Very Basic: replace value flanked by tabs

snakefact · 12-14-2009, 05:29 PM

I should know how to do this, but I'm very new to unix.

I have a table of data, which is tab delimited. I need something, preferably sed, that will delete the 2nd value here, while preserving the placeholder:

3 1 1.0 0.10 0.1 1.0e-1 1.0010

to this

3 1.0 0.10 0.1 1.0e-1 1.0010

Thanks for helping a newb-

kbp · 12-14-2009, 05:36 PM

Give this a try:

Code:

sed -i 's/^\([[:digit:]]\{1,\}\t\)[[:digit:]]\{1,\}\t\(.*\)/\1\2/' /path/to/my/file

cheers

snakefact · 12-14-2009, 05:47 PM

Quote:

sed -i 's/^$[[:digit::]\{1,\}\t$[[:digit:]]\{1,\}\t$.*$/\1\2/g'

says unterminated command

pixellany · 12-14-2009, 06:06 PM

Code:

sed 's/\t[^\t]*\t/\t \t/1' filename > newfilename

Simple, but not very flexible---e.g it does not adapt well to changing any field.

Why not AWK? Simply print the fields you want, and print your placeholder (I used a space) for the ones you don't want.

PS: Have you read the material suggested in your earlier thread?

kbp · 12-14-2009, 06:18 PM

Edited my post above due to a typo, sorry.. didn't test it

And again! geez... you should always test !

Here's a working one:

Code:

sed 's/^\([[:digit:]]\{1,\}[[:blank:]]\)[[:digit:]]\{1,\}[[:blank:]]\(.*\)/\1\2/' ./myfile

Output:

Code:

sed 's/^\([[:digit:]]\{1,\}[[:blank:]]\)[[:digit:]]\{1,\}[[:blank:]]\(.*\)/\1\2/' ./test
3 1.0 0.10 0.1 1.0e-1 1.0010

snakefact · 12-14-2009, 06:32 PM

For some reason was unable to get either of the above suggestions to work.

However, this works except on values at the beginning or end of the line:

sed -r 's/(\t)1(\t)/\1\2/g'

I tried this to replace the ones the beginning, but it didn't work:

sed -r 's/(\t|^)1(\t)/\1\2/g'

Does anyone have a solution to this problem??

The reason why I'd prefer sed is because I know the syntax a little, whereas with awk, I don't know it at all. But if there's an easy awk solution, then by all means please share.

snakefact · 12-14-2009, 07:19 PM

Ended up using this:

sed -r 's/(\t)1(\t)/\1\2/g' | sed -r 's/^1(\t)/\1/g' | sed -r 's/(\t)1$/\1/g'

Tinkster · 12-14-2009, 07:55 PM

Note: I've actually inserted TABs where you had spaces.

Code:

$ cat tabs 
3       1       1.0     0.10    0.1     1.0e-1  1.0010
awk 'BEGIN{FS=OFS="\t"}{$2=""; print }' tabs
3               1.0     0.10    0.1     1.0e-1  1.0010

Cheers,
Tink

ghostdog74 · 12-14-2009, 08:07 PM

Quote:

Originally Posted by snakefact

The reason why I'd prefer sed is because I know the syntax a little, whereas with awk, I don't know it at all. But if there's an easy awk solution, then by all means please share.

you have seen the simpler and more readable awk solution, now start your learning journey. See my sig for gawk manual