How to check the total number of columns in a text file

cliffyao · 11-30-2009, 03:57 PM

Hi, all

I have a .txt file which has many many tab-delimited columns. I wonder if there is any command like "wc -l" that can tell me the total number of columns in this file.

All your help will be greatly appreciated.

-Cliff

acid_kewpie · 11-30-2009, 04:14 PM

awk -F\t 'BEGIN{print NF}' file.txt

should do it.

rweaver · 11-30-2009, 04:15 PM

This could be more cleanly done in perl, but this method has pretty low overhead.

Code:

head -n1 test |  sed 's/\t/\n/g' | wc -l

Could also be done using a single awk/sed line, but I don't have time to test at the moment.

pixellany · 11-30-2009, 04:27 PM

Could not find something in less time than it took to write it:

linenum = the number of the line to test

Code:

sed -n "${linenum}s/\t/\n/gp" | wc -l <filename >newfilename

subtract 1 to get the number of tabs

<<EDIT: Other answers here are better>>

colucix · 11-30-2009, 06:33 PM

My

using awk:

Code:

awk '{print split($0,a,"\t"); exit}' testfile

ghostdog74 · 11-30-2009, 06:38 PM

with awk,

Code:

awk -F"\t" '{print NF;exit}' file

or the shell

Code:

$ exec 4<"file"
$ read -r line <&4
$ IFS=$'\t'
$ set -- $line
$ echo "Number of columns: $#"
$ exec >&4-

ghostdog74 · 11-30-2009, 06:47 PM

Quote:

Originally Posted by acid_kewpie

awk -F\t 'BEGIN{print NF}' file.txt

should do it.

NF is not yet initialized in BEGIN block

acid_kewpie · 12-01-2009, 12:30 AM

No, of course it's not. My bad. I just didn't want to print it for every line, thought it'd just take the first, but sure, exiting after the first line is clearly the way to do that.

colucix · 12-01-2009, 04:42 PM

Quote:

Originally Posted by ghostdog74

with awk,

Code:

awk -F"\t" '{print NF;exit}' file

The number of fields does not reflect the fact you can have null values in a column. Suppose in the first row one or more values are empty and you have two consecutive tabs: awk and many other commands/languages ignore them and count the actual number of fields. In this way you don't retrieve the actual number of columns.

On the other hand, if the file has not empty (null) values and all columns are filled up, we can simply use:

Code:

head -1 testfile | wc -w

chrism01 · 12-01-2009, 04:49 PM

That's a possibility, but only if the cols are single valued ... if not you have to count the tab chars.