One may want to run grep-like operation on a specific column.

For example, when I need the lines from /home/me/test.tsv where the second column equals to a:

1	a	2.1
1	b	2.2
2	a	2.5
2	c	2.6

grep does not come with an option to specify column. I don't want to break the lines into pieces or put pieces back, that's too much pain.

The trick is to use join. It performs join (recall relational algebra) on lines of 2 input files.


% join -1 2 -2 1 -o 1.1,1.2,1.3 -t $'\t' <(sort test --key=2) <(echo a)
1 a 2.1
2 a 2.5

Arguments explained:

  • -1, -2: which column to join on
  • -o: which columns to output (I had to specify all columns explicitly)
  • -t: column delimiter of input/output
  • <(sort test --key=2): lines sorted by 2nd column
  • <(echo a): just a

BTW this join solution may be to some degree cleaner, but not much shorter than a perl one-liner:

% perl -ne '@cs=split;$cs[1] eq "a" and print;' < test

and awk code is even shorter:

% awk '$2=="a"' < test