exporting variables from awk

[Nicolás]

Hi there!

I'm trying to split a string in an unknown number of strings separated by spaces.

I'm trying with awk, but i can't export variables back to shell.

this is what i have, it doesn't work:

str='one two three four five six'
i=0
awk -v string="$str" 'BEGIN { split(string,a); for(i in a) print a[i]; }' | \
while read name
do
filename[$i]=$name
echo ${filename[$i]}
let i=$i+1
done


I know i can append awk output in a file and then do
cat filename | \
while read name
....

but i don't want to use external files.

i also know about variable=`awk statement`
but this lastone can't return me more than one variable.

any idea?

[Dennis Handly]

>but i can't export variables back to shell.
>it doesn't work:

You don't explain why it fails and what you want to do?

>I know i can append awk output in a file and then do

This is not much different than the pipe except you don't need a name.

>i also know about variable=$(awk statement)
but this last one can't return me more than one variable.

It returns you a string. Which you can fiddle with.

The standard trick is to return a command and use eval on it:
eval $(echo export x=1 b=2)

>str='one two three four five six'

You could use:
set -A filename $str

Or you could do:
for i in $str; do
echo $i
done

[Mark Grant · Accepted solution]

Firstly it is not possible for anything to export variables to the parent shell (at least not without being extremely "hacky" about it).

To me, I think perl is your answer here but if you don't like it perhaps it would be simpler to change the "while" to a "for" loop and cycle through the output like this.

for $result in `awk -v string="$str" 'BEGIN { split(string,a); for(i in a) print a[i]; }'`
do
filename[$i]=$result
etc etc etc
done

You might want to look at the venerable "expr" with a nice regular expression command instead of awk for something like this too.

[Ralph Grothe]

If you only need to populate an array of whitespace delimited tokens,
why the "awk" awkwardness?
The Posix shell can handle this

$ myvirtues="lust gluttony greed sloth wrath envy\npride"

$ set -A sins $(echo $myvirtues)

$ echo ${#sins[*]}
7

$ echo ${sins[*]}
lust gluttony greed sloth wrath envy pride

[James R. Ferguson]

Hi:

Adding a bit to Ralph's virtues:

# myvirtues=" myvirtues="lust gluttony greed sloth wrath envy\npride\nimpatience\nlaziness\nhubris"

# echo ${myvirtures}|xargs
lust gluttony greed sloth wrath envy pride impatience laziness hubris

Regards!

...JRF...

[Nicolás]

Ralph and James:
that works great on ksh, but unfourtunly i'm working on bash.

i'll explain what i need
what i have to do is to look for a file and do something with it.
points are:
-I've to search in many directories
-I'll probably found not one but two, three or four files, with the same name but diferent extension. and perhaps some completly similar files in diferent directories
-I have to chose only one of those, which one? the one has not null lenght (wc -l) and the one has the higher value in a list of extension priorities

so, my idea was to loop the dirs, and make finds with the only file mask for any dir. then get the find result in a variable, split it, and make stuff.

but perhaps that's not the best idea.
time ago "somebody" used to do it like this:

file_1=`echo $fullname | cut -d' ' -f1`
file_2=`echo $fullname | cut -d' ' -f2`

but he didn't concern into the posibility have more than two filenames in the string.

[Dennis Handly]

>that works great on ksh, but unfortunately i'm working on bash.

(Scripts should be in sh or ksh.)

>file_1=`echo $fullname | cut -d' ' -f1`
>but he didn't concern into the possibility have more than two filenames in the string.

If they are whitespace separated then "for" or arrays should automatically separate them.

[Ralph Grothe]

Hi Nicolas,

is your bash run on Linux or HP-UX?

Btw, bash can equally well handle arrays
like hpux-sh or ksh.
It only has a slightly deviating syntax
(a peek at man bash always helps ;-)

Because we haven't installed Bash on our HP-UX boxes (simply no need for it there)
I am running this on Linux.

$ uname
Linux


$ echo $BASH_VERSION
3.1.17(1)-release


To stick with James' extended set of virtues
(btw. I would have thought that laziness was covered by sloth, and hubris by pride?)

$ myvirtues=" myvirtues="lust gluttony greed sloth wrath envy\npride\nimpatience\nlaziness\nhubris"

as I'm doing this on Linux I need to cater the escape option along with the echo command (omit on hpux)

$ sins=($(echo -e "$myvirtues"))

$ echo ${#sins[*]}
10

$ echo ${sins[*]}
lust gluttony greed sloth wrath envy pride impatience laziness hubris


But this doesn't help you much, I suppose?

