Using sed instead of cut utility

[Danny Fang]

Hi,

I have a UNIX shell script which is to extract the timestamp values into their separate components such as the year, month, day, hour, min, secs.

The timestamp is denoted by the variable $timestamp below. The $timestamp denotes the UNIX epoch time since 1970.

timestamp=1159445287
From the $timestamp value, I've converted the UNIX timestamp into a format of year, month, day, hour, min, seconds into the variable
$newTimeToSecs.
newTimeToSecs=060922113366

Now, I'm attempting to extract values of the year, month, day, hour, min, seconds from the
variable "$newTimeToSecs" and place them into separate variables. I've used the 'cut' utility to extract them.

year=`/usr/bin/echo $newTimeToSecs |/usr/bin/cut -b1-2`
month=`/usr/bin/echo $newTimeToSecs |/usr/bin/cut -b3-4`
day=`/usr/bin/echo $newTimeToSecs |/usr/bin/cut -b5-6`
hour=`/usr/bin/echo $newTimeToSecs |/usr/bin/cut -b7-8`
MIN=`/usr/bin/echo $newTimeToSecs |/usr/bin/cut -b9-10`
secs=`/usr/bin/echo $newTimeToSecs |/usr/bin/cut -b11-12`

However, I'd would like to know how I could used the 'sed' utility to perform the same task.

year=`echo newTimeToSecs | sed -e 's/\([0-9]\{2\}\)\([0-9]\{2\}\)\([0-9]\{2\}\)\([0-9]\{2\}\)\([0-9]\{2\}\)\([0-9]\{2\}\)/\1/'`

month=`echo newTimeToSecs | sed -e 's/\([0-9]\{2\}\)\([0-9]\{2\}\)\([0-9]\{2\}\)\([0-9]\{2\}\)\([0-9]\{2\}\)\([0-9]\{2\}\)/\2/'`

day=`echo newTimeToSecs | sed -e 's/\([0-9]\{2\}\)\([0-9]\{2\}\)\([0-9]\{2\}\)\([0-9]\{2\}\)\([0-9]\{2\}\)\([0-9]\{2\}\)/\3/'`

hour=`echo newTimeToSecs | sed -e 's/\([0-9]\{2\}\)\([0-9]\{2\}\)\([0-9]\{2\}\)\([0-9]\{2\}\)\([0-9]\{2\}\)\([0-9]\{2\}\)/\4/'`

min=`echo newTimeToSecs | sed -e 's/\([0-9]\{2\}\)\([0-9]\{2\}\)\([0-9]\{2\}\)\([0-9]\{2\}\)\([0-9]\{2\}\)\([0-9]\{2\}\)/\5/'`

seconds=`echo newTimeToSecs | sed -e 's/\([0-9]\{2\}\)\([0-9]\{2\}\)\([0-9]\{2\}\)\([0-9]\{2\}\)\([0-9]\{2\}\)\([0-9]\{2\}\)/\6/'`

Could anyone show me a better way using 'sed' to achieve the same task?

Thanks

[Jonathan Fife]

There's no way for sed to assign multiple variables at once, so you're probably better off going with cut for readability. If you have your heart bent on using sed I would eliminate all the pattern string after the text you're looking for -- eg.

year=$(echo $newTimeToSecs | sed -e 's/^\([0-9]\{2\}\).*/\1/')
month=$(echo $newTimeToSecs | sed -e 's/[0-9]\{2\}\([0-9]\{2\}\).*/\1/')
etc.

[James R. Ferguson]

Hi:

For something like this, I'd use Perl or 'awk'. With Perl you could 'unpack' or use 'substr' to perform the extractions. With 'awk' you have 'substr'.

Otherwise, my choice would be to use 'cut'. It is the most natural corrollary.

Regards!

...JRF...

[Peter Godron]

Danny,
last few lines of section 36.23.4 in http://www.sm.luth.se/~alapaa/file_fetch/unixcdbookshelf/upt/ch36_23.htm seems to suggest you could assign values to multiple variables in one go, with eval.

I have never used this method.

I tend to choose the easiest path to a solution, which in this case seems to be cut.

[Peter Nikitka]

Hi,

what about

echo $newTimeToSecs | sed 's/\(..\)/\1 /g' | read year month day hour min seconds

This will not work in a (Linux-)Pdksh or bash, but in the HP-Posix shell and other KSHs.

mfG Peter

[OldSchool]

the easy way is:

date +'%y %m %d %H %M %S' | read yr mo da hr mn sec

change %H to %I for 12hr clock

[OldSchool]

disregard the previous. i didn't note that you weren't creating the stamps, but working w/ existing....more coffed please

[Sandman!]

How about the following:

year=`echo $newTimeToSecs | sed 's/^\(.\{2\}\).*$/\1/p'`
month=`echo $newTimeToSecs | sed 's/^..\(.\{2\}\).*$/\1/p'`
day=`echo $newTimeToSecs | sed 's/^....\(.\{2\}\).*$/\1/p'`
hour=`echo $newTimeToSecs | sed 's/^.\{6\}\(.\{2\}\).*$/\1/p'`
min=`echo $newTimeToSecs | sed 's/^.\{8\}\(.\{2\}\).*$/\1/p'`
seconds=`echo $newTimeToSecs | sed 's/^.\{10\}\(.\{2\}\)$/\1/p'`

[Arturo Galbiati]

Hi,
on my Hp-UX 11i:
% newTimeToSecs=060922113366
%>echo $newTimeToSecs | sed 's/\(..\)/\1 /g' | read yy mm dd hh mm ss
% echo $yy $mm $dd $hh $mm $ss

HTH,
Art

[Danny Fang]

HI Sandman,

I attempted your suggested method but I noticed it duplicated the values being output.

mz$ echo 1159528031 | sed 's/^..\(.\{2\}\).*$/\1/p'
59
59
mz$ echo 1159528031 |sed 's/^....\(.\{2\}\).*$/\1/p'
52
52
mz$ echo 1159528031 |sed 's/^.\{6\}\(.\{2\}\).*$/\1/p'
80
80
mz$ echo 1159528031 |sed 's/^\(.\{2\}\).*$/\1/p'
11
11
mz$ echo 1159528031|sed 's/^.\{8\}\(.\{2\}\).*$/\1/p'
31
31

Could you point out where did I go wrong?

Thanks
Danny

[Peter Nikitka]

Hi,

did you try my solution?

mfG Peter

[Sandman!]

>HI Sandman,

>I attempted your suggested method but I noticed it duplicated the values being >output.

>mz$ echo 1159528031 | sed 's/^..\(.\{2\}\).*$/\1/p'
>59
>59

Provide the "-n" switch to all the sed commands for removing duplicates i.e.

mz$ echo 1159528031 | sed -n 's/^..\(.\{2\}\).*$/\1/p'

[Peter Nikitka]

Hi,

in Sandmans solution you can just drop the 'p(rint)' directive AND the '-n' option:
Use
echo 1159528031 | sed 's/^..\(.\{2\}\).*$/\1/'
59

mfG Peter