5

Sample File

a.b.gg3d 0000000001 0000000001 y
a.b.112a 0000000000 0000000001 y
a.b.gg3s 0000000001 0000000001 y
a.b.1111 0000000000 0000000001 m

I have a file that includes about three thousand unique lines, similar to the sample above. I want to use sed to remove the "xxxxxxxxxx xxxxxxxxxx" numbers from each line. Using the sample above, I would like the sed script to produce the following output:

a.b.gg3d y
a.b.112a y
a.b.gg3s y
a.b.1111 m

How can I accomplish this? Is there a string of commands to tell sed something like "delete all numbers after the first whitespace, but leave the last character"?

Improve this question

4 Answers 4

Reset to default
4

Given this input, you want to keep the first and last fields. Pretty simple with awk:

awk '{print $1, $NF}' filename

Using sed, this will replace all space delimited digit-only words:

sed ':a; s/ [[:digit:]]\+ / /; ta'
Improve this answer
6
  • I understand that the s means substitute and the forward slashes are delimiters, but what do the following commands mean? :a; \+ and ; ta
    – user46865
    Commented Sep 10, 2013 at 20:19
  • :a is a label, ta means "go to a": in short, it is a loop
    – enzotib
    Commented Sep 10, 2013 at 20:31
  • 1
    And sed 's/ [[:digit:] ]\+ / /' can do it the same
    – enzotib
    Commented Sep 10, 2013 at 20:34
  • 1
    I was thinking along the lines of enzotib, but went with sed 's/[[:space:]][[:digit:]]\+//g' which removes one or more numbers [[:digit:]]\+ proceeded by whitespace (tabs, spaces) [:space:], and also specified /g to the options, to do it for all occurrences (and not just the first)
    – Drav Sloan
    Commented Sep 10, 2013 at 20:41
  • The reason I searched for the trailing space in a loop was to avoid matching " 123abc" -- i.e. to ensure that the word was composed only with digits.
    – glenn jackman
    Commented Sep 11, 2013 at 12:41
1

Another solution only using cut, that will print the first and fourth column:

cut -d' ' -f1,4 < samplefile.txt
Improve this answer
1
  • 1
    By far the most appropriate solution IMO. If all that needs to be done is to remove columns that are well defined by their delimiters, than cut is the right tool for the job. However, the -1 should be just 1 as there are no columns to the left of the first one.
    – Kusalananda ♦
    Commented Feb 16, 2017 at 8:52
0 
sed -e 's/[ ].*[ ]/ /' yourfile

This looks at the leftmost space and up till the rightmost space and replaces this with a single space. IOW, first & last fields are retained.

Assuming that there are no leading &/or trailing spaces. There are no TABs.

Otherwise, we can always do the following:

SPC=`echo 'x' | tr 'x' '\040'`
TAB=`echo 'x' | tr 'x' '\011'`
WS="[$TAB$SPC]"
sed -e "s/^$WS*//;s/${WS}\$//;s/${WS}.*${WS}/${SPC}/" < yourfile
Improve this answer
0 
$ echo "a.b.gg3d 0000000001 0000000001 y" |sed -r 's/[ ]+[0-9]+//g'
a.b.gg3d y

-r regex mode

You're searching for one or more spaces, followed by one or more numbers of any in range 0-9. Technically this is matching twice, leaving your third space in between a.b.gg3d and y.

Improve this answer

You must log in to answer this question.