I was recently surprised to find out, that the POSIX
list
of utilities doesn't include the
column
utility. column first appeared way back in 4.3BSD, and
it's is pretty useful. Is there a POSIX equivalent of it?
The exact command I would like to replace is:
$ more -- ./tmp.txt 1 SMALL 000a 2 VERY_VERY_VERY_VERY_LONG 000b 3 SMALL 000c $ column -t -- ./tmp.txt 1 SMALL 000a 2 VERY_VERY_VERY_VERY_LONG 000b 3 SMALL 000c $
That is, expand tabs to create prettier columns.
POSIX prefers to codify existing behavior, and to only mandate new features or features that are not widely adopted when the existing behavior is unsatisfactory. Nowadays there isn't much call for presentation of data using unformatted text in fixed-width fonts, so column is unlikely to become mandatory. Unlike pr, it's from BSD, it isn't present System V and other historical unices, so it isn't grandfathered in.
Like any other text utility, you can express it in awk with a moderate amount of work. Here's a minimally tested implementation of column -t. Awk's -F option is similar to column's -s in simple cases (with a single character).
#!/usr/bin/env awk
{
if (max_column < NF) max_column = NF;
for (i = 1; i <= NF; i++) {
if (width[i] < length($i)) width[i] = length($i);
data[NR, i] = $i;
}
}
END {
for (i = 1; i < max_column; i++) format[i] = sprintf("%%-%ds ", width[i]);
format[max_column] = "%s\n";
for (k = 1; k <= NR; k++) {
for (i = 1; i <= max_column; i++) printf format[i], data[k, i];
}
}
sprintf with a tab, so that fullwidth characters (like the Han characters or emoji) don't screw up the alignment.
-v, do you mean -F? The -s option to column sets a delimiter, like -F; the -v option to awk sets an variable within the context of awk, at least on my system.
Commented
Dec 3, 2022 at 1:56
pr does a lot of what column does. If you specify an actual command line we can work on the replacement one.
NAME pr - convert text files for printing
SYNOPSIS pr [OPTION]... [FILE]...
DESCRIPTION Paginate or columnate FILE(s) for printing.
For your particular case, you could use awk:
awk '{printf("%2s %30s %10s", $1, $2, $3);}' file.txt
Note that the width specifiers 2, 30 and 10 are hardcoded, you have to choose the values for your input, or read the file twice (once to get the maximum widths for each column and the second time to print it out with the values you got).
column -t.
As far as I know there is no direct equivalent to column among the POSIX set of utilities.
A possible (cheap) POSIX solution could be:
$ tr -s ' ' '\n' <<"EOT" | pr -3 -a -t
1 SMALL 000a
2 VERY_VERY_VERY_VERY_LONG 000b
3 SMALL 000c
EOT
1 SMALL 000a
2 VERY_VERY_VERY_VERY_LON 000b
3 SMALL 000
After copying/pasting it, your sample data had fields separated by spaces. In case they were actually separated by a single TAB, you can use tr '\t' '\n'.
Of course this assumes that all the input lines have the same number of fields. The rather evident downside is that columns are evenly sized, which make them look significantly less nice than the output of column -t.
Also, as shown in the above sample, pr uses a fixed length for lines (defaulting to 72 characters), and truncates fields that don't fit into their column's width. It may be then necessary to explicitly set the page width (-w) to a suitable value.
