Find and replace confined to a specific column (one-liner)

Question

Is there a way (in a one-liner) to confine find and replace to within a specific column of a tab-delimited .txt file?

For example:

perl -pe 's/A/B/g;'

But which only finds and replaces 'A' with 'B' within column-2 and ignores all other instances of A, converting this:

Into:

Almost certainly. Sample data would help massively though. – Sobrique Jan 22 '16 at 22:26 — Sobrique, Jan 22 '16 at 22:26

Sobrique · Accepted Answer · 2016-01-22T22:40:21.097

2

Something like this will transform the second field of a tab separated row.

perl -lne '@row = split /\t/; $row[1] =~ s/A/B/; print join ( "\t", @row );'

edited Jan 22 '16 at 22:40

answered Jan 22 '16 at 22:30

Sobrique

52,974
7
60
101

This is returning `Substitution replacement not terminated at -e line 1.` What is causing that? – AnnaSchumann Jan 22 '16 at 22:35
A missing `/`. Edited – Sobrique Jan 22 '16 at 22:35
It now runs without error but only acts on the first line of the .txt file - or at least only replaces the first 'A' it finds in column-2. – AnnaSchumann Jan 22 '16 at 22:36
Sorry - issues with the .txt file caused by hidden characters. Now it works. – AnnaSchumann Jan 22 '16 at 22:39
Given you don't have any other whitespace, you could probably just `split;` which defaults to any whitespace. – Sobrique Jan 22 '16 at 22:40
2

You can shorten that up a bit by using the `-a` flag: `perl -anE '$F[1] =~ s/A/B/; say join "\t", @F' in.txt`. Search for it in [perlrun](http://perldoc.perl.org/perlrun.html) – stevieb Jan 22 '16 at 22:50
@stevieb Thanks! Is there a shorter way to perform multiple search/replacements other than simply chaining them together e.g. `perl -anE '$F[1] =~ s/A/B/; $F[1] =~ s/D/F/; say join "\t", @F'`? – AnnaSchumann Jan 22 '16 at 22:52
You could say `$F[1] =~ tr/AD/BF/;` – Chris Charley Jan 22 '16 at 23:10

score 2 · Answer 2 · answered Jan 23 '16 at 00:55

2

You can write this:

perl -lane '$F[1]=~tr/A/B/;print join("\t", @F)' file

answered Jan 23 '16 at 00:55

Casimir et Hippolyte

88,009
5
94
125

Benjamin W. · Answer 3 · 2016-01-23T08:27:27.490

A sed one-liner

$ sed 'h;s/^.*\t\(.*\)\t.*$/\1/;y/A/B/;G;s/\(.*\)\n\(.*\t\).*\(\t.*\)/\2\1\3/' infile
1       B       A2
2       B       A
3       B       B1
4       B       B
5       B       A
6       B       A3
7       B       A
8       C       A
9       E       B
10      D       B
11      B       C6
12      B       G
13      B       L
14      B       E
15      B       A
16      B       A

Explained:

# Copy pattern space to hold space
h

# Remove first and third column from pattern space
s/^.*\t\(.*\)\t.*$/\1/

# Transliterate A to B (can use s/// for more complex substitutions)
y/A/B/

# Append hold space to pattern space
G

# Replace second column with transliterated value
s/\(.*\)\n\(.*\t\).*\(\t.*\)/\2\1\3/

A sed-free alternative

$ paste <(cut -f 1 infile) <(cut -f 2 infile | tr A B) <(cut -f 3 infile)

This uses paste , cut and tr with process substitution. The tr can of course be replaced by a sed substitution for more complex changes.

Find and replace confined to a specific column (one-liner)

3 Answers3

A sed one-liner

A sed-free alternative

Linked