Apropos, the GNU find available on Linux
has a nifty -print0 option,
which separates the output records by \0
so that find can also cope with (Unix)unusual file names that contain whitespace etc.

[James R. Ferguson]

Hi (again):

And I would add that the use of 'xargs' applies to the Bash environment too:

$ virtues=" laziness\timpatience hubris\nare\nvirtures"
$ echo "${virtues}"
laziness impatience hubris
are
virtures
$ values=$(echo "${virtues}"|xargs)
$ echo ${values}
laziness impatience hubris are virtures

By the way, I teasingly used "hubris ~ pride" and "laziness ~ sloth" as Ralph pointed out, since these are Perl virtues he appreciates too :-)

Regards!

...JRF...

[James R. Ferguson]

Hi (again):

> Ralph: Apropos, the GNU find available on Linux has a nifty -print0 option, which separates the output records by \0 so that find can also cope with (Unix)unusual file names that contain whitespace etc.

HP-UX introduced this to 11.23 with patch: PHCO_32269.

The 'find' command for 11.31 supports this option, too.

Regards!

...JRF...

[Nicolás]

Ralph,

i'm trying it on linux, but it must work on HP-UX at work.

Yes, i have learned to handle arrays in bash. well, i'm learning it right now, trying these stuff

matter is that i have resolved in a simple way:

would it work on HP-UX?


for dir in $directories
do
echo "Searching in $dir" >> $log
find $dir -name "$filemask*" 1> fullnames 2> /dev/null
#cat fullnames
if [ -n `cat fullnames` ]; then
break;
fi;
done
i=0
flag=0
echo
while read fullname
do
echo "Checking for posible duplicated file" >> $log
for (( e=0 ; e <= i ; e++ ))
do
echo "checking file $i : $fullname iteration number: $e" >> $log
if [ "$fullname" = "${fullnames_array[$e]}" ];
then
echo "$fullname number $i equals another, it's Discarded" >> $log
flag=1
else
echo "$fullname ins unique!" >> $log
fi
done
if (( flag == 0 ))
then
fullnames_array[$i]=$fullname
i=$(( i + 1 ))
#echo ${fullnames_array[*]}
#echo "print $fullname"
fi
done echo "found files: ${fullnames_array[*]}" >> $log

[Dennis Handly]

>would it work on HP-UX?

This works in ksh. You have to replace that funny for loop by a while:
for (( e=0 ; e <= i ; e++ ))
for dir in $directories; do
echo "Searching in $dir" >> $log
find $dir -name "$filemask*" 1> fullnames 2> /dev/null
#cat fullnames
if [ -n "$(< fullnames)" ]; then
break;
fi
done
i=0
flag=0
echo
while read fullname; do
echo "Checking for possible duplicated file" >> $log
(( e = 0 ))
while (( e <= i )); do
echo "checking file $i : $fullname iteration number: $e" >> $log
if [ "$fullname" = "${fullnames_array[$e]}" ]; then
echo "$fullname number $i equals another, it's Discarded" >> $log
flag=1
else
echo "$fullname is unique!" >> $log
fi
(( e += 1 ))
done
if (( flag == 0 )); then
fullnames_array[$i]=$fullname
(( i += 1 ))
#echo ${fullnames_array[*]}
#echo "print $fullname"
fi
done < fullnames
echo "found files: ${fullnames_array[*]}" >> $log

I'm not sure why you want to check for a duplicated file, find won't return those.
And you can always use "sort -u" to get rid of dups.

[Nicolás]

Thanks dennis,
it looks better with a while, but what looks really better for me is "$(< fullnames)" insteed the horrible `cat $fullnames`.
in a deeper way, what are the differences between both?


you're right, find won't return me duplicated...

but i'll cut string routes before doing comparison

[Hein van den Heuvel]

Well, in the 'cat' case you fork a process to run the cat code which reads the file and pipe back line after line. In the $(< file) solution you tell the shell to just go read the file itself.... best I know.

hth,
Hein.

[Dennis Handly]

>it looks better with a while

The only problem with that while is that a continue won't work, it will skip the increment.

>what are the differences between both?

Besides what Hein said, it helps you in a catless style of scripting. You really only need cat(1) in a few cases, multiple inputs, 2 GB limitations or copying a file to stdout/stderr.