it in a couple pipes, but I wonder if there's a better way
<5> I should learn how to do this in sed.
<5> I can do it in awk.
<5> awk '/[^{]*{/,/}/(/[^{]*{/){a=$0}(/float/){print a}'
<4> what am I not understanding such that sed -n '/foo/,/bar/p' -e '/baz/p' file fails?
<4> I mean, I know it doesn't do what I want, but why does it error out and say "can't read /foo/,/bar/p, no such file or directory"
<4> ohh
<4> need another -e
<5> Ah right.
<4> too bad you can't just do /{/,/float/,/}/p/
<4> er, you get what I mean
<5> Yeah.
<5> Well, it involves the Hold Space
<5> and I've never gotten around to practising it
<5> sed -n '/.*{/,/}/{/.*{/{h};/float/{g;p}}'
<5> you can exclude the {} surround the h
<5> Maybe one of those sickos, prec or BearPerson can show a better way :-)
<6> hmm
<5> hey man
<6> so basically, you want "if between 'selector {' and '}' there is the substring 'float', print the string 'selector {'" ?
<6> or do you want to print the entire selector{} block if and only if it contains 'float' somewhere?
<5> selector {
<6> sed '/selector {/,/}/ {/float/ i selector {; d}'
<4> er
<4> but I don't know what 'selector' is
<4> or am I not understanding?
<6> plus proper syntax
<4> BearPerson: something like sed '/^\([^{]*\){/,/}/ {/float/ i \1 {; d}' ?
<6> I don't think // groups can capture
<6> kojiro, ah, so it's not the string 'selector', it's "the string in front of the {" ?
<4> BearPerson: yeah, and I just discovered it's more complicated, some of these CSS have multi-line selectors
<6> hrm
<4> a,\nb,\nc {...
<4> Someone suggested I use cscope
<6> where does a selector start, then?
<4> BearPerson: anything that's not between { } or /* */ is a selector
<6> hrm
<6> yes, a css parser would work wonders
<4> well, cscope is a C parser, isn't it?
<4> I haven't been able to find a decent CSS parser yet
<5> make one!
<5> There's probably a perl module to do it.
<4> goldfish: no kidding, I should make one
<5> heh
<4> With sed! (not)
<5> The other one works, but not for multi-line selectors
<4> I suppose the simplest (not the most efficient) way to do a parser would be to clear all newlines and create a literal hash
<4> CSS really is just a hash of hashes
<6> yeah, that's my favorite last-measure sed technique, though it is highly mem-inefficient
<6> run stuff through tr '\n' '\a' and work on that
<6> nowadays, I usually prefer smart usage of the 'N' command
<6> sed -n '/{/,/}/ ! h ; /{/,/}/ {/float/ {s/.*//; x;p}} ; /}/ {s/.*//;h;}'
<6> that might work somewhat rudimentarily
<6> hrm, not quite, actually
<6> sed -n '/{/,/}/ ! H ; /{/ H; /{/,/}/ {/float/ {s/.*//; x;p}} ; /}/ {s/.*//;h;}'
<6> I'll probably get killed for that line if anyone who knows sed a bunch reads it, but it might work somewhat ;)
<5> is there way to empty the current hold space?
<5> ah.
<5> {s/.*//;h;} is how you achieve it?
<6> aye
<6> put something empty into the hold space
<5> cool
<5> thanks
<5> maybe they ran out of letters when they thought about adding that feature :)
<6> gnu features are interesting
<6> normally, extra nonstandard features are a standard method of vendor lock-in
<6> I wonder if they're trying to achieve the same effect ;)
<5> heh
<7> hello
<5> hi
<7> i'm trying to filter binary files for certain patterns. the main problem i'm facing is that sed/grep seem to be line orientated, so i always get the whole line where my target expression resides. i just want to get the exact sequence of characters matching a certain regex, can sed do this for me?
<1> Of course it's easy when this sequence does not span a newline.
<1> sed and grep work with text files however, not "binary" files.
<1> Consider converting your binary file to a text file, then using sed, then converting it back (if necessary).
<7> i've tried 'grep -a'
<1> grep is _still_ a tool to work with text files.
<7> and converted with hexdump .. but it gets worse and worse.
<1> grep will only print the lines which match.
<1> If your input file does not consist of lines, then this isn't useful for you.
<1> text files consist of lines of text, and so text-oriented tools are line-oriented.
<7> is there a tool capable of returning just the matching string?
<7> string meaning lower ascii range
<1> grisu: Are you searching for strings in binary files? You may want to use 'strings' then grep its output.
<7> okay.. that would be worth a try
<1> strings some_executable_image |grep pattern
<1> strings has special functionality when it works with object files and executables.
<7> oh.. that looks much better now :-)
<7> this should be easy from this point on
<7> well, just that i didn't figure how to make grep/sed/awk returning just the match and nothing else (no characters before and after)
<1> I know GNU grep has an option for this.
<1> sed can do it too ...
<1> http://rafb.net/p/gxaBoP87.html
<7> i had "sed -n '/expression/p'" but it didn't really do what i want
<7> -n '/expr/&/p' even
<7> yes, it works - but is there away around this multiline solution?
<7> maybe another tool which trims the output to just the matching region?
<7> when grep is capable of colorizing the matching portion.. why shouldn't it be capable of printing just that?
<8> grisu, "man grep" should tell you.
<7> -o"hhh" :-)
<7> thanks prec and Rado!
Return to
#sed
or
Go to some related
logs:
#debian
at line 2: Duplicate entry 'localhost-mail' for key 1
ubuntu php5 mssql.so
readlines remove \n
Kernel panic - not syncing: Attemted to kill init
setting sensor limits fail
#debian
Error inserting scsi_transport_spi
confog cah
centos pam_mysql
|
|