Extract from stdout the string in [ ], in a nicely way

frenchn00b · 11-21-2007, 05:07 PM

cat myfile.txt

Code:

    * [82] Download the cdrom [83] Linux Ubuntu Forum

I would like that a script i.e.

Code:

mysuperscript.sh myfile.txt

simply just outputs all intances that are between [ ]:

Code:

82 
83

I am using awk, but that is not really easy with it. I like awk since it is much easier than C, perl, and python. ... but that is not made for that at all I guess

Has someone an idea ?
thank you

theNbomr · 11-21-2007, 06:11 PM

Not awk, but a one-liner at least:

Code:

cat myfile.txt | perl -e 'while (<>){ print join "\n", $_ =~ m/\[([0-9]+)\]/g; }'

--- rod.

bigearsbilly · 11-22-2007, 02:31 AM

my go

here's a little script,

Code:

#!/usr/bin/perl

$/ = undef;             # go into slurp mode into the array @L
$slurp = <>;            # slurp the file

@L = $slurp =~ m/\[(.*?)\]/g;   # get the match

$" = "\n";              # set array separator which works if you...
print "@L\n";           # print array in quotes

makyo · 11-22-2007, 07:50 AM

Hi.

With standard commands:

Code:

#!/usr/bin/env sh

# @(#) s1       Demonstrate extraction of bracket-bounded numeric string.

set -o nounset
echo

debug=":"
debug="echo"

## Use local command version for the commands in this demonstration.

echo "(Versions displayed with local utility \"version\")"
version >/dev/null 2>&1 && version bash grep sed

echo

echo '* [82] Download the cdrom [83] Linux Ubuntu Forum' >data1
FILE=data1

grep -E -o '\[[0-9]+\]' $FILE |
sed -e 's/\[//' -e 's/\]//'

exit 0

Producing:

Code:

% ./s1

(Versions displayed with local utility "version")
GNU bash 2.05b.0
grep (GNU grep) 2.5.1
GNU sed version 4.1.2

82
83

See man pages for details ... cheers, makyo

radoulov · 11-22-2007, 08:00 AM

Corrected (missread the question):

Code:

set -- $(grep -oE '\[[^]]+\]' filename)
printf "%s\n" "${@//[\[\]]}"

zsh:

Code:

set -- $(<filename)
print  "${(F)${(Mz)@#*\[*\]}//[\[\]]}"

jschiwal · 11-22-2007, 08:17 AM

How's this for a cheat:

Code:

sed 's/[^][:digit:][]//g' test

It assumes all square brackets & numbers are in the pattern to keep. However, sed is only 1/24th the size of perl, and 1/3rd the size of grep.

The grep solution gets my vote.

radoulov · 11-22-2007, 08:32 AM

If the pattern is always numeric (without the brackets,
as the OP wants)

Code:

awk NF RS='[^0-9]]*'

Otherwise:

Code:

awk '$0=$2' RS="]" FS="["

frenchn00b · 11-22-2007, 12:48 PM

Amazing replies !! Amazing

PAix · 11-22-2007, 01:07 PM

Radoulov,

Code:

#1
awk '$0=$2' RS="]" FS="{" < test.txt

I see it works, but don't understand how. Can you explain at all please?

My difficulty: I see the primes (') which normally delimit an awk program. So it's essentially

Code:

#2
 '$0=$2'

If I run this on it's own, I can see that it is the complete program and what it does in the default environment; it takes the second field and makes it into the total field and prints it ( the default action - in the same way that print infers print $0 ).
Now because the RS="]" FS="[" are outside the program (I would normally have expected to see this type of assignment inside a BEGIN block)[code]

Code:

#3
awk 'BEGIN { RS="]";  FS="[" }
$0 = $2' < test.txt

, I expect that they are being passed as parameters and would have expected to see

Code:

#4
awk  -v FS="[" -v RS="]" '
$0 = $2' < test.txt

, which without the -v will throw a syntax error.

Well, Radoulov, I have to thank you for the explanation thus far, but here comes the bit I couldn't work out; what makes putting the command line assignments(?) after the program and before the redirected stdin input work?

I hope that this discussion will also be of use to others that hadn't quite thought about the subtleties implicit in your VERY NEAT bit of code. Please share your magic with me/us.

Personally (before today) I would have written the code as shown in #3 above, had I been smart enough to work it out for myself. Honest, I would prefer to leave you guessing

. The version at #4 I might have used if it had occurred to me that It was valid to pass values to the awk built-in variables. It didn't although I am otherwise familiar with parameter passing in this manner.

So a super thread at several levels with lots of learning content. Well done everyone.
Sorry Bigears, I'm just an awk kinda guy. Perl is just too rich for me. One day perhaps, but don't hold your breath.

angrybanana · 11-22-2007, 01:27 PM

Another one line Perl solution.

Code:

perl -lne 'print for /\[(\d+)\]/g' file

frenchn00b · 11-22-2007, 01:34 PM

Code:

| grep -o 'http:[^"]*'

this thing, for instance, I am not really understanding

^ is for begin of line

-o, --only-matching
Show only the part of a matching line that matches PATTERN.

* to say whatever

[ ] and ", not idea

frenchn00b · 11-22-2007, 01:36 PM

Now my final script looks like this:

I am not sure if it is "nicely"

Code:

#!/bin/sh
cat do.html | grep MYSTRING1 | grep MYSTRING1 |  awk '$0=$2' RS="]" FS="["  > "/tmp/.TMP1.txt" 
for each in $(cat "/tmp/.TMP1.txt" ) ; do
	cat do.html | grep "^  $each." -A1 | grep "MYSTRING3" grep -o 'http:[^"]*'
done

druuna · 11-22-2007, 01:37 PM

Hi,

@PAix:

awk '$0=$2' RS="]" FS="[" infile

and

awk -v RS="]" -v FS="[" '$0=$2' infile

Are both according to the command line syntax rules: From the Sed & Awk manual:
awk [-v var=value] [-Fre] [--] 'pattern { action }' var=value datafile(s)

radoulov · 11-22-2007, 02:24 PM

Hi,
@PAix:

Yes,
as already stated, the variable assignment variable=text
is part of the Awk syntax and its usage may depend on
personal style.
But note that -v variable=text is not the same as variable=text,
for examle:

Code:

zsh-4.3.4-dev-2% awk -vv=x 'BEGIN{print v}'
x
zsh-4.3.4-dev-2% awk 'BEGIN{print v}' v=x

zsh-4.3.4-dev-2%

or:

Code:

zsh-4.3.4-dev-2% awk -vv=x 'BEGIN{print v}{print v;exit}' v=z <(yes)
x
z

Another important point is that
older awks don't support the -v syntax.

Anyway,
my Awk examples were kind of joke:
one doesn't have to use Awk for such trivial
task, like Ed Morton says it's like
suggesting a pneumatic drill, when it's posisble
to break it up with a toothpick.

PAix · 11-22-2007, 07:08 PM

Hi Radoulov,

Thank you for taking the time.
It took a moment, but yes, all understood, the post-code assignment is only good where there are instances of data read in.
My sed is weak (working on it) and my awk/nawk (Solaris) is probably a little dated.

Re the pneumatic drill/toothpick anology, when you have a hammer, every problem is a nail!

Joke or not, it served a very useful update for me. It's appreciated.